Add benchmark tool

.gitignore

@@ -21,6 +21,8 @@ test/test.xcodeproj/*/xcuser*
 test/*.o
 test/*.pem
 test/*.srl
+benchmark/server
+benchmark/server-crow
 
 *.swp
 
@@ -34,6 +36,7 @@ Release
 *.db
 ipch
 *.dSYM
+*.pyc
 .*
 !/.gitattributes
 !/.travis.yml

benchmark/Makefile

@@ -0,0 +1,31 @@
+CXXFLAGS = -std=c++14 -O2 -I..
+
+THEAD_POOL_COUNT = 16
+BENCH_FLAGS = -c 8 -d 5s
+
+# cpp-httplib
+bench: server
+	@./server & export PID=$$!; bombardier $(BENCH_FLAGS) localhost:8080; kill $${PID}
+
+server : cpp-httplib/main.cpp ../httplib.h
+	g++ -o $@ $(CXXFLAGS) -DCPPHTTPLIB_THREAD_POOL_COUNT=$(THEAD_POOL_COUNT) cpp-httplib/main.cpp
+
+run : server
+	@./server
+
+# crow
+server-crow : crow/main.cpp
+	g++ -o $@ $(CXXFLAGS) crow/main.cpp
+
+bench-crow: server-crow
+	@./server-crow & export PID=$$!; bombardier $(BENCH_FLAGS) localhost:8080; kill $${PID}
+
+# flask
+bench-flask:
+	@FLASK_APP=flask/main.py flask run --port=8080 & export PID=$$!; bombardier $(BENCH_FLAGS) localhost:8080; kill $${PID}
+
+# misc
+bench-all: bench bench-crow bench-flask
+
+clean:
+	rm -rf server*

benchmark/cpp-httplib/main.cpp

@@ -0,0 +1,12 @@
+#include "httplib.h"
+using namespace httplib;
+
+int main() {
+  Server svr;
+
+  svr.Get("/", [](const Request &, Response &res) {
+    res.set_content("Hello World!", "text/plain");
+  });
+
+  svr.listen("0.0.0.0", 8080);
+}

benchmark/crow/crow_all.h

@@ -0,0 +1,14316 @@
+// SPDX-License-Identifier: BSD-3-Clause AND ISC AND MIT
+/*BSD 3-Clause License
+
+Copyright (c) 2014-2017, ipkn
+              2020-2022, CrowCpp
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
+
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+
+* Neither the name of the author nor the names of its
+  contributors may be used to endorse or promote products derived from
+  this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+The Crow logo and other graphic material (excluding third party logos) used are under exclusive Copyright (c) 2021-2022, Farook Al-Sammarraie (The-EDev), All rights reserved.
+*/
+#pragma once
+// This file is generated from nginx/conf/mime.types using nginx_mime2cpp.py on 2021-12-03.
+#include <unordered_map>
+#include <string>
+
+namespace crow
+{
+    const std::unordered_map<std::string, std::string> mime_types{
+      {"shtml", "text/html"},
+      {"htm", "text/html"},
+      {"html", "text/html"},
+      {"css", "text/css"},
+      {"xml", "text/xml"},
+      {"gif", "image/gif"},
+      {"jpg", "image/jpeg"},
+      {"jpeg", "image/jpeg"},
+      {"js", "application/javascript"},
+      {"atom", "application/atom+xml"},
+      {"rss", "application/rss+xml"},
+      {"mml", "text/mathml"},
+      {"txt", "text/plain"},
+      {"jad", "text/vnd.sun.j2me.app-descriptor"},
+      {"wml", "text/vnd.wap.wml"},
+      {"htc", "text/x-component"},
+      {"avif", "image/avif"},
+      {"png", "image/png"},
+      {"svgz", "image/svg+xml"},
+      {"svg", "image/svg+xml"},
+      {"tiff", "image/tiff"},
+      {"tif", "image/tiff"},
+      {"wbmp", "image/vnd.wap.wbmp"},
+      {"webp", "image/webp"},
+      {"ico", "image/x-icon"},
+      {"jng", "image/x-jng"},
+      {"bmp", "image/x-ms-bmp"},
+      {"woff", "font/woff"},
+      {"woff2", "font/woff2"},
+      {"ear", "application/java-archive"},
+      {"war", "application/java-archive"},
+      {"jar", "application/java-archive"},
+      {"json", "application/json"},
+      {"hqx", "application/mac-binhex40"},
+      {"doc", "application/msword"},
+      {"pdf", "application/pdf"},
+      {"ai", "application/postscript"},
+      {"eps", "application/postscript"},
+      {"ps", "application/postscript"},
+      {"rtf", "application/rtf"},
+      {"m3u8", "application/vnd.apple.mpegurl"},
+      {"kml", "application/vnd.google-earth.kml+xml"},
+      {"kmz", "application/vnd.google-earth.kmz"},
+      {"xls", "application/vnd.ms-excel"},
+      {"eot", "application/vnd.ms-fontobject"},
+      {"ppt", "application/vnd.ms-powerpoint"},
+      {"odg", "application/vnd.oasis.opendocument.graphics"},
+      {"odp", "application/vnd.oasis.opendocument.presentation"},
+      {"ods", "application/vnd.oasis.opendocument.spreadsheet"},
+      {"odt", "application/vnd.oasis.opendocument.text"},
+      {"pptx", "application/vnd.openxmlformats-officedocument.presentationml.presentation"},
+      {"xlsx", "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"},
+      {"docx", "application/vnd.openxmlformats-officedocument.wordprocessingml.document"},
+      {"wmlc", "application/vnd.wap.wmlc"},
+      {"wasm", "application/wasm"},
+      {"7z", "application/x-7z-compressed"},
+      {"cco", "application/x-cocoa"},
+      {"jardiff", "application/x-java-archive-diff"},
+      {"jnlp", "application/x-java-jnlp-file"},
+      {"run", "application/x-makeself"},
+      {"pm", "application/x-perl"},
+      {"pl", "application/x-perl"},
+      {"pdb", "application/x-pilot"},
+      {"prc", "application/x-pilot"},
+      {"rar", "application/x-rar-compressed"},
+      {"rpm", "application/x-redhat-package-manager"},
+      {"sea", "application/x-sea"},
+      {"swf", "application/x-shockwave-flash"},
+      {"sit", "application/x-stuffit"},
+      {"tk", "application/x-tcl"},
+      {"tcl", "application/x-tcl"},
+      {"crt", "application/x-x509-ca-cert"},
+      {"pem", "application/x-x509-ca-cert"},
+      {"der", "application/x-x509-ca-cert"},
+      {"xpi", "application/x-xpinstall"},
+      {"xhtml", "application/xhtml+xml"},
+      {"xspf", "application/xspf+xml"},
+      {"zip", "application/zip"},
+      {"dll", "application/octet-stream"},
+      {"exe", "application/octet-stream"},
+      {"bin", "application/octet-stream"},
+      {"deb", "application/octet-stream"},
+      {"dmg", "application/octet-stream"},
+      {"img", "application/octet-stream"},
+      {"iso", "application/octet-stream"},
+      {"msm", "application/octet-stream"},
+      {"msp", "application/octet-stream"},
+      {"msi", "application/octet-stream"},
+      {"kar", "audio/midi"},
+      {"midi", "audio/midi"},
+      {"mid", "audio/midi"},
+      {"mp3", "audio/mpeg"},
+      {"ogg", "audio/ogg"},
+      {"m4a", "audio/x-m4a"},
+      {"ra", "audio/x-realaudio"},
+      {"3gp", "video/3gpp"},
+      {"3gpp", "video/3gpp"},
+      {"ts", "video/mp2t"},
+      {"mp4", "video/mp4"},
+      {"mpg", "video/mpeg"},
+      {"mpeg", "video/mpeg"},
+      {"mov", "video/quicktime"},
+      {"webm", "video/webm"},
+      {"flv", "video/x-flv"},
+      {"m4v", "video/x-m4v"},
+      {"mng", "video/x-mng"},
+      {"asf", "video/x-ms-asf"},
+      {"asx", "video/x-ms-asf"},
+      {"wmv", "video/x-ms-wmv"},
+      {"avi", "video/x-msvideo"}};
+}
+
+
+#include <string>
+
+namespace crow
+{
+    /// An abstract class that allows any other class to be returned by a handler.
+    struct returnable
+    {
+        std::string content_type;
+        virtual std::string dump() const = 0;
+
+        returnable(std::string ctype):
+          content_type{ctype}
+        {}
+
+        virtual ~returnable(){};
+    };
+} // namespace crow
+
+
+#include <stdio.h>
+#include <string.h>
+#include <string>
+#include <vector>
+#include <unordered_map>
+#include <iostream>
+#include <memory>
+
+namespace crow
+{
+
+// ----------------------------------------------------------------------------
+// qs_parse (modified)
+// https://github.com/bartgrantham/qs_parse
+// ----------------------------------------------------------------------------
+/*  Similar to strncmp, but handles URL-encoding for either string  */
+int qs_strncmp(const char* s, const char* qs, size_t n);
+
+
+/*  Finds the beginning of each key/value pair and stores a pointer in qs_kv.
+ *  Also decodes the value portion of the k/v pair *in-place*.  In a future
+ *  enhancement it will also have a compile-time option of sorting qs_kv
+ *  alphabetically by key.  */
+size_t qs_parse(char* qs, char* qs_kv[], size_t qs_kv_size, bool parse_url);
+
+
+/*  Used by qs_parse to decode the value portion of a k/v pair  */
+int qs_decode(char * qs);
+
+
+/*  Looks up the value according to the key on a pre-processed query string
+ *  A future enhancement will be a compile-time option to look up the key
+ *  in a pre-sorted qs_kv array via a binary search.  */
+//char * qs_k2v(const char * key, char * qs_kv[], int qs_kv_size);
+ char * qs_k2v(const char * key, char * const * qs_kv, size_t qs_kv_size, int nth);
+
+
+/*  Non-destructive lookup of value, based on key.  User provides the
+ *  destinaton string and length.  */
+char * qs_scanvalue(const char * key, const char * qs, char * val, size_t val_len);
+
+// TODO: implement sorting of the qs_kv array; for now ensure it's not compiled
+#undef _qsSORTING
+
+// isxdigit _is_ available in <ctype.h>, but let's avoid another header instead
+#define CROW_QS_ISHEX(x)    ((((x)>='0'&&(x)<='9') || ((x)>='A'&&(x)<='F') || ((x)>='a'&&(x)<='f')) ? 1 : 0)
+#define CROW_QS_HEX2DEC(x)  (((x)>='0'&&(x)<='9') ? (x)-48 : ((x)>='A'&&(x)<='F') ? (x)-55 : ((x)>='a'&&(x)<='f') ? (x)-87 : 0)
+#define CROW_QS_ISQSCHR(x) ((((x)=='=')||((x)=='#')||((x)=='&')||((x)=='\0')) ? 0 : 1)
+
+inline int qs_strncmp(const char * s, const char * qs, size_t n)
+{
+    unsigned char u1, u2, unyb, lnyb;
+
+    while(n-- > 0)
+    {
+        u1 = static_cast<unsigned char>(*s++);
+        u2 = static_cast<unsigned char>(*qs++);
+
+        if ( ! CROW_QS_ISQSCHR(u1) ) {  u1 = '\0';  }
+        if ( ! CROW_QS_ISQSCHR(u2) ) {  u2 = '\0';  }
+
+        if ( u1 == '+' ) {  u1 = ' ';  }
+        if ( u1 == '%' ) // easier/safer than scanf
+        {
+            unyb = static_cast<unsigned char>(*s++);
+            lnyb = static_cast<unsigned char>(*s++);
+            if ( CROW_QS_ISHEX(unyb) && CROW_QS_ISHEX(lnyb) )
+                u1 = (CROW_QS_HEX2DEC(unyb) * 16) + CROW_QS_HEX2DEC(lnyb);
+            else
+                u1 = '\0';
+        }
+
+        if ( u2 == '+' ) {  u2 = ' ';  }
+        if ( u2 == '%' ) // easier/safer than scanf
+        {
+            unyb = static_cast<unsigned char>(*qs++);
+            lnyb = static_cast<unsigned char>(*qs++);
+            if ( CROW_QS_ISHEX(unyb) && CROW_QS_ISHEX(lnyb) )
+                u2 = (CROW_QS_HEX2DEC(unyb) * 16) + CROW_QS_HEX2DEC(lnyb);
+            else
+                u2 = '\0';
+        }
+
+        if ( u1 != u2 )
+            return u1 - u2;
+        if ( u1 == '\0' )
+            return 0;
+    }
+    if ( CROW_QS_ISQSCHR(*qs) )
+        return -1;
+    else
+        return 0;
+}
+
+
+inline size_t qs_parse(char* qs, char* qs_kv[], size_t qs_kv_size, bool parse_url = true)
+{
+    size_t i, j;
+    char * substr_ptr;
+
+    for(i=0; i<qs_kv_size; i++)  qs_kv[i] = NULL;
+
+    // find the beginning of the k/v substrings or the fragment
+    substr_ptr = parse_url ? qs + strcspn(qs, "?#") : qs;
+    if (parse_url)
+    {
+        if (substr_ptr[0] != '\0')
+            substr_ptr++;
+        else
+            return 0; // no query or fragment
+    }
+
+    i=0;
+    while(i<qs_kv_size)
+    {
+        qs_kv[i] = substr_ptr;
+        j = strcspn(substr_ptr, "&");
+        if ( substr_ptr[j] == '\0' ) { i++; break;  } // x &'s -> means x iterations of this loop -> means *x+1* k/v pairs
+        substr_ptr += j + 1;
+        i++;
+    }
+
+    // we only decode the values in place, the keys could have '='s in them
+    // which will hose our ability to distinguish keys from values later
+    for(j=0; j<i; j++)
+    {
+        substr_ptr = qs_kv[j] + strcspn(qs_kv[j], "=&#");
+        if ( substr_ptr[0] == '&' || substr_ptr[0] == '\0')  // blank value: skip decoding
+            substr_ptr[0] = '\0';
+        else
+            qs_decode(++substr_ptr);
+    }
+
+#ifdef _qsSORTING
+// TODO: qsort qs_kv, using qs_strncmp() for the comparison
+#endif
+
+    return i;
+}
+
+
+inline int qs_decode(char * qs)
+{
+    int i=0, j=0;
+
+    while( CROW_QS_ISQSCHR(qs[j]) )
+    {
+        if ( qs[j] == '+' ) {  qs[i] = ' ';  }
+        else if ( qs[j] == '%' ) // easier/safer than scanf
+        {
+            if ( ! CROW_QS_ISHEX(qs[j+1]) || ! CROW_QS_ISHEX(qs[j+2]) )
+            {
+                qs[i] = '\0';
+                return i;
+            }
+            qs[i] = (CROW_QS_HEX2DEC(qs[j+1]) * 16) + CROW_QS_HEX2DEC(qs[j+2]);
+            j+=2;
+        }
+        else
+        {
+            qs[i] = qs[j];
+        }
+        i++;  j++;
+    }
+    qs[i] = '\0';
+
+    return i;
+}
+
+
+inline char * qs_k2v(const char * key, char * const * qs_kv, size_t qs_kv_size, int nth = 0)
+{
+    size_t i;
+    size_t key_len, skip;
+
+    key_len = strlen(key);
+
+#ifdef _qsSORTING
+// TODO: binary search for key in the sorted qs_kv
+#else  // _qsSORTING
+    for(i=0; i<qs_kv_size; i++)
+    {
+        // we rely on the unambiguous '=' to find the value in our k/v pair
+        if ( qs_strncmp(key, qs_kv[i], key_len) == 0 )
+        {
+            skip = strcspn(qs_kv[i], "=");
+            if ( qs_kv[i][skip] == '=' )
+                skip++;
+            // return (zero-char value) ? ptr to trailing '\0' : ptr to value
+            if(nth == 0)
+                return qs_kv[i] + skip;
+            else
+                --nth;
+        }
+    }
+#endif  // _qsSORTING
+
+    return nullptr;
+}
+
+inline std::unique_ptr<std::pair<std::string, std::string>> qs_dict_name2kv(const char * dict_name, char * const * qs_kv, size_t qs_kv_size, int nth = 0)
+{
+    size_t i;
+    size_t name_len, skip_to_eq, skip_to_brace_open, skip_to_brace_close;
+
+    name_len = strlen(dict_name);
+
+#ifdef _qsSORTING
+// TODO: binary search for key in the sorted qs_kv
+#else  // _qsSORTING
+    for(i=0; i<qs_kv_size; i++)
+    {
+        if ( strncmp(dict_name, qs_kv[i], name_len) == 0 )
+        {
+            skip_to_eq = strcspn(qs_kv[i], "=");
+            if ( qs_kv[i][skip_to_eq] == '=' )
+                skip_to_eq++;
+            skip_to_brace_open = strcspn(qs_kv[i], "[");
+            if ( qs_kv[i][skip_to_brace_open] == '[' )
+                skip_to_brace_open++;
+            skip_to_brace_close = strcspn(qs_kv[i], "]");
+
+            if ( skip_to_brace_open <= skip_to_brace_close &&
+                 skip_to_brace_open > 0 &&
+                 skip_to_brace_close > 0 &&
+                 nth == 0 )
+            {
+                auto key = std::string(qs_kv[i] + skip_to_brace_open, skip_to_brace_close - skip_to_brace_open);
+                auto value = std::string(qs_kv[i] + skip_to_eq);
+                return std::unique_ptr<std::pair<std::string, std::string>>(new std::pair<std::string, std::string>(key, value));
+            }
+            else
+            {
+                --nth;
+            }
+        }
+    }
+#endif  // _qsSORTING
+
+    return nullptr;
+}
+
+
+inline char * qs_scanvalue(const char * key, const char * qs, char * val, size_t val_len)
+{
+    size_t i, key_len;
+    const char * tmp;
+
+    // find the beginning of the k/v substrings
+    if ( (tmp = strchr(qs, '?')) != NULL )
+        qs = tmp + 1;
+
+    key_len = strlen(key);
+    while(qs[0] != '#' && qs[0] != '\0')
+    {
+        if ( qs_strncmp(key, qs, key_len) == 0 )
+            break;
+        qs += strcspn(qs, "&") + 1;
+    }
+
+    if ( qs[0] == '\0' ) return NULL;
+
+    qs += strcspn(qs, "=&#");
+    if ( qs[0] == '=' )
+    {
+        qs++;
+        i = strcspn(qs, "&=#");
+#ifdef _MSC_VER
+        strncpy_s(val, val_len, qs, (val_len - 1)<(i + 1) ? (val_len - 1) : (i + 1));
+#else
+        strncpy(val, qs, (val_len - 1)<(i + 1) ? (val_len - 1) : (i + 1));
+#endif
+		qs_decode(val);
+    }
+    else
+    {
+        if ( val_len > 0 )
+            val[0] = '\0';
+    }
+
+    return val;
+}
+}
+// ----------------------------------------------------------------------------
+
+
+namespace crow
+{
+    struct request;
+    /// A class to represent any data coming after the `?` in the request URL into key-value pairs.
+    class query_string
+    {
+    public:
+        static const int MAX_KEY_VALUE_PAIRS_COUNT = 256;
+
+        query_string() = default;
+
+        query_string(const query_string& qs):
+          url_(qs.url_)
+        {
+            for (auto p : qs.key_value_pairs_)
+            {
+                key_value_pairs_.push_back((char*)(p - qs.url_.c_str() + url_.c_str()));
+            }
+        }
+
+        query_string& operator=(const query_string& qs)
+        {
+            url_ = qs.url_;
+            key_value_pairs_.clear();
+            for (auto p : qs.key_value_pairs_)
+            {
+                key_value_pairs_.push_back((char*)(p - qs.url_.c_str() + url_.c_str()));
+            }
+            return *this;
+        }
+
+        query_string& operator=(query_string&& qs) noexcept
+        {
+            key_value_pairs_ = std::move(qs.key_value_pairs_);
+            char* old_data = (char*)qs.url_.c_str();
+            url_ = std::move(qs.url_);
+            for (auto& p : key_value_pairs_)
+            {
+                p += (char*)url_.c_str() - old_data;
+            }
+            return *this;
+        }
+
+
+        query_string(std::string params, bool url = true):
+          url_(std::move(params))
+        {
+            if (url_.empty())
+                return;
+
+            key_value_pairs_.resize(MAX_KEY_VALUE_PAIRS_COUNT);
+            size_t count = qs_parse(&url_[0], &key_value_pairs_[0], MAX_KEY_VALUE_PAIRS_COUNT, url);
+
+            key_value_pairs_.resize(count);
+            key_value_pairs_.shrink_to_fit();
+        }
+
+        void clear()
+        {
+            key_value_pairs_.clear();
+            url_.clear();
+        }
+
+        friend std::ostream& operator<<(std::ostream& os, const query_string& qs)
+        {
+            os << "[ ";
+            for (size_t i = 0; i < qs.key_value_pairs_.size(); ++i)
+            {
+                if (i)
+                    os << ", ";
+                os << qs.key_value_pairs_[i];
+            }
+            os << " ]";
+            return os;
+        }
+
+        /// Get a value from a name, used for `?name=value`.
+
+        ///
+        /// Note: this method returns the value of the first occurrence of the key only, to return all occurrences, see \ref get_list().
+        char* get(const std::string& name) const
+        {
+            char* ret = qs_k2v(name.c_str(), key_value_pairs_.data(), key_value_pairs_.size());
+            return ret;
+        }
+
+        /// Works similar to \ref get() except it removes the item from the query string.
+        char* pop(const std::string& name)
+        {
+            char* ret = get(name);
+            if (ret != nullptr)
+            {
+                for (unsigned int i = 0; i < key_value_pairs_.size(); i++)
+                {
+                    std::string str_item(key_value_pairs_[i]);
+                    if (str_item.substr(0, name.size() + 1) == name + '=')
+                    {
+                        key_value_pairs_.erase(key_value_pairs_.begin() + i);
+                        break;
+                    }
+                }
+            }
+            return ret;
+        }
+
+        /// Returns a list of values, passed as `?name[]=value1&name[]=value2&...name[]=valuen` with n being the size of the list.
+
+        ///
+        /// Note: Square brackets in the above example are controlled by `use_brackets` boolean (true by default). If set to false, the example becomes `?name=value1,name=value2...name=valuen`
+        std::vector<char*> get_list(const std::string& name, bool use_brackets = true) const
+        {
+            std::vector<char*> ret;
+            std::string plus = name + (use_brackets ? "[]" : "");
+            char* element = nullptr;
+
+            int count = 0;
+            while (1)
+            {
+                element = qs_k2v(plus.c_str(), key_value_pairs_.data(), key_value_pairs_.size(), count++);
+                if (!element)
+                    break;
+                ret.push_back(element);
+            }
+            return ret;
+        }
+
+        /// Similar to \ref get_list() but it removes the
+        std::vector<char*> pop_list(const std::string& name, bool use_brackets = true)
+        {
+            std::vector<char*> ret = get_list(name, use_brackets);
+            if (!ret.empty())
+            {
+                for (unsigned int i = 0; i < key_value_pairs_.size(); i++)
+                {
+                    std::string str_item(key_value_pairs_[i]);
+                    if ((use_brackets ? (str_item.substr(0, name.size() + 3) == name + "[]=") : (str_item.substr(0, name.size() + 1) == name + '=')))
+                    {
+                        key_value_pairs_.erase(key_value_pairs_.begin() + i--);
+                    }
+                }
+            }
+            return ret;
+        }
+
+        /// Works similar to \ref get_list() except the brackets are mandatory must not be empty.
+
+        ///
+        /// For example calling `get_dict(yourname)` on `?yourname[sub1]=42&yourname[sub2]=84` would give a map containing `{sub1 : 42, sub2 : 84}`.
+        ///
+        /// if your query string has both empty brackets and ones with a key inside, use pop_list() to get all the values without a key before running this method.
+        std::unordered_map<std::string, std::string> get_dict(const std::string& name) const
+        {
+            std::unordered_map<std::string, std::string> ret;
+
+            int count = 0;
+            while (1)
+            {
+                if (auto element = qs_dict_name2kv(name.c_str(), key_value_pairs_.data(), key_value_pairs_.size(), count++))
+                    ret.insert(*element);
+                else
+                    break;
+            }
+            return ret;
+        }
+
+        /// Works the same as \ref get_dict() but removes the values from the query string.
+        std::unordered_map<std::string, std::string> pop_dict(const std::string& name)
+        {
+            std::unordered_map<std::string, std::string> ret = get_dict(name);
+            if (!ret.empty())
+            {
+                for (unsigned int i = 0; i < key_value_pairs_.size(); i++)
+                {
+                    std::string str_item(key_value_pairs_[i]);
+                    if (str_item.substr(0, name.size() + 1) == name + '[')
+                    {
+                        key_value_pairs_.erase(key_value_pairs_.begin() + i--);
+                    }
+                }
+            }
+            return ret;
+        }
+
+        std::vector<std::string> keys() const
+        {
+            std::vector<std::string> keys;
+            keys.reserve(key_value_pairs_.size());
+
+            for (const char* const element : key_value_pairs_)
+            {
+                const char* delimiter = strchr(element, '=');
+                if (delimiter)
+                    keys.emplace_back(element, delimiter);
+                else
+                    keys.emplace_back(element);
+            }
+
+            return keys;
+        }
+
+    private:
+        std::string url_;
+        std::vector<char*> key_value_pairs_;
+    };
+
+} // namespace crow
+
+#ifdef CROW_ENABLE_COMPRESSION
+
+#include <string>
+#include <zlib.h>
+
+// http://zlib.net/manual.html
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+    namespace compression
+    {
+        // Values used in the 'windowBits' parameter for deflateInit2.
+        enum algorithm
+        {
+            // 15 is the default value for deflate
+            DEFLATE = 15,
+            // windowBits can also be greater than 15 for optional gzip encoding.
+            // Add 16 to windowBits to write a simple gzip header and trailer around the compressed data instead of a zlib wrapper.
+            GZIP = 15 | 16,
+        };
+
+        inline std::string compress_string(std::string const& str, algorithm algo)
+        {
+            std::string compressed_str;
+            z_stream stream{};
+            // Initialize with the default values
+            if (::deflateInit2(&stream, Z_DEFAULT_COMPRESSION, Z_DEFLATED, algo, 8, Z_DEFAULT_STRATEGY) == Z_OK)
+            {
+                char buffer[8192];
+
+                stream.avail_in = str.size();
+                // zlib does not take a const pointer. The data is not altered.
+                stream.next_in = const_cast<Bytef*>(reinterpret_cast<const Bytef*>(str.c_str()));
+
+                int code = Z_OK;
+                do
+                {
+                    stream.avail_out = sizeof(buffer);
+                    stream.next_out = reinterpret_cast<Bytef*>(&buffer[0]);
+
+                    code = ::deflate(&stream, Z_FINISH);
+                    // Successful and non-fatal error code returned by deflate when used with Z_FINISH flush
+                    if (code == Z_OK || code == Z_STREAM_END)
+                    {
+                        std::copy(&buffer[0], &buffer[sizeof(buffer) - stream.avail_out], std::back_inserter(compressed_str));
+                    }
+
+                } while (code == Z_OK);
+
+                if (code != Z_STREAM_END)
+                    compressed_str.clear();
+
+                ::deflateEnd(&stream);
+            }
+
+            return compressed_str;
+        }
+
+        inline std::string decompress_string(std::string const& deflated_string)
+        {
+            std::string inflated_string;
+            Bytef tmp[8192];
+
+            z_stream zstream{};
+            zstream.avail_in = deflated_string.size();
+            // Nasty const_cast but zlib won't alter its contents
+            zstream.next_in = const_cast<Bytef*>(reinterpret_cast<Bytef const*>(deflated_string.c_str()));
+            // Initialize with automatic header detection, for gzip support
+            if (::inflateInit2(&zstream, MAX_WBITS | 32) == Z_OK)
+            {
+                do
+                {
+                    zstream.avail_out = sizeof(tmp);
+                    zstream.next_out = &tmp[0];
+
+                    auto ret = ::inflate(&zstream, Z_NO_FLUSH);
+                    if (ret == Z_OK || ret == Z_STREAM_END)
+                    {
+                        std::copy(&tmp[0], &tmp[sizeof(tmp) - zstream.avail_out], std::back_inserter(inflated_string));
+                    }
+                    else
+                    {
+                        // Something went wrong with inflate; make sure we return an empty string
+                        inflated_string.clear();
+                        break;
+                    }
+
+                } while (zstream.avail_out == 0);
+
+                // Free zlib's internal memory
+                ::inflateEnd(&zstream);
+            }
+
+            return inflated_string;
+        }
+    } // namespace compression
+} // namespace crow
+
+#endif
+
+/*
+ * SHA1 Wikipedia Page: http://en.wikipedia.org/wiki/SHA-1
+ * 
+ * Copyright (c) 2012-22 SAURAV MOHAPATRA <[email protected]>
+ *
+ * Permission to use, copy, modify, and distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+/** 
+ * \file TinySHA1.hpp
+ * \author SAURAV MOHAPATRA <[email protected]>
+ * \date 2012-22
+ * \brief TinySHA1 - a header only implementation of the SHA1 algorithm in C++. Based
+ * on the implementation in boost::uuid::details.
+ *
+ * In this file are defined:
+ * - sha1::SHA1
+ */
+#ifndef _TINY_SHA1_HPP_
+#define _TINY_SHA1_HPP_
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <stdint.h>
+
+/**
+ * \namespace sha1
+ * \brief Here is defined the SHA1 class
+ */
+namespace sha1
+{
+	/**
+	 * \class SHA1
+	 * \brief A tiny SHA1 algorithm implementation used internally in the
+	 * Crow server (specifically in crow/websocket.h).
+	 */
+	class SHA1
+	{
+	public:
+		typedef uint32_t digest32_t[5];
+		typedef uint8_t digest8_t[20];
+		inline static uint32_t LeftRotate(uint32_t value, size_t count) {
+			return (value << count) ^ (value >> (32-count));
+		}
+		SHA1(){ reset(); }
+		virtual ~SHA1() {}
+		SHA1(const SHA1& s) { *this = s; }
+		const SHA1& operator = (const SHA1& s) {
+			memcpy(m_digest, s.m_digest, 5 * sizeof(uint32_t));
+			memcpy(m_block, s.m_block, 64);
+			m_blockByteIndex = s.m_blockByteIndex;
+			m_byteCount = s.m_byteCount;
+			return *this;
+		}
+		SHA1& reset() {
+			m_digest[0] = 0x67452301;
+			m_digest[1] = 0xEFCDAB89;
+			m_digest[2] = 0x98BADCFE;
+			m_digest[3] = 0x10325476;
+			m_digest[4] = 0xC3D2E1F0;
+			m_blockByteIndex = 0;
+			m_byteCount = 0;
+			return *this;
+		}
+		SHA1& processByte(uint8_t octet) {
+			this->m_block[this->m_blockByteIndex++] = octet;
+			++this->m_byteCount;
+			if(m_blockByteIndex == 64) {
+				this->m_blockByteIndex = 0;
+				processBlock();
+			}
+			return *this;
+		}
+		SHA1& processBlock(const void* const start, const void* const end) {
+			const uint8_t* begin = static_cast<const uint8_t*>(start);
+			const uint8_t* finish = static_cast<const uint8_t*>(end);
+			while(begin != finish) {
+				processByte(*begin);
+				begin++;
+			}
+			return *this;
+		}
+		SHA1& processBytes(const void* const data, size_t len) {
+			const uint8_t* block = static_cast<const uint8_t*>(data);
+			processBlock(block, block + len);
+			return *this;
+		}
+		const uint32_t* getDigest(digest32_t digest) {
+			size_t bitCount = this->m_byteCount * 8;
+			processByte(0x80);
+			if (this->m_blockByteIndex > 56) {
+				while (m_blockByteIndex != 0) {
+					processByte(0);
+				}
+				while (m_blockByteIndex < 56) {
+					processByte(0);
+				}
+			} else {
+				while (m_blockByteIndex < 56) {
+					processByte(0);
+				}
+			}
+			processByte(0);
+			processByte(0);
+			processByte(0);
+			processByte(0);
+			processByte( static_cast<unsigned char>((bitCount>>24) & 0xFF));
+			processByte( static_cast<unsigned char>((bitCount>>16) & 0xFF));
+			processByte( static_cast<unsigned char>((bitCount>>8 ) & 0xFF));
+			processByte( static_cast<unsigned char>((bitCount)     & 0xFF));
+	
+			memcpy(digest, m_digest, 5 * sizeof(uint32_t));
+			return digest;
+		}
+		const uint8_t* getDigestBytes(digest8_t digest) {
+			digest32_t d32;
+			getDigest(d32);
+			size_t di = 0;
+			digest[di++] = ((d32[0] >> 24) & 0xFF);
+			digest[di++] = ((d32[0] >> 16) & 0xFF);
+			digest[di++] = ((d32[0] >> 8) & 0xFF);
+			digest[di++] = ((d32[0]) & 0xFF);
+			
+			digest[di++] = ((d32[1] >> 24) & 0xFF);
+			digest[di++] = ((d32[1] >> 16) & 0xFF);
+			digest[di++] = ((d32[1] >> 8) & 0xFF);
+			digest[di++] = ((d32[1]) & 0xFF);
+			
+			digest[di++] = ((d32[2] >> 24) & 0xFF);
+			digest[di++] = ((d32[2] >> 16) & 0xFF);
+			digest[di++] = ((d32[2] >> 8) & 0xFF);
+			digest[di++] = ((d32[2]) & 0xFF);
+			
+			digest[di++] = ((d32[3] >> 24) & 0xFF);
+			digest[di++] = ((d32[3] >> 16) & 0xFF);
+			digest[di++] = ((d32[3] >> 8) & 0xFF);
+			digest[di++] = ((d32[3]) & 0xFF);
+			
+			digest[di++] = ((d32[4] >> 24) & 0xFF);
+			digest[di++] = ((d32[4] >> 16) & 0xFF);
+			digest[di++] = ((d32[4] >> 8) & 0xFF);
+			digest[di++] = ((d32[4]) & 0xFF);
+			return digest;
+		}
+	
+	protected:
+		void processBlock() {
+			uint32_t w[80];
+			for (size_t i = 0; i < 16; i++) {
+				w[i]  = (m_block[i*4 + 0] << 24);
+				w[i] |= (m_block[i*4 + 1] << 16);
+				w[i] |= (m_block[i*4 + 2] << 8);
+				w[i] |= (m_block[i*4 + 3]);
+			}
+			for (size_t i = 16; i < 80; i++) {
+				w[i] = LeftRotate((w[i-3] ^ w[i-8] ^ w[i-14] ^ w[i-16]), 1);
+			}
+	
+			uint32_t a = m_digest[0];
+			uint32_t b = m_digest[1];
+			uint32_t c = m_digest[2];
+			uint32_t d = m_digest[3];
+			uint32_t e = m_digest[4];
+	
+			for (std::size_t i=0; i<80; ++i) {
+				uint32_t f = 0;
+				uint32_t k = 0;
+	
+				if (i<20) {
+					f = (b & c) | (~b & d);
+					k = 0x5A827999;
+				} else if (i<40) {
+					f = b ^ c ^ d;
+					k = 0x6ED9EBA1;
+				} else if (i<60) {
+					f = (b & c) | (b & d) | (c & d);
+					k = 0x8F1BBCDC;
+				} else {
+					f = b ^ c ^ d;
+					k = 0xCA62C1D6;
+				}
+				uint32_t temp = LeftRotate(a, 5) + f + e + k + w[i];
+				e = d;
+				d = c;
+				c = LeftRotate(b, 30);
+				b = a;
+				a = temp;
+			}
+	
+			m_digest[0] += a;
+			m_digest[1] += b;
+			m_digest[2] += c;
+			m_digest[3] += d;
+			m_digest[4] += e;
+		}
+	private:
+		digest32_t m_digest;
+		uint8_t m_block[64];
+		size_t m_blockByteIndex;
+		size_t m_byteCount;
+	};
+}
+#endif
+
+// settings for crow
+// TODO(ipkn) replace with runtime config. libucl?
+
+/* #ifdef - enables debug mode */
+//#define CROW_ENABLE_DEBUG
+
+/* #ifdef - enables logging */
+#define CROW_ENABLE_LOGGING
+
+/* #ifdef - enforces section 5.2 and 6.1 of RFC6455 (only accepting masked messages from clients) */
+//#define CROW_ENFORCE_WS_SPEC
+
+/* #define - specifies log level */
+/*
+    Debug       = 0
+    Info        = 1
+    Warning     = 2
+    Error       = 3
+    Critical    = 4
+
+    default to INFO
+*/
+#ifndef CROW_LOG_LEVEL
+#define CROW_LOG_LEVEL 1
+#endif
+
+#ifndef CROW_STATIC_DIRECTORY
+#define CROW_STATIC_DIRECTORY "static/"
+#endif
+#ifndef CROW_STATIC_ENDPOINT
+#define CROW_STATIC_ENDPOINT "/static/<path>"
+#endif
+
+// compiler flags
+#if defined(_MSVC_LANG) && _MSVC_LANG >= 201402L
+#define CROW_CAN_USE_CPP14
+#endif
+#if __cplusplus >= 201402L
+#define CROW_CAN_USE_CPP14
+#endif
+
+#if defined(_MSVC_LANG) && _MSVC_LANG >= 201703L
+#define CROW_CAN_USE_CPP17
+#endif
+#if __cplusplus >= 201703L
+#define CROW_CAN_USE_CPP17
+#if defined(__GNUC__) && __GNUC__ < 8
+#define CROW_FILESYSTEM_IS_EXPERIMENTAL
+#endif
+#endif
+
+#if defined(_MSC_VER)
+#if _MSC_VER < 1900
+#define CROW_MSVC_WORKAROUND
+#define constexpr const
+#define noexcept throw()
+#endif
+#endif
+
+#if defined(__GNUC__) && __GNUC__ == 8 && __GNUC_MINOR__ < 4
+#if __cplusplus > 201103L
+#define CROW_GCC83_WORKAROUND
+#else
+#error "GCC 8.1 - 8.3 has a bug that prevents Crow from compiling with C++11. Please update GCC to > 8.3 or use C++ > 11."
+#endif
+#endif
+
+
+#ifdef CROW_USE_BOOST
+#include <boost/asio.hpp>
+#include <boost/asio/version.hpp>
+#ifdef CROW_ENABLE_SSL
+#include <boost/asio/ssl.hpp>
+#endif
+#else
+#ifndef ASIO_STANDALONE
+#define ASIO_STANDALONE
+#endif
+#include <asio.hpp>
+#include <asio/version.hpp>
+#ifdef CROW_ENABLE_SSL
+#include <asio/ssl.hpp>
+#endif
+#endif
+
+#if (CROW_USE_BOOST && BOOST_VERSION >= 107000) || (ASIO_VERSION >= 101300)
+#define GET_IO_SERVICE(s) ((asio::io_context&)(s).get_executor().context())
+#else
+#define GET_IO_SERVICE(s) ((s).get_io_service())
+#endif
+
+namespace crow
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+    using error_code = boost::system::error_code;
+#else
+    using error_code = asio::error_code;
+#endif
+    using tcp = asio::ip::tcp;
+
+    /// A wrapper for the asio::ip::tcp::socket and asio::ssl::stream
+    struct SocketAdaptor
+    {
+        using context = void;
+        SocketAdaptor(asio::io_service& io_service, context*):
+          socket_(io_service)
+        {}
+
+        asio::io_service& get_io_service()
+        {
+            return GET_IO_SERVICE(socket_);
+        }
+
+        /// Get the TCP socket handling data trasfers, regardless of what layer is handling transfers on top of the socket.
+        tcp::socket& raw_socket()
+        {
+            return socket_;
+        }
+
+        /// Get the object handling data transfers, this can be either a TCP socket or an SSL stream (if SSL is enabled).
+        tcp::socket& socket()
+        {
+            return socket_;
+        }
+
+        tcp::endpoint remote_endpoint()
+        {
+            return socket_.remote_endpoint();
+        }
+
+        bool is_open()
+        {
+            return socket_.is_open();
+        }
+
+        void close()
+        {
+            error_code ec;
+            socket_.close(ec);
+        }
+
+        void shutdown_readwrite()
+        {
+            error_code ec;
+            socket_.shutdown(asio::socket_base::shutdown_type::shutdown_both, ec);
+        }
+
+        void shutdown_write()
+        {
+            error_code ec;
+            socket_.shutdown(asio::socket_base::shutdown_type::shutdown_send, ec);
+        }
+
+        void shutdown_read()
+        {
+            error_code ec;
+            socket_.shutdown(asio::socket_base::shutdown_type::shutdown_receive, ec);
+        }
+
+        template<typename F>
+        void start(F f)
+        {
+            f(error_code());
+        }
+
+        tcp::socket socket_;
+    };
+
+#ifdef CROW_ENABLE_SSL
+    struct SSLAdaptor
+    {
+        using context = asio::ssl::context;
+        using ssl_socket_t = asio::ssl::stream<tcp::socket>;
+        SSLAdaptor(asio::io_service& io_service, context* ctx):
+          ssl_socket_(new ssl_socket_t(io_service, *ctx))
+        {}
+
+        asio::ssl::stream<tcp::socket>& socket()
+        {
+            return *ssl_socket_;
+        }
+
+        tcp::socket::lowest_layer_type&
+          raw_socket()
+        {
+            return ssl_socket_->lowest_layer();
+        }
+
+        tcp::endpoint remote_endpoint()
+        {
+            return raw_socket().remote_endpoint();
+        }
+
+        bool is_open()
+        {
+            return ssl_socket_ ? raw_socket().is_open() : false;
+        }
+
+        void close()
+        {
+            if (is_open())
+            {
+                error_code ec;
+                raw_socket().close(ec);
+            }
+        }
+
+        void shutdown_readwrite()
+        {
+            if (is_open())
+            {
+                error_code ec;
+                raw_socket().shutdown(asio::socket_base::shutdown_type::shutdown_both, ec);
+            }
+        }
+
+        void shutdown_write()
+        {
+            if (is_open())
+            {
+                error_code ec;
+                raw_socket().shutdown(asio::socket_base::shutdown_type::shutdown_send, ec);
+            }
+        }
+
+        void shutdown_read()
+        {
+            if (is_open())
+            {
+                error_code ec;
+                raw_socket().shutdown(asio::socket_base::shutdown_type::shutdown_receive, ec);
+            }
+        }
+
+        asio::io_service& get_io_service()
+        {
+            return GET_IO_SERVICE(raw_socket());
+        }
+
+        template<typename F>
+        void start(F f)
+        {
+            ssl_socket_->async_handshake(asio::ssl::stream_base::server,
+                                         [f](const error_code& ec) {
+                                             f(ec);
+                                         });
+        }
+
+        std::unique_ptr<asio::ssl::stream<tcp::socket>> ssl_socket_;
+    };
+#endif
+} // namespace crow
+
+
+#include <cstdint>
+#include <stdexcept>
+#include <tuple>
+#include <type_traits>
+#include <cstring>
+#include <cctype>
+#include <functional>
+#include <string>
+#include <sstream>
+#include <unordered_map>
+#include <random>
+#include <algorithm>
+
+
+#if defined(CROW_CAN_USE_CPP17) && !defined(CROW_FILESYSTEM_IS_EXPERIMENTAL)
+#include <filesystem>
+#endif
+
+// TODO(EDev): Adding C++20's [[likely]] and [[unlikely]] attributes might be useful
+#if defined(__GNUG__) || defined(__clang__)
+#define CROW_LIKELY(X) __builtin_expect(!!(X), 1)
+#define CROW_UNLIKELY(X) __builtin_expect(!!(X), 0)
+#else
+#define CROW_LIKELY(X) (X)
+#define CROW_UNLIKELY(X) (X)
+#endif
+
+namespace crow
+{
+    /// @cond SKIP
+    namespace black_magic
+    {
+#ifndef CROW_MSVC_WORKAROUND
+        /// Out of Range Exception for const_str
+        struct OutOfRange
+        {
+            OutOfRange(unsigned /*pos*/, unsigned /*length*/) {}
+        };
+        /// Helper function to throw an exception if i is larger than len
+        constexpr unsigned requires_in_range(unsigned i, unsigned len)
+        {
+            return i >= len ? throw OutOfRange(i, len) : i;
+        }
+
+        /// A constant string implementation.
+        class const_str
+        {
+            const char* const begin_;
+            unsigned size_;
+
+        public:
+            template<unsigned N>
+            constexpr const_str(const char (&arr)[N]):
+              begin_(arr), size_(N - 1)
+            {
+                static_assert(N >= 1, "not a string literal");
+            }
+            constexpr char operator[](unsigned i) const
+            {
+                return requires_in_range(i, size_), begin_[i];
+            }
+
+            constexpr operator const char*() const
+            {
+                return begin_;
+            }
+
+            constexpr const char* begin() const { return begin_; }
+            constexpr const char* end() const { return begin_ + size_; }
+
+            constexpr unsigned size() const
+            {
+                return size_;
+            }
+        };
+
+        constexpr unsigned find_closing_tag(const_str s, unsigned p)
+        {
+            return s[p] == '>' ? p : find_closing_tag(s, p + 1);
+        }
+
+        /// Check that the CROW_ROUTE string is valid
+        constexpr bool is_valid(const_str s, unsigned i = 0, int f = 0)
+        {
+            return i == s.size()   ? f == 0 :
+                   f < 0 || f >= 2 ? false :
+                   s[i] == '<'     ? is_valid(s, i + 1, f + 1) :
+                   s[i] == '>'     ? is_valid(s, i + 1, f - 1) :
+                                     is_valid(s, i + 1, f);
+        }
+
+        constexpr bool is_equ_p(const char* a, const char* b, unsigned n)
+        {
+            return *a == 0 && *b == 0 && n == 0 ? true :
+                   (*a == 0 || *b == 0)         ? false :
+                   n == 0                       ? true :
+                   *a != *b                     ? false :
+                                                  is_equ_p(a + 1, b + 1, n - 1);
+        }
+
+        constexpr bool is_equ_n(const_str a, unsigned ai, const_str b, unsigned bi, unsigned n)
+        {
+            return ai + n > a.size() || bi + n > b.size() ? false :
+                   n == 0                                 ? true :
+                   a[ai] != b[bi]                         ? false :
+                                                            is_equ_n(a, ai + 1, b, bi + 1, n - 1);
+        }
+
+        constexpr bool is_int(const_str s, unsigned i)
+        {
+            return is_equ_n(s, i, "<int>", 0, 5);
+        }
+
+        constexpr bool is_uint(const_str s, unsigned i)
+        {
+            return is_equ_n(s, i, "<uint>", 0, 6);
+        }
+
+        constexpr bool is_float(const_str s, unsigned i)
+        {
+            return is_equ_n(s, i, "<float>", 0, 7) ||
+                   is_equ_n(s, i, "<double>", 0, 8);
+        }
+
+        constexpr bool is_str(const_str s, unsigned i)
+        {
+            return is_equ_n(s, i, "<str>", 0, 5) ||
+                   is_equ_n(s, i, "<string>", 0, 8);
+        }
+
+        constexpr bool is_path(const_str s, unsigned i)
+        {
+            return is_equ_n(s, i, "<path>", 0, 6);
+        }
+#endif
+        template<typename T>
+        struct parameter_tag
+        {
+            static const int value = 0;
+        };
+#define CROW_INTERNAL_PARAMETER_TAG(t, i) \
+    template<>                            \
+    struct parameter_tag<t>               \
+    {                                     \
+        static const int value = i;       \
+    }
+        CROW_INTERNAL_PARAMETER_TAG(int, 1);
+        CROW_INTERNAL_PARAMETER_TAG(char, 1);
+        CROW_INTERNAL_PARAMETER_TAG(short, 1);
+        CROW_INTERNAL_PARAMETER_TAG(long, 1);
+        CROW_INTERNAL_PARAMETER_TAG(long long, 1);
+        CROW_INTERNAL_PARAMETER_TAG(unsigned int, 2);
+        CROW_INTERNAL_PARAMETER_TAG(unsigned char, 2);
+        CROW_INTERNAL_PARAMETER_TAG(unsigned short, 2);
+        CROW_INTERNAL_PARAMETER_TAG(unsigned long, 2);
+        CROW_INTERNAL_PARAMETER_TAG(unsigned long long, 2);
+        CROW_INTERNAL_PARAMETER_TAG(double, 3);
+        CROW_INTERNAL_PARAMETER_TAG(std::string, 4);
+#undef CROW_INTERNAL_PARAMETER_TAG
+        template<typename... Args>
+        struct compute_parameter_tag_from_args_list;
+
+        template<>
+        struct compute_parameter_tag_from_args_list<>
+        {
+            static const int value = 0;
+        };
+
+        template<typename Arg, typename... Args>
+        struct compute_parameter_tag_from_args_list<Arg, Args...>
+        {
+            static const int sub_value =
+              compute_parameter_tag_from_args_list<Args...>::value;
+            static const int value =
+              parameter_tag<typename std::decay<Arg>::type>::value ? sub_value * 6 + parameter_tag<typename std::decay<Arg>::type>::value : sub_value;
+        };
+
+        static inline bool is_parameter_tag_compatible(uint64_t a, uint64_t b)
+        {
+            if (a == 0)
+                return b == 0;
+            if (b == 0)
+                return a == 0;
+            int sa = a % 6;
+            int sb = a % 6;
+            if (sa == 5) sa = 4;
+            if (sb == 5) sb = 4;
+            if (sa != sb)
+                return false;
+            return is_parameter_tag_compatible(a / 6, b / 6);
+        }
+
+        static inline unsigned find_closing_tag_runtime(const char* s, unsigned p)
+        {
+            return s[p] == 0   ? throw std::runtime_error("unmatched tag <") :
+                   s[p] == '>' ? p :
+                                 find_closing_tag_runtime(s, p + 1);
+        }
+
+        static inline uint64_t get_parameter_tag_runtime(const char* s, unsigned p = 0)
+        {
+            return s[p] == 0   ? 0 :
+                   s[p] == '<' ? (
+                                   std::strncmp(s + p, "<int>", 5) == 0  ? get_parameter_tag_runtime(s, find_closing_tag_runtime(s, p)) * 6 + 1 :
+                                   std::strncmp(s + p, "<uint>", 6) == 0 ? get_parameter_tag_runtime(s, find_closing_tag_runtime(s, p)) * 6 + 2 :
+                                   (std::strncmp(s + p, "<float>", 7) == 0 ||
+                                    std::strncmp(s + p, "<double>", 8) == 0) ?
+                                                                           get_parameter_tag_runtime(s, find_closing_tag_runtime(s, p)) * 6 + 3 :
+                                   (std::strncmp(s + p, "<str>", 5) == 0 ||
+                                    std::strncmp(s + p, "<string>", 8) == 0) ?
+                                                                           get_parameter_tag_runtime(s, find_closing_tag_runtime(s, p)) * 6 + 4 :
+                                   std::strncmp(s + p, "<path>", 6) == 0 ? get_parameter_tag_runtime(s, find_closing_tag_runtime(s, p)) * 6 + 5 :
+                                                                           throw std::runtime_error("invalid parameter type")) :
+                                 get_parameter_tag_runtime(s, p + 1);
+        }
+#ifndef CROW_MSVC_WORKAROUND
+        constexpr uint64_t get_parameter_tag(const_str s, unsigned p = 0)
+        {
+            return p == s.size() ? 0 :
+                   s[p] == '<'   ? (
+                                   is_int(s, p)   ? get_parameter_tag(s, find_closing_tag(s, p)) * 6 + 1 :
+                                     is_uint(s, p)  ? get_parameter_tag(s, find_closing_tag(s, p)) * 6 + 2 :
+                                     is_float(s, p) ? get_parameter_tag(s, find_closing_tag(s, p)) * 6 + 3 :
+                                     is_str(s, p)   ? get_parameter_tag(s, find_closing_tag(s, p)) * 6 + 4 :
+                                     is_path(s, p)  ? get_parameter_tag(s, find_closing_tag(s, p)) * 6 + 5 :
+                                                      throw std::runtime_error("invalid parameter type")) :
+                                 get_parameter_tag(s, p + 1);
+        }
+#endif
+
+        template<typename... T>
+        struct S
+        {
+            template<typename U>
+            using push = S<U, T...>;
+            template<typename U>
+            using push_back = S<T..., U>;
+            template<template<typename... Args> class U>
+            using rebind = U<T...>;
+        };
+
+        // Check whether the template function can be called with specific arguments
+        template<typename F, typename Set>
+        struct CallHelper;
+        template<typename F, typename... Args>
+        struct CallHelper<F, S<Args...>>
+        {
+            template<typename F1, typename... Args1, typename = decltype(std::declval<F1>()(std::declval<Args1>()...))>
+            static char __test(int);
+
+            template<typename...>
+            static int __test(...);
+
+            static constexpr bool value = sizeof(__test<F, Args...>(0)) == sizeof(char);
+        };
+
+        // Check Tuple contains type T
+        template<typename T, typename Tuple>
+        struct has_type;
+
+        template<typename T>
+        struct has_type<T, std::tuple<>> : std::false_type
+        {};
+
+        template<typename T, typename U, typename... Ts>
+        struct has_type<T, std::tuple<U, Ts...>> : has_type<T, std::tuple<Ts...>>
+        {};
+
+        template<typename T, typename... Ts>
+        struct has_type<T, std::tuple<T, Ts...>> : std::true_type
+        {};
+
+        // Find index of type in tuple
+        template<class T, class Tuple>
+        struct tuple_index;
+
+        template<class T, class... Types>
+        struct tuple_index<T, std::tuple<T, Types...>>
+        {
+            static const int value = 0;
+        };
+
+        template<class T, class U, class... Types>
+        struct tuple_index<T, std::tuple<U, Types...>>
+        {
+            static const int value = 1 + tuple_index<T, std::tuple<Types...>>::value;
+        };
+
+        // Extract element from forward tuple or get default
+#ifdef CROW_CAN_USE_CPP14
+        template<typename T, typename Tup>
+        typename std::enable_if<has_type<T&, Tup>::value, typename std::decay<T>::type&&>::type
+          tuple_extract(Tup& tup)
+        {
+            return std::move(std::get<T&>(tup));
+        }
+#else
+        template<typename T, typename Tup>
+        typename std::enable_if<has_type<T&, Tup>::value, T&&>::type
+          tuple_extract(Tup& tup)
+        {
+            return std::move(std::get<tuple_index<T&, Tup>::value>(tup));
+        }
+#endif
+
+        template<typename T, typename Tup>
+        typename std::enable_if<!has_type<T&, Tup>::value, T>::type
+          tuple_extract(Tup&)
+        {
+            return T{};
+        }
+
+        // Kind of fold expressions in C++11
+        template<bool...>
+        struct bool_pack;
+        template<bool... bs>
+        using all_true = std::is_same<bool_pack<bs..., true>, bool_pack<true, bs...>>;
+
+        template<int N>
+        struct single_tag_to_type
+        {};
+
+        template<>
+        struct single_tag_to_type<1>
+        {
+            using type = int64_t;
+        };
+
+        template<>
+        struct single_tag_to_type<2>
+        {
+            using type = uint64_t;
+        };
+
+        template<>
+        struct single_tag_to_type<3>
+        {
+            using type = double;
+        };
+
+        template<>
+        struct single_tag_to_type<4>
+        {
+            using type = std::string;
+        };
+
+        template<>
+        struct single_tag_to_type<5>
+        {
+            using type = std::string;
+        };
+
+
+        template<uint64_t Tag>
+        struct arguments
+        {
+            using subarguments = typename arguments<Tag / 6>::type;
+            using type =
+              typename subarguments::template push<typename single_tag_to_type<Tag % 6>::type>;
+        };
+
+        template<>
+        struct arguments<0>
+        {
+            using type = S<>;
+        };
+
+        template<typename... T>
+        struct last_element_type
+        {
+            using type = typename std::tuple_element<sizeof...(T) - 1, std::tuple<T...>>::type;
+        };
+
+
+        template<>
+        struct last_element_type<>
+        {};
+
+
+        // from http://stackoverflow.com/questions/13072359/c11-compile-time-array-with-logarithmic-evaluation-depth
+        template<class T>
+        using Invoke = typename T::type;
+
+        template<unsigned...>
+        struct seq
+        {
+            using type = seq;
+        };
+
+        template<class S1, class S2>
+        struct concat;
+
+        template<unsigned... I1, unsigned... I2>
+        struct concat<seq<I1...>, seq<I2...>> : seq<I1..., (sizeof...(I1) + I2)...>
+        {};
+
+        template<class S1, class S2>
+        using Concat = Invoke<concat<S1, S2>>;
+
+        template<unsigned N>
+        struct gen_seq;
+        template<unsigned N>
+        using GenSeq = Invoke<gen_seq<N>>;
+
+        template<unsigned N>
+        struct gen_seq : Concat<GenSeq<N / 2>, GenSeq<N - N / 2>>
+        {};
+
+        template<>
+        struct gen_seq<0> : seq<>
+        {};
+        template<>
+        struct gen_seq<1> : seq<0>
+        {};
+
+        template<typename Seq, typename Tuple>
+        struct pop_back_helper;
+
+        template<unsigned... N, typename Tuple>
+        struct pop_back_helper<seq<N...>, Tuple>
+        {
+            template<template<typename... Args> class U>
+            using rebind = U<typename std::tuple_element<N, Tuple>::type...>;
+        };
+
+        template<typename... T>
+        struct pop_back //: public pop_back_helper<typename gen_seq<sizeof...(T)-1>::type, std::tuple<T...>>
+        {
+            template<template<typename... Args> class U>
+            using rebind = typename pop_back_helper<typename gen_seq<sizeof...(T) - 1>::type, std::tuple<T...>>::template rebind<U>;
+        };
+
+        template<>
+        struct pop_back<>
+        {
+            template<template<typename... Args> class U>
+            using rebind = U<>;
+        };
+
+        // from http://stackoverflow.com/questions/2118541/check-if-c0x-parameter-pack-contains-a-type
+        template<typename Tp, typename... List>
+        struct contains : std::true_type
+        {};
+
+        template<typename Tp, typename Head, typename... Rest>
+        struct contains<Tp, Head, Rest...> : std::conditional<std::is_same<Tp, Head>::value, std::true_type, contains<Tp, Rest...>>::type
+        {};
+
+        template<typename Tp>
+        struct contains<Tp> : std::false_type
+        {};
+
+        template<typename T>
+        struct empty_context
+        {};
+
+        template<typename T>
+        struct promote
+        {
+            using type = T;
+        };
+
+#define CROW_INTERNAL_PROMOTE_TYPE(t1, t2) \
+    template<>                             \
+    struct promote<t1>                     \
+    {                                      \
+        using type = t2;                   \
+    }
+
+        CROW_INTERNAL_PROMOTE_TYPE(char, int64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(short, int64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(int, int64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(long, int64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(long long, int64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(unsigned char, uint64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(unsigned short, uint64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(unsigned int, uint64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(unsigned long, uint64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(unsigned long long, uint64_t);
+        CROW_INTERNAL_PROMOTE_TYPE(float, double);
+#undef CROW_INTERNAL_PROMOTE_TYPE
+
+        template<typename T>
+        using promote_t = typename promote<T>::type;
+
+    } // namespace black_magic
+
+    namespace detail
+    {
+
+        template<class T, std::size_t N, class... Args>
+        struct get_index_of_element_from_tuple_by_type_impl
+        {
+            static constexpr auto value = N;
+        };
+
+        template<class T, std::size_t N, class... Args>
+        struct get_index_of_element_from_tuple_by_type_impl<T, N, T, Args...>
+        {
+            static constexpr auto value = N;
+        };
+
+        template<class T, std::size_t N, class U, class... Args>
+        struct get_index_of_element_from_tuple_by_type_impl<T, N, U, Args...>
+        {
+            static constexpr auto value = get_index_of_element_from_tuple_by_type_impl<T, N + 1, Args...>::value;
+        };
+    } // namespace detail
+
+    namespace utility
+    {
+        template<class T, class... Args>
+        T& get_element_by_type(std::tuple<Args...>& t)
+        {
+            return std::get<detail::get_index_of_element_from_tuple_by_type_impl<T, 0, Args...>::value>(t);
+        }
+
+        template<typename T>
+        struct function_traits;
+
+#ifndef CROW_MSVC_WORKAROUND
+        template<typename T>
+        struct function_traits : public function_traits<decltype(&T::operator())>
+        {
+            using parent_t = function_traits<decltype(&T::operator())>;
+            static const size_t arity = parent_t::arity;
+            using result_type = typename parent_t::result_type;
+            template<size_t i>
+            using arg = typename parent_t::template arg<i>;
+        };
+#endif
+
+        template<typename ClassType, typename R, typename... Args>
+        struct function_traits<R (ClassType::*)(Args...) const>
+        {
+            static const size_t arity = sizeof...(Args);
+
+            typedef R result_type;
+
+            template<size_t i>
+            using arg = typename std::tuple_element<i, std::tuple<Args...>>::type;
+        };
+
+        template<typename ClassType, typename R, typename... Args>
+        struct function_traits<R (ClassType::*)(Args...)>
+        {
+            static const size_t arity = sizeof...(Args);
+
+            typedef R result_type;
+
+            template<size_t i>
+            using arg = typename std::tuple_element<i, std::tuple<Args...>>::type;
+        };
+
+        template<typename R, typename... Args>
+        struct function_traits<std::function<R(Args...)>>
+        {
+            static const size_t arity = sizeof...(Args);
+
+            typedef R result_type;
+
+            template<size_t i>
+            using arg = typename std::tuple_element<i, std::tuple<Args...>>::type;
+        };
+        /// @endcond
+
+        inline static std::string base64encode(const unsigned char* data, size_t size, const char* key = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/")
+        {
+            std::string ret;
+            ret.resize((size + 2) / 3 * 4);
+            auto it = ret.begin();
+            while (size >= 3)
+            {
+                *it++ = key[(static_cast<unsigned char>(*data) & 0xFC) >> 2];
+                unsigned char h = (static_cast<unsigned char>(*data++) & 0x03) << 4;
+                *it++ = key[h | ((static_cast<unsigned char>(*data) & 0xF0) >> 4)];
+                h = (static_cast<unsigned char>(*data++) & 0x0F) << 2;
+                *it++ = key[h | ((static_cast<unsigned char>(*data) & 0xC0) >> 6)];
+                *it++ = key[static_cast<unsigned char>(*data++) & 0x3F];
+
+                size -= 3;
+            }
+            if (size == 1)
+            {
+                *it++ = key[(static_cast<unsigned char>(*data) & 0xFC) >> 2];
+                unsigned char h = (static_cast<unsigned char>(*data++) & 0x03) << 4;
+                *it++ = key[h];
+                *it++ = '=';
+                *it++ = '=';
+            }
+            else if (size == 2)
+            {
+                *it++ = key[(static_cast<unsigned char>(*data) & 0xFC) >> 2];
+                unsigned char h = (static_cast<unsigned char>(*data++) & 0x03) << 4;
+                *it++ = key[h | ((static_cast<unsigned char>(*data) & 0xF0) >> 4)];
+                h = (static_cast<unsigned char>(*data++) & 0x0F) << 2;
+                *it++ = key[h];
+                *it++ = '=';
+            }
+            return ret;
+        }
+
+        inline static std::string base64encode(std::string data, size_t size, const char* key = "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/")
+        {
+            return base64encode((const unsigned char*)data.c_str(), size, key);
+        }
+
+        inline static std::string base64encode_urlsafe(const unsigned char* data, size_t size)
+        {
+            return base64encode(data, size, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_");
+        }
+
+        inline static std::string base64encode_urlsafe(std::string data, size_t size)
+        {
+            return base64encode((const unsigned char*)data.c_str(), size, "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789-_");
+        }
+
+        inline static std::string base64decode(const char* data, size_t size)
+        {
+            // We accept both regular and url encoding here, as there does not seem to be any downside to that.
+            // If we want to distinguish that we should use +/ for non-url and -_ for url.
+
+            // Mapping logic from characters to [0-63]
+            auto key = [](char c) -> unsigned char {
+                if ((c >= 'A') && (c <= 'Z')) return c - 'A';
+                if ((c >= 'a') && (c <= 'z')) return c - 'a' + 26;
+                if ((c >= '0') && (c <= '9')) return c - '0' + 52;
+                if ((c == '+') || (c == '-')) return 62;
+                if ((c == '/') || (c == '_')) return 63;
+                return 0;
+            };
+
+            // Not padded
+            if (size % 4 == 2)             // missing last 2 characters
+                size = (size / 4 * 3) + 1; // Not subtracting extra characters because they're truncated in int division
+            else if (size % 4 == 3)        // missing last character
+                size = (size / 4 * 3) + 2; // Not subtracting extra characters because they're truncated in int division
+
+            // Padded
+            else if (data[size - 2] == '=') // padded with '=='
+                size = (size / 4 * 3) - 2;  // == padding means the last block only has 1 character instead of 3, hence the '-2'
+            else if (data[size - 1] == '=') // padded with '='
+                size = (size / 4 * 3) - 1;  // = padding means the last block only has 2 character instead of 3, hence the '-1'
+
+            // Padding not needed
+            else
+                size = size / 4 * 3;
+
+            std::string ret;
+            ret.resize(size);
+            auto it = ret.begin();
+
+            // These will be used to decode 1 character at a time
+            unsigned char odd;  // char1 and char3
+            unsigned char even; // char2 and char4
+
+            // Take 4 character blocks to turn into 3
+            while (size >= 3)
+            {
+                // dec_char1 = (char1 shifted 2 bits to the left) OR ((char2 AND 00110000) shifted 4 bits to the right))
+                odd = key(*data++);
+                even = key(*data++);
+                *it++ = (odd << 2) | ((even & 0x30) >> 4);
+                // dec_char2 = ((char2 AND 00001111) shifted 4 bits left) OR ((char3 AND 00111100) shifted 2 bits right))
+                odd = key(*data++);
+                *it++ = ((even & 0x0F) << 4) | ((odd & 0x3C) >> 2);
+                // dec_char3 = ((char3 AND 00000011) shifted 6 bits left) OR (char4)
+                even = key(*data++);
+                *it++ = ((odd & 0x03) << 6) | (even);
+
+                size -= 3;
+            }
+            if (size == 2)
+            {
+                // d_char1 = (char1 shifted 2 bits to the left) OR ((char2 AND 00110000) shifted 4 bits to the right))
+                odd = key(*data++);
+                even = key(*data++);
+                *it++ = (odd << 2) | ((even & 0x30) >> 4);
+                // d_char2 = ((char2 AND 00001111) shifted 4 bits left) OR ((char3 AND 00111100) shifted 2 bits right))
+                odd = key(*data++);
+                *it++ = ((even & 0x0F) << 4) | ((odd & 0x3C) >> 2);
+            }
+            else if (size == 1)
+            {
+                // d_char1 = (char1 shifted 2 bits to the left) OR ((char2 AND 00110000) shifted 4 bits to the right))
+                odd = key(*data++);
+                even = key(*data++);
+                *it++ = (odd << 2) | ((even & 0x30) >> 4);
+            }
+            return ret;
+        }
+
+        inline static std::string base64decode(const std::string& data, size_t size)
+        {
+            return base64decode(data.data(), size);
+        }
+
+        inline static std::string base64decode(const std::string& data)
+        {
+            return base64decode(data.data(), data.length());
+        }
+
+        inline static std::string normalize_path(const std::string& directoryPath)
+        {
+            std::string normalizedPath = directoryPath;
+            std::replace(normalizedPath.begin(), normalizedPath.end(), '\\', '/');
+            if (!normalizedPath.empty() && normalizedPath.back() != '/')
+                normalizedPath += '/';
+            return normalizedPath;
+        }
+
+        inline static void sanitize_filename(std::string& data, char replacement = '_')
+        {
+            if (data.length() > 255)
+                data.resize(255);
+
+            static const auto toUpper = [](char c) {
+                return ((c >= 'a') && (c <= 'z')) ? (c - ('a' - 'A')) : c;
+            };
+            // Check for special device names. The Windows behavior is really odd here, it will consider both AUX and AUX.txt
+            // a special device. Thus we search for the string (case-insensitive), and then check if the string ends or if
+            // is has a dangerous follow up character (.:\/)
+            auto sanitizeSpecialFile = [](std::string& source, unsigned ofs, const char* pattern, bool includeNumber, char replacement) {
+                unsigned i = ofs;
+                size_t len = source.length();
+                const char* p = pattern;
+                while (*p)
+                {
+                    if (i >= len) return;
+                    if (toUpper(source[i]) != *p) return;
+                    ++i;
+                    ++p;
+                }
+                if (includeNumber)
+                {
+                    if ((i >= len) || (source[i] < '1') || (source[i] > '9')) return;
+                    ++i;
+                }
+                if ((i >= len) || (source[i] == '.') || (source[i] == ':') || (source[i] == '/') || (source[i] == '\\'))
+                {
+                    source.erase(ofs + 1, (i - ofs) - 1);
+                    source[ofs] = replacement;
+                }
+            };
+            bool checkForSpecialEntries = true;
+            for (unsigned i = 0; i < data.length(); ++i)
+            {
+                // Recognize directory traversals and the special devices CON/PRN/AUX/NULL/COM[1-]/LPT[1-9]
+                if (checkForSpecialEntries)
+                {
+                    checkForSpecialEntries = false;
+                    switch (toUpper(data[i]))
+                    {
+                        case 'A':
+                            sanitizeSpecialFile(data, i, "AUX", false, replacement);
+                            break;
+                        case 'C':
+                            sanitizeSpecialFile(data, i, "CON", false, replacement);
+                            sanitizeSpecialFile(data, i, "COM", true, replacement);
+                            break;
+                        case 'L':
+                            sanitizeSpecialFile(data, i, "LPT", true, replacement);
+                            break;
+                        case 'N':
+                            sanitizeSpecialFile(data, i, "NUL", false, replacement);
+                            break;
+                        case 'P':
+                            sanitizeSpecialFile(data, i, "PRN", false, replacement);
+                            break;
+                        case '.':
+                            sanitizeSpecialFile(data, i, "..", false, replacement);
+                            break;
+                    }
+                }
+
+                // Sanitize individual characters
+                unsigned char c = data[i];
+                if ((c < ' ') || ((c >= 0x80) && (c <= 0x9F)) || (c == '?') || (c == '<') || (c == '>') || (c == ':') || (c == '*') || (c == '|') || (c == '\"'))
+                {
+                    data[i] = replacement;
+                }
+                else if ((c == '/') || (c == '\\'))
+                {
+                    if (CROW_UNLIKELY(i == 0)) //Prevent Unix Absolute Paths (Windows Absolute Paths are prevented with `(c == ':')`)
+                    {
+                        data[i] = replacement;
+                    }
+                    else
+                    {
+                        checkForSpecialEntries = true;
+                    }
+                }
+            }
+        }
+
+        inline static std::string random_alphanum(std::size_t size)
+        {
+            static const char alphabet[] = "0123456789abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ";
+            std::random_device dev;
+            std::mt19937 rng(dev());
+            std::uniform_int_distribution<std::mt19937::result_type> dist(0, sizeof(alphabet) - 2);
+            std::string out;
+            out.reserve(size);
+            for (std::size_t i = 0; i < size; i++)
+                out.push_back(alphabet[dist(rng)]);
+            return out;
+        }
+
+        inline static std::string join_path(std::string path, const std::string& fname)
+        {
+#if defined(CROW_CAN_USE_CPP17) && !defined(CROW_FILESYSTEM_IS_EXPERIMENTAL)
+            return (std::filesystem::path(path) / fname).string();
+#else
+            if (!(path.back() == '/' || path.back() == '\\'))
+                path += '/';
+            path += fname;
+            return path;
+#endif
+        }
+
+        /**
+         * @brief Checks two string for equality.
+         * Always returns false if strings differ in size.
+         * Defaults to case-insensitive comparison.
+         */
+        inline static bool string_equals(const std::string& l, const std::string& r, bool case_sensitive = false)
+        {
+            if (l.length() != r.length())
+                return false;
+
+            for (size_t i = 0; i < l.length(); i++)
+            {
+                if (case_sensitive)
+                {
+                    if (l[i] != r[i])
+                        return false;
+                }
+                else
+                {
+                    if (std::toupper(l[i]) != std::toupper(r[i]))
+                        return false;
+                }
+            }
+
+            return true;
+        }
+
+        template<typename T, typename U>
+        inline static T lexical_cast(const U& v)
+        {
+            std::stringstream stream;
+            T res;
+
+            stream << v;
+            stream >> res;
+
+            return res;
+        }
+
+        template<typename T>
+        inline static T lexical_cast(const char* v, size_t count)
+        {
+            std::stringstream stream;
+            T res;
+
+            stream.write(v, count);
+            stream >> res;
+
+            return res;
+        }
+
+
+        /// Return a copy of the given string with its
+        /// leading and trailing whitespaces removed.
+        inline static std::string trim(const std::string& v)
+        {
+            if (v.empty())
+                return "";
+
+            size_t begin = 0, end = v.length();
+
+            size_t i;
+            for (i = 0; i < v.length(); i++)
+            {
+                if (!std::isspace(v[i]))
+                {
+                    begin = i;
+                    break;
+                }
+            }
+
+            if (i == v.length())
+                return "";
+
+            for (i = v.length(); i > 0; i--)
+            {
+                if (!std::isspace(v[i - 1]))
+                {
+                    end = i;
+                    break;
+                }
+            }
+
+            return v.substr(begin, end - begin);
+        }
+    } // namespace utility
+} // namespace crow
+
+
+#include <locale>
+#include <unordered_map>
+
+namespace crow
+{
+    /// Hashing function for ci_map (unordered_multimap).
+    struct ci_hash
+    {
+        size_t operator()(const std::string& key) const
+        {
+            std::size_t seed = 0;
+            std::locale locale;
+
+            for (auto c : key)
+                hash_combine(seed, std::toupper(c, locale));
+
+            return seed;
+        }
+
+    private:
+        static inline void hash_combine(std::size_t& seed, char v)
+        {
+            std::hash<char> hasher;
+            seed ^= hasher(v) + 0x9e3779b9 + (seed << 6) + (seed >> 2);
+        }
+    };
+
+    /// Equals function for ci_map (unordered_multimap).
+    struct ci_key_eq
+    {
+        bool operator()(const std::string& l, const std::string& r) const
+        {
+            return utility::string_equals(l, r);
+        }
+    };
+
+    using ci_map = std::unordered_multimap<std::string, std::string, ci_hash, ci_key_eq>;
+} // namespace crow
+
+
+#include <vector>
+#include <string>
+#include <stdexcept>
+#include <iostream>
+
+namespace crow
+{
+    const char cr = '\r';
+    const char lf = '\n';
+    const std::string crlf("\r\n");
+
+    enum class HTTPMethod : char
+    {
+#ifndef DELETE
+        DELETE = 0,
+        GET,
+        HEAD,
+        POST,
+        PUT,
+
+        CONNECT,
+        OPTIONS,
+        TRACE,
+
+        PATCH,
+        PURGE,
+
+        COPY,
+        LOCK,
+        MKCOL,
+        MOVE,
+        PROPFIND,
+        PROPPATCH,
+        SEARCH,
+        UNLOCK,
+        BIND,
+        REBIND,
+        UNBIND,
+        ACL,
+
+        REPORT,
+        MKACTIVITY,
+        CHECKOUT,
+        MERGE,
+
+        MSEARCH,
+        NOTIFY,
+        SUBSCRIBE,
+        UNSUBSCRIBE,
+
+        MKCALENDAR,
+
+        LINK,
+        UNLINK,
+
+        SOURCE,
+#endif
+
+        Delete = 0,
+        Get,
+        Head,
+        Post,
+        Put,
+
+        Connect,
+        Options,
+        Trace,
+
+        Patch,
+        Purge,
+
+        Copy,
+        Lock,
+        MkCol,
+        Move,
+        Propfind,
+        Proppatch,
+        Search,
+        Unlock,
+        Bind,
+        Rebind,
+        Unbind,
+        Acl,
+
+        Report,
+        MkActivity,
+        Checkout,
+        Merge,
+
+        MSearch,
+        Notify,
+        Subscribe,
+        Unsubscribe,
+
+        MkCalendar,
+
+        Link,
+        Unlink,
+
+        Source,
+
+
+        InternalMethodCount,
+        // should not add an item below this line: used for array count
+    };
+
+    constexpr const char* method_strings[] =
+      {
+        "DELETE",
+        "GET",
+        "HEAD",
+        "POST",
+        "PUT",
+
+        "CONNECT",
+        "OPTIONS",
+        "TRACE",
+
+        "PATCH",
+        "PURGE",
+
+        "COPY",
+        "LOCK",
+        "MKCOL",
+        "MOVE",
+        "PROPFIND",
+        "PROPPATCH",
+        "SEARCH",
+        "UNLOCK",
+        "BIND",
+        "REBIND",
+        "UNBIND",
+        "ACL",
+
+        "REPORT",
+        "MKACTIVITY",
+        "CHECKOUT",
+        "MERGE",
+
+        "M-SEARCH",
+        "NOTIFY",
+        "SUBSCRIBE",
+        "UNSUBSCRIBE",
+
+        "MKCALENDAR",
+
+        "LINK",
+        "UNLINK",
+
+        "SOURCE"};
+
+
+    inline std::string method_name(HTTPMethod method)
+    {
+        if (CROW_LIKELY(method < HTTPMethod::InternalMethodCount))
+        {
+            return method_strings[(unsigned char)method];
+        }
+        return "invalid";
+    }
+
+    // clang-format off
+
+    enum status
+    {
+        CONTINUE                      = 100,
+        SWITCHING_PROTOCOLS           = 101,
+
+        OK                            = 200,
+        CREATED                       = 201,
+        ACCEPTED                      = 202,
+        NON_AUTHORITATIVE_INFORMATION = 203,
+        NO_CONTENT                    = 204,
+        RESET_CONTENT                 = 205,
+        PARTIAL_CONTENT               = 206,
+
+        MULTIPLE_CHOICES              = 300,
+        MOVED_PERMANENTLY             = 301,
+        FOUND                         = 302,
+        SEE_OTHER                     = 303,
+        NOT_MODIFIED                  = 304,
+        TEMPORARY_REDIRECT            = 307,
+        PERMANENT_REDIRECT            = 308,
+
+        BAD_REQUEST                   = 400,
+        UNAUTHORIZED                  = 401,
+        FORBIDDEN                     = 403,
+        NOT_FOUND                     = 404,
+        METHOD_NOT_ALLOWED            = 405,
+        NOT_ACCEPTABLE                = 406,
+        PROXY_AUTHENTICATION_REQUIRED = 407,
+        CONFLICT                      = 409,
+        GONE                          = 410,
+        PAYLOAD_TOO_LARGE             = 413,
+        UNSUPPORTED_MEDIA_TYPE        = 415,
+        RANGE_NOT_SATISFIABLE         = 416,
+        EXPECTATION_FAILED            = 417,
+        PRECONDITION_REQUIRED         = 428,
+        TOO_MANY_REQUESTS             = 429,
+        UNAVAILABLE_FOR_LEGAL_REASONS = 451,
+
+        INTERNAL_SERVER_ERROR         = 500,
+        NOT_IMPLEMENTED               = 501,
+        BAD_GATEWAY                   = 502,
+        SERVICE_UNAVAILABLE           = 503,
+        GATEWAY_TIMEOUT               = 504,
+        VARIANT_ALSO_NEGOTIATES       = 506
+    };
+
+    // clang-format on
+
+    enum class ParamType : char
+    {
+        INT,
+        UINT,
+        DOUBLE,
+        STRING,
+        PATH,
+
+        MAX
+    };
+
+    /// @cond SKIP
+    struct routing_params
+    {
+        std::vector<int64_t> int_params;
+        std::vector<uint64_t> uint_params;
+        std::vector<double> double_params;
+        std::vector<std::string> string_params;
+
+        void debug_print() const
+        {
+            std::cerr << "routing_params" << std::endl;
+            for (auto i : int_params)
+                std::cerr << i << ", ";
+            std::cerr << std::endl;
+            for (auto i : uint_params)
+                std::cerr << i << ", ";
+            std::cerr << std::endl;
+            for (auto i : double_params)
+                std::cerr << i << ", ";
+            std::cerr << std::endl;
+            for (auto& i : string_params)
+                std::cerr << i << ", ";
+            std::cerr << std::endl;
+        }
+
+        template<typename T>
+        T get(unsigned) const;
+    };
+
+    template<>
+    inline int64_t routing_params::get<int64_t>(unsigned index) const
+    {
+        return int_params[index];
+    }
+
+    template<>
+    inline uint64_t routing_params::get<uint64_t>(unsigned index) const
+    {
+        return uint_params[index];
+    }
+
+    template<>
+    inline double routing_params::get<double>(unsigned index) const
+    {
+        return double_params[index];
+    }
+
+    template<>
+    inline std::string routing_params::get<std::string>(unsigned index) const
+    {
+        return string_params[index];
+    }
+    /// @endcond
+
+    struct routing_handle_result
+    {
+        uint16_t rule_index;
+        std::vector<uint16_t> blueprint_indices;
+        routing_params r_params;
+        HTTPMethod method;
+
+        routing_handle_result() {}
+
+        routing_handle_result(uint16_t rule_index_, std::vector<uint16_t> blueprint_indices_, routing_params r_params_):
+          rule_index(rule_index_),
+          blueprint_indices(blueprint_indices_),
+          r_params(r_params_) {}
+
+        routing_handle_result(uint16_t rule_index_, std::vector<uint16_t> blueprint_indices_, routing_params r_params_, HTTPMethod method_):
+          rule_index(rule_index_),
+          blueprint_indices(blueprint_indices_),
+          r_params(r_params_),
+          method(method_) {}
+    };
+} // namespace crow
+
+// clang-format off
+#ifndef CROW_MSVC_WORKAROUND
+constexpr crow::HTTPMethod method_from_string(const char* str)
+{
+    return crow::black_magic::is_equ_p(str, "GET", 3)    ? crow::HTTPMethod::Get :
+           crow::black_magic::is_equ_p(str, "DELETE", 6) ? crow::HTTPMethod::Delete :
+           crow::black_magic::is_equ_p(str, "HEAD", 4)   ? crow::HTTPMethod::Head :
+           crow::black_magic::is_equ_p(str, "POST", 4)   ? crow::HTTPMethod::Post :
+           crow::black_magic::is_equ_p(str, "PUT", 3)    ? crow::HTTPMethod::Put :
+
+           crow::black_magic::is_equ_p(str, "OPTIONS", 7) ? crow::HTTPMethod::Options :
+           crow::black_magic::is_equ_p(str, "CONNECT", 7) ? crow::HTTPMethod::Connect :
+           crow::black_magic::is_equ_p(str, "TRACE", 5)   ? crow::HTTPMethod::Trace :
+
+           crow::black_magic::is_equ_p(str, "PATCH", 5)     ? crow::HTTPMethod::Patch :
+           crow::black_magic::is_equ_p(str, "PURGE", 5)     ? crow::HTTPMethod::Purge :
+           crow::black_magic::is_equ_p(str, "COPY", 4)      ? crow::HTTPMethod::Copy :
+           crow::black_magic::is_equ_p(str, "LOCK", 4)      ? crow::HTTPMethod::Lock :
+           crow::black_magic::is_equ_p(str, "MKCOL", 5)     ? crow::HTTPMethod::MkCol :
+           crow::black_magic::is_equ_p(str, "MOVE", 4)      ? crow::HTTPMethod::Move :
+           crow::black_magic::is_equ_p(str, "PROPFIND", 8)  ? crow::HTTPMethod::Propfind :
+           crow::black_magic::is_equ_p(str, "PROPPATCH", 9) ? crow::HTTPMethod::Proppatch :
+           crow::black_magic::is_equ_p(str, "SEARCH", 6)    ? crow::HTTPMethod::Search :
+           crow::black_magic::is_equ_p(str, "UNLOCK", 6)    ? crow::HTTPMethod::Unlock :
+           crow::black_magic::is_equ_p(str, "BIND", 4)      ? crow::HTTPMethod::Bind :
+           crow::black_magic::is_equ_p(str, "REBIND", 6)    ? crow::HTTPMethod::Rebind :
+           crow::black_magic::is_equ_p(str, "UNBIND", 6)    ? crow::HTTPMethod::Unbind :
+           crow::black_magic::is_equ_p(str, "ACL", 3)       ? crow::HTTPMethod::Acl :
+
+           crow::black_magic::is_equ_p(str, "REPORT", 6)      ? crow::HTTPMethod::Report :
+           crow::black_magic::is_equ_p(str, "MKACTIVITY", 10) ? crow::HTTPMethod::MkActivity :
+           crow::black_magic::is_equ_p(str, "CHECKOUT", 8)    ? crow::HTTPMethod::Checkout :
+           crow::black_magic::is_equ_p(str, "MERGE", 5)       ? crow::HTTPMethod::Merge :
+
+           crow::black_magic::is_equ_p(str, "MSEARCH", 7)      ? crow::HTTPMethod::MSearch :
+           crow::black_magic::is_equ_p(str, "NOTIFY", 6)       ? crow::HTTPMethod::Notify :
+           crow::black_magic::is_equ_p(str, "SUBSCRIBE", 9)    ? crow::HTTPMethod::Subscribe :
+           crow::black_magic::is_equ_p(str, "UNSUBSCRIBE", 11) ? crow::HTTPMethod::Unsubscribe :
+
+           crow::black_magic::is_equ_p(str, "MKCALENDAR", 10) ? crow::HTTPMethod::MkCalendar :
+
+           crow::black_magic::is_equ_p(str, "LINK", 4)   ? crow::HTTPMethod::Link :
+           crow::black_magic::is_equ_p(str, "UNLINK", 6) ? crow::HTTPMethod::Unlink :
+
+           crow::black_magic::is_equ_p(str, "SOURCE", 6) ? crow::HTTPMethod::Source :
+                                                           throw std::runtime_error("invalid http method");
+}
+
+constexpr crow::HTTPMethod operator"" _method(const char* str, size_t /*len*/)
+{
+    return method_from_string( str );
+}
+#endif
+// clang-format on
+
+
+#ifdef CROW_USE_BOOST
+#include <boost/asio.hpp>
+#else
+#ifndef ASIO_STANDALONE
+#define ASIO_STANDALONE
+#endif
+#include <asio.hpp>
+#endif
+
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+#endif
+
+    /// Find and return the value associated with the key. (returns an empty string if nothing is found)
+    template<typename T>
+    inline const std::string& get_header_value(const T& headers, const std::string& key)
+    {
+        if (headers.count(key))
+        {
+            return headers.find(key)->second;
+        }
+        static std::string empty;
+        return empty;
+    }
+
+    /// An HTTP request.
+    struct request
+    {
+        HTTPMethod method;
+        std::string raw_url;     ///< The full URL containing the `?` and URL parameters.
+        std::string url;         ///< The endpoint without any parameters.
+        query_string url_params; ///< The parameters associated with the request. (everything after the `?` in the URL)
+        ci_map headers;
+        std::string body;
+        std::string remote_ip_address; ///< The IP address from which the request was sent.
+        unsigned char http_ver_major, http_ver_minor;
+        bool keep_alive,    ///< Whether or not the server should send a `connection: Keep-Alive` header to the client.
+          close_connection, ///< Whether or not the server should shut down the TCP connection once a response is sent.
+          upgrade;          ///< Whether or noth the server should change the HTTP connection to a different connection.
+
+        void* middleware_context{};
+        void* middleware_container{};
+        asio::io_service* io_service{};
+
+        /// Construct an empty request. (sets the method to `GET`)
+        request():
+          method(HTTPMethod::Get)
+        {}
+
+        /// Construct a request with all values assigned.
+        request(HTTPMethod method, std::string raw_url, std::string url, query_string url_params, ci_map headers, std::string body, unsigned char http_major, unsigned char http_minor, bool has_keep_alive, bool has_close_connection, bool is_upgrade):
+          method(method), raw_url(std::move(raw_url)), url(std::move(url)), url_params(std::move(url_params)), headers(std::move(headers)), body(std::move(body)), http_ver_major(http_major), http_ver_minor(http_minor), keep_alive(has_keep_alive), close_connection(has_close_connection), upgrade(is_upgrade)
+        {}
+
+        void add_header(std::string key, std::string value)
+        {
+            headers.emplace(std::move(key), std::move(value));
+        }
+
+        const std::string& get_header_value(const std::string& key) const
+        {
+            return crow::get_header_value(headers, key);
+        }
+
+        bool check_version(unsigned char major, unsigned char minor) const
+        {
+            return http_ver_major == major && http_ver_minor == minor;
+        }
+
+        /// Get the body as parameters in QS format.
+
+        ///
+        /// This is meant to be used with requests of type "application/x-www-form-urlencoded"
+        const query_string get_body_params() const
+        {
+            return query_string(body, false);
+        }
+
+        /// Send data to whoever made this request with a completion handler and return immediately.
+        template<typename CompletionHandler>
+        void post(CompletionHandler handler)
+        {
+            io_service->post(handler);
+        }
+
+        /// Send data to whoever made this request with a completion handler.
+        template<typename CompletionHandler>
+        void dispatch(CompletionHandler handler)
+        {
+            io_service->dispatch(handler);
+        }
+    };
+} // namespace crow
+
+
+#include <string>
+#include <vector>
+#include <sstream>
+
+
+namespace crow
+{
+
+    /// Encapsulates anything related to processing and organizing `multipart/xyz` messages
+    namespace multipart
+    {
+
+        const std::string dd = "--";
+
+        /// The first part in a section, contains metadata about the part
+        struct header
+        {
+            std::string value;                                   ///< The first part of the header, usually `Content-Type` or `Content-Disposition`
+            std::unordered_map<std::string, std::string> params; ///< The parameters of the header, come after the `value`
+
+            operator int() const { return std::stoi(value); }    ///< Returns \ref value as integer
+            operator double() const { return std::stod(value); } ///< Returns \ref value as double
+        };
+
+        /// Multipart header map (key is header key).
+        using mph_map = std::unordered_multimap<std::string, header, ci_hash, ci_key_eq>;
+
+        /// Find and return the value object associated with the key. (returns an empty class if nothing is found)
+        template<typename O, typename T>
+        inline const O& get_header_value_object(const T& headers, const std::string& key)
+        {
+            if (headers.count(key))
+            {
+                return headers.find(key)->second;
+            }
+            static O empty;
+            return empty;
+        }
+
+        /// Same as \ref get_header_value_object() but for \ref multipart.header
+        template<typename T>
+        inline const header& get_header_object(const T& headers, const std::string& key)
+        {
+            return get_header_value_object<header>(headers, key);
+        }
+
+        ///One part of the multipart message
+
+        ///
+        /// It is usually separated from other sections by a `boundary`
+        struct part
+        {
+            mph_map headers;  ///< (optional) The first part before the data, Contains information regarding the type of data and encoding
+            std::string body; ///< The actual data in the part
+
+            operator int() const { return std::stoi(body); }    ///< Returns \ref body as integer
+            operator double() const { return std::stod(body); } ///< Returns \ref body as double
+
+            const header& get_header_object(const std::string& key) const
+            {
+                return multipart::get_header_object(headers, key);
+            }
+        };
+
+        /// Multipart map (key is the name parameter).
+        using mp_map = std::unordered_multimap<std::string, part, ci_hash, ci_key_eq>;
+
+        /// The parsed multipart request/response
+        struct message : public returnable
+        {
+            ci_map headers;          ///< The request/response headers
+            std::string boundary;    ///< The text boundary that separates different `parts`
+            std::vector<part> parts; ///< The individual parts of the message
+            mp_map part_map;         ///< The individual parts of the message, organized in a map with the `name` header parameter being the key
+
+            const std::string& get_header_value(const std::string& key) const
+            {
+                return crow::get_header_value(headers, key);
+            }
+
+            part get_part_by_name(const std::string& name)
+            {
+                mp_map::iterator result = part_map.find(name);
+                if (result != part_map.end())
+                    return result->second;
+                else
+                    return {};
+            }
+
+            /// Represent all parts as a string (**does not include message headers**)
+            std::string dump() const override
+            {
+                std::stringstream str;
+                std::string delimiter = dd + boundary;
+
+                for (unsigned i = 0; i < parts.size(); i++)
+                {
+                    str << delimiter << crlf;
+                    str << dump(i);
+                }
+                str << delimiter << dd << crlf;
+                return str.str();
+            }
+
+            /// Represent an individual part as a string
+            std::string dump(int part_) const
+            {
+                std::stringstream str;
+                part item = parts[part_];
+                for (auto& item_h : item.headers)
+                {
+                    str << item_h.first << ": " << item_h.second.value;
+                    for (auto& it : item_h.second.params)
+                    {
+                        str << "; " << it.first << '=' << pad(it.second);
+                    }
+                    str << crlf;
+                }
+                str << crlf;
+                str << item.body << crlf;
+                return str.str();
+            }
+
+            /// Default constructor using default values
+            message(const ci_map& headers, const std::string& boundary, const std::vector<part>& sections):
+              returnable("multipart/form-data; boundary=CROW-BOUNDARY"), headers(headers), boundary(boundary), parts(sections)
+            {
+                if (!boundary.empty())
+                    content_type = "multipart/form-data; boundary=" + boundary;
+                for (auto& item : parts)
+                {
+                    part_map.emplace(
+                      (get_header_object(item.headers, "Content-Disposition").params.find("name")->second),
+                      item);
+                }
+            }
+
+            /// Create a multipart message from a request data
+            message(const request& req):
+              returnable("multipart/form-data; boundary=CROW-BOUNDARY"),
+              headers(req.headers),
+              boundary(get_boundary(get_header_value("Content-Type")))
+            {
+                if (!boundary.empty())
+                    content_type = "multipart/form-data; boundary=" + boundary;
+                parse_body(req.body, parts, part_map);
+            }
+
+        private:
+            std::string get_boundary(const std::string& header) const
+            {
+                constexpr char boundary_text[] = "boundary=";
+                size_t found = header.find(boundary_text);
+                if (found != std::string::npos)
+                {
+                    std::string to_return(header.substr(found + strlen(boundary_text)));
+                    if (to_return[0] == '\"')
+                    {
+                        to_return = to_return.substr(1, to_return.length() - 2);
+                    }
+                    return to_return;
+                }
+                return std::string();
+            }
+
+            void parse_body(std::string body, std::vector<part>& sections, mp_map& part_map)
+            {
+
+                std::string delimiter = dd + boundary;
+
+                // TODO(EDev): Exit on error
+                while (body != (crlf))
+                {
+                    size_t found = body.find(delimiter);
+                    if (found == std::string::npos)
+                    {
+                        // did not find delimiter; probably an ill-formed body; ignore the rest
+                        break;
+                    }
+                    std::string section = body.substr(0, found);
+
+                    // +2 is the CRLF.
+                    // We don't check it and delete it so that the same delimiter can be used for The last delimiter (--delimiter--CRLF).
+                    body.erase(0, found + delimiter.length() + 2);
+                    if (!section.empty())
+                    {
+                        part parsed_section(parse_section(section));
+                        part_map.emplace(
+                          (get_header_object(parsed_section.headers, "Content-Disposition").params.find("name")->second),
+                          parsed_section);
+                        sections.push_back(std::move(parsed_section));
+                    }
+                }
+            }
+
+            part parse_section(std::string& section)
+            {
+                struct part to_return;
+
+                size_t found = section.find(crlf + crlf);
+                std::string head_line = section.substr(0, found + 2);
+                section.erase(0, found + 4);
+
+                parse_section_head(head_line, to_return);
+                to_return.body = section.substr(0, section.length() - 2);
+                return to_return;
+            }
+
+            void parse_section_head(std::string& lines, part& part)
+            {
+                while (!lines.empty())
+                {
+                    header to_add;
+
+                    size_t found = lines.find(crlf);
+                    std::string line = lines.substr(0, found);
+                    std::string key;
+                    lines.erase(0, found + 2);
+                    // Add the header if available
+                    if (!line.empty())
+                    {
+                        size_t found = line.find("; ");
+                        std::string header = line.substr(0, found);
+                        if (found != std::string::npos)
+                            line.erase(0, found + 2);
+                        else
+                            line = std::string();
+
+                        size_t header_split = header.find(": ");
+                        key = header.substr(0, header_split);
+
+                        to_add.value = header.substr(header_split + 2);
+                    }
+
+                    // Add the parameters
+                    while (!line.empty())
+                    {
+                        size_t found = line.find("; ");
+                        std::string param = line.substr(0, found);
+                        if (found != std::string::npos)
+                            line.erase(0, found + 2);
+                        else
+                            line = std::string();
+
+                        size_t param_split = param.find('=');
+
+                        std::string value = param.substr(param_split + 1);
+
+                        to_add.params.emplace(param.substr(0, param_split), trim(value));
+                    }
+                    part.headers.emplace(key, to_add);
+                }
+            }
+
+            inline std::string trim(std::string& string, const char& excess = '"') const
+            {
+                if (string.length() > 1 && string[0] == excess && string[string.length() - 1] == excess)
+                    return string.substr(1, string.length() - 2);
+                return string;
+            }
+
+            inline std::string pad(std::string& string, const char& padding = '"') const
+            {
+                return (padding + string + padding);
+            }
+        };
+    } // namespace multipart
+} // namespace crow
+
+/* merged revision: 5b951d74bd66ec9d38448e0a85b1cf8b85d97db3 */
+/* updated to     : e13b274770da9b82a1085dec29182acfea72e7a7 (beyond v2.9.5) */
+/* commits not included:
+ * 091ebb87783a58b249062540bbea07de2a11e9cf
+ * 6132d1fefa03f769a3979355d1f5da0b8889cad2
+ * 7ba312397c2a6c851a4b5efe6c1603b1e1bda6ff
+ * d7675453a6c03180572f084e95eea0d02df39164
+ * dff604db203986e532e5a679bafd0e7382c6bdd9 (Might be useful to actually add [upgrade requests with a body])
+ * e01811e7f4894d7f0f7f4bd8492cccec6f6b4038 (related to above)
+ * 05525c5fde1fc562481f6ae08fa7056185325daf (also related to above)
+ * 350258965909f249f9c59823aac240313e0d0120 (cannot be implemented due to upgrade)
+ */
+
+// clang-format off
+extern "C" {
+#include <stddef.h>
+#if defined(_WIN32) && !defined(__MINGW32__) && \
+  (!defined(_MSC_VER) || _MSC_VER<1600) && !defined(__WINE__)
+#include <BaseTsd.h>
+typedef __int8 int8_t;
+typedef unsigned __int8 uint8_t;
+typedef __int16 int16_t;
+typedef unsigned __int16 uint16_t;
+typedef __int32 int32_t;
+typedef unsigned __int32 uint32_t;
+typedef __int64 int64_t;
+typedef unsigned __int64 uint64_t;
+#elif (defined(__sun) || defined(__sun__)) && defined(__SunOS_5_9)
+#include <sys/inttypes.h>
+#else
+#include <stdint.h>
+#endif
+#include <assert.h>
+#include <ctype.h>
+#include <string.h>
+#include <limits.h>
+}
+
+namespace crow
+{
+/* Maximium header size allowed. If the macro is not defined
+ * before including this header then the default is used. To
+ * change the maximum header size, define the macro in the build
+ * environment (e.g. -DHTTP_MAX_HEADER_SIZE=<value>). To remove
+ * the effective limit on the size of the header, define the macro
+ * to a very large number (e.g. -DCROW_HTTP_MAX_HEADER_SIZE=0x7fffffff)
+ */
+#ifndef CROW_HTTP_MAX_HEADER_SIZE
+# define CROW_HTTP_MAX_HEADER_SIZE (80*1024)
+#endif
+
+typedef struct http_parser http_parser;
+typedef struct http_parser_settings http_parser_settings;
+
+/* Callbacks should return non-zero to indicate an error. The parser will
+ * then halt execution.
+ *
+ * The one exception is on_headers_complete. In a HTTP_RESPONSE parser
+ * returning '1' from on_headers_complete will tell the parser that it
+ * should not expect a body. This is used when receiving a response to a
+ * HEAD request which may contain 'Content-Length' or 'Transfer-Encoding:
+ * chunked' headers that indicate the presence of a body.
+ *
+ * Returning `2` from on_headers_complete will tell parser that it should not
+ * expect neither a body nor any futher responses on this connection. This is
+ * useful for handling responses to a CONNECT request which may not contain
+ * `Upgrade` or `Connection: upgrade` headers.
+ *
+ * http_data_cb does not return data chunks. It will be called arbitrarally
+ * many times for each string. E.G. you might get 10 callbacks for "on_url"
+ * each providing just a few characters more data.
+ */
+typedef int (*http_data_cb) (http_parser*, const char *at, size_t length);
+typedef int (*http_cb) (http_parser*);
+
+
+/* Flag values for http_parser.flags field */
+enum http_connection_flags // This is basically 7 booleans placed into 1 integer. Uses 4 bytes instead of n bytes (7 currently).
+  { F_CHUNKED               = 1 << 0 // 00000000 00000000 00000000 00000001
+  , F_CONNECTION_KEEP_ALIVE = 1 << 1 // 00000000 00000000 00000000 00000010
+  , F_CONNECTION_CLOSE      = 1 << 2 // 00000000 00000000 00000000 00000100
+  , F_TRAILING              = 1 << 3 // 00000000 00000000 00000000 00001000
+  , F_UPGRADE               = 1 << 4 // 00000000 00000000 00000000 00010000
+  , F_SKIPBODY              = 1 << 5 // 00000000 00000000 00000000 00100000
+  , F_CONTENTLENGTH         = 1 << 6 // 00000000 00000000 00000000 01000000
+  };
+
+
+/* Map for errno-related constants
+ *
+ * The provided argument should be a macro that takes 2 arguments.
+ */
+#define CROW_HTTP_ERRNO_MAP(CROW_XX)                                                    \
+  /* No error */                                                                        \
+  CROW_XX(OK, "success")                                                                \
+                                                                                        \
+  /* Callback-related errors */                                                         \
+  CROW_XX(CB_message_begin, "the on_message_begin callback failed")                     \
+  CROW_XX(CB_method, "the on_method callback failed")                                   \
+  CROW_XX(CB_url, "the \"on_url\" callback failed")                                     \
+  CROW_XX(CB_header_field, "the \"on_header_field\" callback failed")                   \
+  CROW_XX(CB_header_value, "the \"on_header_value\" callback failed")                   \
+  CROW_XX(CB_headers_complete, "the \"on_headers_complete\" callback failed")           \
+  CROW_XX(CB_body, "the \"on_body\" callback failed")                                   \
+  CROW_XX(CB_message_complete, "the \"on_message_complete\" callback failed")           \
+  CROW_XX(CB_status, "the \"on_status\" callback failed")                               \
+                                                                                        \
+  /* Parsing-related errors */                                                          \
+  CROW_XX(INVALID_EOF_STATE, "stream ended at an unexpected time")                      \
+  CROW_XX(HEADER_OVERFLOW, "too many header bytes seen; overflow detected")             \
+  CROW_XX(CLOSED_CONNECTION, "data received after completed connection: close message") \
+  CROW_XX(INVALID_VERSION, "invalid HTTP version")                                      \
+  CROW_XX(INVALID_STATUS, "invalid HTTP status code")                                   \
+  CROW_XX(INVALID_METHOD, "invalid HTTP method")                                        \
+  CROW_XX(INVALID_URL, "invalid URL")                                                   \
+  CROW_XX(INVALID_HOST, "invalid host")                                                 \
+  CROW_XX(INVALID_PORT, "invalid port")                                                 \
+  CROW_XX(INVALID_PATH, "invalid path")                                                 \
+  CROW_XX(INVALID_QUERY_STRING, "invalid query string")                                 \
+  CROW_XX(INVALID_FRAGMENT, "invalid fragment")                                         \
+  CROW_XX(LF_EXPECTED, "LF character expected")                                         \
+  CROW_XX(INVALID_HEADER_TOKEN, "invalid character in header")                          \
+  CROW_XX(INVALID_CONTENT_LENGTH, "invalid character in content-length header")         \
+  CROW_XX(UNEXPECTED_CONTENT_LENGTH, "unexpected content-length header")                \
+  CROW_XX(INVALID_CHUNK_SIZE, "invalid character in chunk size header")                 \
+  CROW_XX(INVALID_CONSTANT, "invalid constant string")                                  \
+  CROW_XX(INVALID_INTERNAL_STATE, "encountered unexpected internal state")              \
+  CROW_XX(STRICT, "strict mode assertion failed")                                       \
+  CROW_XX(UNKNOWN, "an unknown error occurred")                                         \
+  CROW_XX(INVALID_TRANSFER_ENCODING, "request has invalid transfer-encoding")           \
+
+
+/* Define CHPE_* values for each errno value above */
+#define CROW_HTTP_ERRNO_GEN(n, s) CHPE_##n,
+enum http_errno {
+  CROW_HTTP_ERRNO_MAP(CROW_HTTP_ERRNO_GEN)
+};
+#undef CROW_HTTP_ERRNO_GEN
+
+
+/* Get an http_errno value from an http_parser */
+#define CROW_HTTP_PARSER_ERRNO(p) ((enum http_errno)(p)->http_errno)
+
+
+    struct http_parser
+    {
+        /** PRIVATE **/
+        unsigned int flags : 7;                  /* F_* values from 'flags' enum; semi-public */
+        unsigned int state : 8;                  /* enum state from http_parser.c */
+        unsigned int header_state : 7;           /* enum header_state from http_parser.c */
+        unsigned int index : 5;                  /* index into current matcher */
+        unsigned int uses_transfer_encoding : 1; /* Transfer-Encoding header is present */
+        unsigned int allow_chunked_length : 1;   /* Allow headers with both `Content-Length` and `Transfer-Encoding: chunked` set */
+        unsigned int lenient_http_headers : 1;
+
+        uint32_t nread;          /* # bytes read in various scenarios */
+        uint64_t content_length; /* # bytes in body. `(uint64_t) -1` (all bits one) if no Content-Length header. */
+        unsigned long qs_point;
+
+        /** READ-ONLY **/
+        unsigned char http_major;
+        unsigned char http_minor;
+        unsigned int method : 8;       /* requests only */
+        unsigned int http_errno : 7;
+
+  /* 1 = Upgrade header was present and the parser has exited because of that.
+   * 0 = No upgrade header present.
+   * Should be checked when http_parser_execute() returns in addition to
+   * error checking.
+   */
+        unsigned int upgrade : 1;
+
+        /** PUBLIC **/
+        void* data; /* A pointer to get hook to the "connection" or "socket" object */
+    };
+
+
+    struct http_parser_settings
+    {
+        http_cb on_message_begin;
+        http_cb on_method;
+        http_data_cb on_url;
+        http_data_cb on_header_field;
+        http_data_cb on_header_value;
+        http_cb on_headers_complete;
+        http_data_cb on_body;
+        http_cb on_message_complete;
+    };
+
+
+
+// SOURCE (.c) CODE
+static uint32_t max_header_size = CROW_HTTP_MAX_HEADER_SIZE;
+
+#ifndef CROW_ULLONG_MAX
+# define CROW_ULLONG_MAX ((uint64_t) -1) /* 2^64-1 */
+#endif
+
+#ifndef CROW_MIN
+# define CROW_MIN(a,b) ((a) < (b) ? (a) : (b))
+#endif
+
+#ifndef CROW_ARRAY_SIZE
+# define CROW_ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
+#endif
+
+#ifndef CROW_BIT_AT
+# define CROW_BIT_AT(a, i)                                           \
+  (!!((unsigned int) (a)[(unsigned int) (i) >> 3] &                  \
+   (1 << ((unsigned int) (i) & 7))))
+#endif
+
+#define CROW_SET_ERRNO(e)                                            \
+do {                                                                 \
+  parser->nread = nread;                                             \
+  parser->http_errno = (e);                                          \
+} while(0)
+
+/* Run the notify callback FOR, returning ER if it fails */
+#define CROW_CALLBACK_NOTIFY_(FOR, ER)                               \
+do {                                                                 \
+  assert(CROW_HTTP_PARSER_ERRNO(parser) == CHPE_OK);                 \
+                                                                     \
+  if (CROW_LIKELY(settings->on_##FOR)) {                             \
+    if (CROW_UNLIKELY(0 != settings->on_##FOR(parser))) {            \
+      CROW_SET_ERRNO(CHPE_CB_##FOR);                                 \
+    }                                                                \
+                                                                     \
+    /* We either errored above or got paused; get out */             \
+    if (CROW_UNLIKELY(CROW_HTTP_PARSER_ERRNO(parser) != CHPE_OK)) {  \
+      return (ER);                                                   \
+    }                                                                \
+  }                                                                  \
+} while (0)
+
+/* Run the notify callback FOR and consume the current byte */
+#define CROW_CALLBACK_NOTIFY(FOR)            CROW_CALLBACK_NOTIFY_(FOR, p - data + 1)
+
+/* Run the notify callback FOR and don't consume the current byte */
+#define CROW_CALLBACK_NOTIFY_NOADVANCE(FOR)  CROW_CALLBACK_NOTIFY_(FOR, p - data)
+
+/* Run data callback FOR with LEN bytes, returning ER if it fails */
+#define CROW_CALLBACK_DATA_(FOR, LEN, ER)                            \
+do {                                                                 \
+  assert(CROW_HTTP_PARSER_ERRNO(parser) == CHPE_OK);                 \
+                                                                     \
+  if (FOR##_mark) {                                                  \
+    if (CROW_LIKELY(settings->on_##FOR)) {                           \
+      if (CROW_UNLIKELY(0 !=                                         \
+          settings->on_##FOR(parser, FOR##_mark, (LEN)))) {          \
+        CROW_SET_ERRNO(CHPE_CB_##FOR);                               \
+      }                                                              \
+                                                                     \
+      /* We either errored above or got paused; get out */           \
+      if (CROW_UNLIKELY(CROW_HTTP_PARSER_ERRNO(parser) != CHPE_OK)) {\
+        return (ER);                                                 \
+      }                                                              \
+    }                                                                \
+    FOR##_mark = NULL;                                               \
+  }                                                                  \
+} while (0)
+
+/* Run the data callback FOR and consume the current byte */
+#define CROW_CALLBACK_DATA(FOR)                                      \
+    CROW_CALLBACK_DATA_(FOR, p - FOR##_mark, p - data + 1)
+
+/* Run the data callback FOR and don't consume the current byte */
+#define CROW_CALLBACK_DATA_NOADVANCE(FOR)                            \
+    CROW_CALLBACK_DATA_(FOR, p - FOR##_mark, p - data)
+
+/* Set the mark FOR; non-destructive if mark is already set */
+#define CROW_MARK(FOR)                                               \
+do {                                                                 \
+  if (!FOR##_mark) {                                                 \
+    FOR##_mark = p;                                                  \
+  }                                                                  \
+} while (0)
+
+/* Don't allow the total size of the HTTP headers (including the status
+ * line) to exceed max_header_size.  This check is here to protect
+ * embedders against denial-of-service attacks where the attacker feeds
+ * us a never-ending header that the embedder keeps buffering.
+ *
+ * This check is arguably the responsibility of embedders but we're doing
+ * it on the embedder's behalf because most won't bother and this way we
+ * make the web a little safer.  max_header_size is still far bigger
+ * than any reasonable request or response so this should never affect
+ * day-to-day operation.
+ */
+#define CROW_COUNT_HEADER_SIZE(V)                                    \
+do {                                                                 \
+  nread += (uint32_t)(V);                                            \
+  if (CROW_UNLIKELY(nread > max_header_size)) {                      \
+    CROW_SET_ERRNO(CHPE_HEADER_OVERFLOW);                            \
+    goto error;                                                      \
+  }                                                                  \
+} while (0)
+#define CROW_REEXECUTE()                                             \
+  goto reexecute;                                                    \
+
+#define CROW_PROXY_CONNECTION "proxy-connection"
+#define CROW_CONNECTION "connection"
+#define CROW_CONTENT_LENGTH "content-length"
+#define CROW_TRANSFER_ENCODING "transfer-encoding"
+#define CROW_UPGRADE "upgrade"
+#define CROW_CHUNKED "chunked"
+#define CROW_KEEP_ALIVE "keep-alive"
+#define CROW_CLOSE "close"
+
+
+
+    enum state
+    {
+        s_dead = 1 /* important that this is > 0 */
+
+        ,
+        s_start_req
+
+        ,
+        s_req_method,
+        s_req_spaces_before_url,
+        s_req_schema,
+        s_req_schema_slash,
+        s_req_schema_slash_slash,
+        s_req_server_start,
+        s_req_server,             // }
+        s_req_server_with_at,     // |
+        s_req_path,               // | The parser recognizes how to switch between these states,
+        s_req_query_string_start, // | however it doesn't process them any differently.
+        s_req_query_string,       // }
+        s_req_http_start,
+        s_req_http_H,
+        s_req_http_HT,
+        s_req_http_HTT,
+        s_req_http_HTTP,
+        s_req_http_I,
+        s_req_http_IC,
+        s_req_http_major,
+        s_req_http_dot,
+        s_req_http_minor,
+        s_req_http_end,
+        s_req_line_almost_done
+
+        ,
+        s_header_field_start,
+        s_header_field,
+        s_header_value_discard_ws,
+        s_header_value_discard_ws_almost_done,
+        s_header_value_discard_lws,
+        s_header_value_start,
+        s_header_value,
+        s_header_value_lws
+
+        ,
+        s_header_almost_done
+
+        ,
+        s_chunk_size_start,
+        s_chunk_size,
+        s_chunk_parameters,
+        s_chunk_size_almost_done
+
+        ,
+        s_headers_almost_done,
+        s_headers_done
+
+        /* Important: 's_headers_done' must be the last 'header' state. All
+         * states beyond this must be 'body' states. It is used for overflow
+         * checking. See the CROW_PARSING_HEADER() macro.
+         */
+
+        ,
+        s_chunk_data,
+        s_chunk_data_almost_done,
+        s_chunk_data_done
+
+        ,
+        s_body_identity,
+        s_body_identity_eof
+
+        ,
+        s_message_done
+    };
+
+
+#define CROW_PARSING_HEADER(state) (state <= s_headers_done)
+
+
+enum header_states
+  { h_general = 0
+  , h_C
+  , h_CO
+  , h_CON
+
+  , h_matching_connection
+  , h_matching_proxy_connection
+  , h_matching_content_length
+  , h_matching_transfer_encoding
+  , h_matching_upgrade
+
+  , h_connection
+  , h_content_length
+  , h_content_length_num
+  , h_content_length_ws
+  , h_transfer_encoding
+  , h_upgrade
+
+  , h_matching_transfer_encoding_token_start
+  , h_matching_transfer_encoding_chunked
+  , h_matching_transfer_encoding_token
+
+  , h_matching_connection_keep_alive
+  , h_matching_connection_close
+
+  , h_transfer_encoding_chunked
+  , h_connection_keep_alive
+  , h_connection_close
+  };
+
+enum http_host_state
+  {
+    s_http_host_dead = 1
+  , s_http_userinfo_start
+  , s_http_userinfo
+  , s_http_host_start
+  , s_http_host_v6_start
+  , s_http_host
+  , s_http_host_v6
+  , s_http_host_v6_end
+  , s_http_host_v6_zone_start
+  , s_http_host_v6_zone
+  , s_http_host_port_start
+  , s_http_host_port
+};
+
+/* Macros for character classes; depends on strict-mode  */
+#define CROW_LOWER(c)            (unsigned char)(c | 0x20)
+#define CROW_IS_ALPHA(c)         (CROW_LOWER(c) >= 'a' && CROW_LOWER(c) <= 'z')
+#define CROW_IS_NUM(c)           ((c) >= '0' && (c) <= '9')
+#define CROW_IS_ALPHANUM(c)      (CROW_IS_ALPHA(c) || CROW_IS_NUM(c))
+//#define CROW_IS_HEX(c)           (CROW_IS_NUM(c) || (CROW_LOWER(c) >= 'a' && CROW_LOWER(c) <= 'f'))
+#define CROW_IS_MARK(c)          ((c) == '-' || (c) == '_' || (c) == '.' || \
+  (c) == '!' || (c) == '~' || (c) == '*' || (c) == '\'' || (c) == '(' ||    \
+  (c) == ')')
+#define CROW_IS_USERINFO_CHAR(c) (CROW_IS_ALPHANUM(c) || CROW_IS_MARK(c) || (c) == '%' || \
+  (c) == ';' || (c) == ':' || (c) == '&' || (c) == '=' || (c) == '+' ||                   \
+  (c) == '$' || (c) == ',')
+
+#define CROW_TOKEN(c)            (tokens[(unsigned char)c])
+#define CROW_IS_URL_CHAR(c)      (CROW_BIT_AT(normal_url_char, (unsigned char)c))
+//#define CROW_IS_HOST_CHAR(c)     (CROW_IS_ALPHANUM(c) || (c) == '.' || (c) == '-')
+
+  /**
+ * Verify that a char is a valid visible (printable) US-ASCII
+ * character or %x80-FF
+ **/
+#define CROW_IS_HEADER_CHAR(ch)                                                     \
+  (ch == cr || ch == lf || ch == 9 || ((unsigned char)ch > 31 && ch != 127))
+
+#define CROW_start_state s_start_req
+
+# define CROW_STRICT_CHECK(cond)                                     \
+do {                                                                 \
+  if (cond) {                                                        \
+    CROW_SET_ERRNO(CHPE_STRICT);                                     \
+    goto error;                                                      \
+  }                                                                  \
+} while (0)
+#define CROW_NEW_MESSAGE() (CROW_start_state)
+
+/* Our URL parser.
+ *
+ * This is designed to be shared by http_parser_execute() for URL validation,
+ * hence it has a state transition + byte-for-byte interface. In addition, it
+ * is meant to be embedded in http_parser_parse_url(), which does the dirty
+ * work of turning state transitions URL components for its API.
+ *
+ * This function should only be invoked with non-space characters. It is
+ * assumed that the caller cares about (and can detect) the transition between
+ * URL and non-URL states by looking for these.
+ */
+inline enum state
+parse_url_char(enum state s, const char ch, http_parser *parser, const char* url_mark, const char* p)
+{
+# define CROW_T(v) 0
+
+
+static const uint8_t normal_url_char[32] = {
+/*   0 nul    1 soh    2 stx    3 etx    4 eot    5 enq    6 ack    7 bel  */
+        0    |   0    |   0    |   0    |   0    |   0    |   0    |   0,
+/*   8 bs     9 ht    10 nl    11 vt    12 np    13 cr    14 so    15 si   */
+        0    |CROW_T(2)|  0    |   0    |CROW_T(16)| 0    |   0    |   0,
+/*  16 dle   17 dc1   18 dc2   19 dc3   20 dc4   21 nak   22 syn   23 etb */
+        0    |   0    |   0    |   0    |   0    |   0    |   0    |   0,
+/*  24 can   25 em    26 sub   27 esc   28 fs    29 gs    30 rs    31 us  */
+        0    |   0    |   0    |   0    |   0    |   0    |   0    |   0,
+/*  32 sp    33  !    34  "    35  #    36  $    37  %    38  &    39  '  */
+        0    |   2    |   4    |   0    |   16   |   32   |   64   |  128,
+/*  40  (    41  )    42  *    43  +    44  ,    45  -    46  .    47  /  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  48  0    49  1    50  2    51  3    52  4    53  5    54  6    55  7  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  56  8    57  9    58  :    59  ;    60  <    61  =    62  >    63  ?  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |   0,
+/*  64  @    65  A    66  B    67  C    68  D    69  E    70  F    71  G  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  72  H    73  I    74  J    75  K    76  L    77  M    78  N    79  O  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  80  P    81  Q    82  R    83  S    84  CROW_T    85  U    86  V    87  W  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  88  X    89  Y    90  Z    91  [    92  \    93  ]    94  ^    95  _  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/*  96  `    97  a    98  b    99  c   100  d   101  e   102  f   103  g  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/* 104  h   105  i   106  j   107  k   108  l   109  m   110  n   111  o  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/* 112  p   113  q   114  r   115  s   116  t   117  u   118  v   119  w  */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |  128,
+/* 120  x   121  y   122  z   123  {   124  |   125  }   126  ~   127 del */
+        1    |   2    |   4    |   8    |   16   |   32   |   64   |   0, };
+
+#undef CROW_T
+
+  if (ch == ' ' || ch == '\r' || ch == '\n') {
+    return s_dead;
+  }
+  if (ch == '\t' || ch == '\f') {
+    return s_dead;
+  }
+
+  switch (s) {
+    case s_req_spaces_before_url:
+      /* Proxied requests are followed by scheme of an absolute URI (alpha).
+       * All methods except CONNECT are followed by '/' or '*'.
+       */
+
+      if (ch == '/' || ch == '*') {
+        return s_req_path;
+      }
+
+      if (CROW_IS_ALPHA(ch)) {
+        return s_req_schema;
+      }
+
+      break;
+
+    case s_req_schema:
+      if (CROW_IS_ALPHA(ch)) {
+        return s;
+      }
+
+      if (ch == ':') {
+        return s_req_schema_slash;
+      }
+
+      break;
+
+    case s_req_schema_slash:
+      if (ch == '/') {
+        return s_req_schema_slash_slash;
+      }
+
+      break;
+
+    case s_req_schema_slash_slash:
+      if (ch == '/') {
+        return s_req_server_start;
+      }
+
+      break;
+
+    case s_req_server_with_at:
+      if (ch == '@') {
+        return s_dead;
+      }
+
+    /* fall through */
+    case s_req_server_start:
+    case s_req_server:
+      if (ch == '/') {
+        return s_req_path;
+      }
+
+      if (ch == '?') {
+          parser->qs_point = p - url_mark;
+        return s_req_query_string_start;
+      }
+
+      if (ch == '@') {
+        return s_req_server_with_at;
+      }
+
+      if (CROW_IS_USERINFO_CHAR(ch) || ch == '[' || ch == ']') {
+        return s_req_server;
+      }
+
+      break;
+
+    case s_req_path:
+      if (CROW_IS_URL_CHAR(ch)) {
+        return s;
+      }
+      else if (ch == '?')
+      {
+          parser->qs_point = p - url_mark;
+          return s_req_query_string_start;
+      }
+
+      break;
+
+    case s_req_query_string_start:
+    case s_req_query_string:
+      if (CROW_IS_URL_CHAR(ch)) {
+        return s_req_query_string;
+      }
+      else if (ch == '?')
+      {
+          return s_req_query_string;
+      }
+
+      break;
+
+    default:
+      break;
+  }
+
+  /* We should never fall out of the switch above unless there's an error */
+  return s_dead;
+}
+
+inline size_t http_parser_execute (http_parser *parser,
+                            const http_parser_settings *settings,
+                            const char *data,
+                            size_t len)
+{
+
+/* Tokens as defined by rfc 2616. Also lowercases them.
+ *        token       = 1*<any CHAR except CTLs or separators>
+ *     separators     = "(" | ")" | "<" | ">" | "@"
+ *                    | "," | ";" | ":" | "\" | <">
+ *                    | "/" | "[" | "]" | "?" | "="
+ *                    | "{" | "}" | SP  | HT
+ */
+static const char tokens[256] = {
+/*   0 nul    1 soh    2 stx    3 etx    4 eot    5 enq    6 ack    7 bel  */
+        0,       0,       0,       0,       0,       0,       0,       0,
+/*   8 bs     9 ht    10 nl    11 vt    12 np    13 cr    14 so    15 si   */
+        0,       0,       0,       0,       0,       0,       0,       0,
+/*  16 dle   17 dc1   18 dc2   19 dc3   20 dc4   21 nak   22 syn   23 etb */
+        0,       0,       0,       0,       0,       0,       0,       0,
+/*  24 can   25 em    26 sub   27 esc   28 fs    29 gs    30 rs    31 us  */
+        0,       0,       0,       0,       0,       0,       0,       0,
+/*  32 sp    33  !    34  "    35  #    36  $    37  %    38  &    39  '  */
+        0,      '!',      0,      '#',     '$',     '%',     '&',    '\'',
+/*  40  (    41  )    42  *    43  +    44  ,    45  -    46  .    47  /  */
+        0,       0,      '*',     '+',      0,      '-',     '.',      0,
+/*  48  0    49  1    50  2    51  3    52  4    53  5    54  6    55  7  */
+       '0',     '1',     '2',     '3',     '4',     '5',     '6',     '7',
+/*  56  8    57  9    58  :    59  ;    60  <    61  =    62  >    63  ?  */
+       '8',     '9',      0,       0,       0,       0,       0,       0,
+/*  64  @    65  A    66  B    67  C    68  D    69  E    70  F    71  G  */
+        0,      'a',     'b',     'c',     'd',     'e',     'f',     'g',
+/*  72  H    73  I    74  J    75  K    76  L    77  M    78  N    79  O  */
+       'h',     'i',     'j',     'k',     'l',     'm',     'n',     'o',
+/*  80  P    81  Q    82  R    83  S    84  T    85  U    86  V    87  W  */
+       'p',     'q',     'r',     's',     't',     'u',     'v',     'w',
+/*  88  X    89  Y    90  Z    91  [    92  \    93  ]    94  ^    95  _  */
+       'x',     'y',     'z',      0,       0,       0,      '^',     '_',
+/*  96  `    97  a    98  b    99  c   100  d   101  e   102  f   103  g  */
+       '`',     'a',     'b',     'c',     'd',     'e',     'f',     'g',
+/* 104  h   105  i   106  j   107  k   108  l   109  m   110  n   111  o  */
+       'h',     'i',     'j',     'k',     'l',     'm',     'n',     'o',
+/* 112  p   113  q   114  r   115  s   116  t   117  u   118  v   119  w  */
+       'p',     'q',     'r',     's',     't',     'u',     'v',     'w',
+/* 120  x   121  y   122  z   123  {   124  |   125  }   126  ~   127 del */
+       'x',     'y',     'z',      0,      '|',      0,      '~',       0 };
+
+
+static const int8_t unhex[256] =
+  {-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  , 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,-1,-1,-1,-1,-1,-1
+  ,-1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  ,-1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  ,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
+  };
+
+
+
+  char c, ch;
+  int8_t unhex_val;
+  const char *p = data;
+  const char *header_field_mark = 0;
+  const char *header_value_mark = 0;
+  const char *url_mark = 0;
+  const char *url_start_mark = 0;
+  const char *body_mark = 0;
+  const unsigned int lenient = parser->lenient_http_headers;
+  const unsigned int allow_chunked_length = parser->allow_chunked_length;
+  
+  uint32_t nread = parser->nread;
+
+  /* We're in an error state. Don't bother doing anything. */
+  if (CROW_HTTP_PARSER_ERRNO(parser) != CHPE_OK) {
+    return 0;
+  }
+
+  if (len == 0) {
+    switch (parser->state) {
+      case s_body_identity_eof:
+        /* Use of CROW_CALLBACK_NOTIFY() here would erroneously return 1 byte read if we got paused. */
+        CROW_CALLBACK_NOTIFY_NOADVANCE(message_complete);
+        return 0;
+
+      case s_dead:
+      case s_start_req:
+        return 0;
+
+      default:
+        CROW_SET_ERRNO(CHPE_INVALID_EOF_STATE);
+        return 1;
+    }
+  }
+
+
+  if (parser->state == s_header_field)
+    header_field_mark = data;
+  if (parser->state == s_header_value)
+    header_value_mark = data;
+  switch (parser->state) {
+  case s_req_path:
+  case s_req_schema:
+  case s_req_schema_slash:
+  case s_req_schema_slash_slash:
+  case s_req_server_start:
+  case s_req_server:
+  case s_req_server_with_at:
+  case s_req_query_string_start:
+  case s_req_query_string:
+    url_mark = data;
+    break;
+  default:
+    break;
+  }
+
+  for (p=data; p != data + len; p++) {
+    ch = *p;
+
+    if (CROW_PARSING_HEADER(parser->state))
+      CROW_COUNT_HEADER_SIZE(1);
+
+reexecute:
+    switch (parser->state) {
+
+      case s_dead:
+        /* this state is used after a 'Connection: close' message
+         * the parser will error out if it reads another message
+         */
+        if (CROW_LIKELY(ch == cr || ch == lf))
+          break;
+
+        CROW_SET_ERRNO(CHPE_CLOSED_CONNECTION);
+        goto error;
+
+      case s_start_req:
+      {
+        if (ch == cr || ch == lf)
+          break;
+        parser->flags = 0;
+        parser->uses_transfer_encoding = 0;
+        parser->content_length = CROW_ULLONG_MAX;
+
+        if (CROW_UNLIKELY(!CROW_IS_ALPHA(ch))) {
+          CROW_SET_ERRNO(CHPE_INVALID_METHOD);
+          goto error;
+        }
+
+        parser->method = 0;
+        parser->index = 1;
+        switch (ch) {
+          case 'A': parser->method = (unsigned)HTTPMethod::Acl;                                                              break;
+          case 'B': parser->method = (unsigned)HTTPMethod::Bind;                                                             break;
+          case 'C': parser->method = (unsigned)HTTPMethod::Connect;   /* or COPY, CHECKOUT */                                break;
+          case 'D': parser->method = (unsigned)HTTPMethod::Delete;                                                           break;
+          case 'G': parser->method = (unsigned)HTTPMethod::Get;                                                              break;
+          case 'H': parser->method = (unsigned)HTTPMethod::Head;                                                             break;
+          case 'L': parser->method = (unsigned)HTTPMethod::Lock;      /* or LINK */                                          break;
+          case 'M': parser->method = (unsigned)HTTPMethod::MkCol;     /* or MOVE, MKACTIVITY, MERGE, M-SEARCH, MKCALENDAR */ break;
+          case 'N': parser->method = (unsigned)HTTPMethod::Notify;                                                           break;
+          case 'O': parser->method = (unsigned)HTTPMethod::Options;                                                          break;
+          case 'P': parser->method = (unsigned)HTTPMethod::Post;      /* or PROPFIND|PROPPATCH|PUT|PATCH|PURGE */            break;
+          case 'R': parser->method = (unsigned)HTTPMethod::Report;    /* or REBIND */                                        break;
+          case 'S': parser->method = (unsigned)HTTPMethod::Subscribe; /* or SEARCH, SOURCE */                                break;
+          case 'T': parser->method = (unsigned)HTTPMethod::Trace;                                                            break;
+          case 'U': parser->method = (unsigned)HTTPMethod::Unlock;    /* or UNSUBSCRIBE, UNBIND, UNLINK */                   break;
+          default:
+            CROW_SET_ERRNO(CHPE_INVALID_METHOD);
+            goto error;
+        }
+        parser->state = s_req_method;
+
+        CROW_CALLBACK_NOTIFY(message_begin);
+
+        break;
+      }
+
+      case s_req_method:
+      {
+        const char *matcher;
+        if (CROW_UNLIKELY(ch == '\0')) {
+          CROW_SET_ERRNO(CHPE_INVALID_METHOD);
+          goto error;
+        }
+
+        matcher = method_strings[parser->method];
+        if (ch == ' ' && matcher[parser->index] == '\0') {
+          parser->state = s_req_spaces_before_url;
+        } else if (ch == matcher[parser->index]) {
+          ; /* nada */
+        } else if ((ch >= 'A' && ch <= 'Z') || ch == '-') {
+
+          switch (parser->method << 16 | parser->index << 8 | ch) {
+#define CROW_XX(meth, pos, ch, new_meth) \
+            case ((unsigned)HTTPMethod::meth << 16 | pos << 8 | ch): \
+              parser->method = (unsigned)HTTPMethod::new_meth; break;
+
+            CROW_XX(Post,      1, 'U', Put)
+            CROW_XX(Post,      1, 'A', Patch)
+            CROW_XX(Post,      1, 'R', Propfind)
+            CROW_XX(Put,       2, 'R', Purge)
+            CROW_XX(Connect,   1, 'H', Checkout)
+            CROW_XX(Connect,   2, 'P', Copy)
+            CROW_XX(MkCol,     1, 'O', Move)
+            CROW_XX(MkCol,     1, 'E', Merge)
+            CROW_XX(MkCol,     1, '-', MSearch)
+            CROW_XX(MkCol,     2, 'A', MkActivity)
+            CROW_XX(MkCol,     3, 'A', MkCalendar)
+            CROW_XX(Subscribe, 1, 'E', Search)
+            CROW_XX(Subscribe, 1, 'O', Source)
+            CROW_XX(Report,    2, 'B', Rebind)
+            CROW_XX(Propfind,  4, 'P', Proppatch)
+            CROW_XX(Lock,      1, 'I', Link)
+            CROW_XX(Unlock,    2, 'S', Unsubscribe)
+            CROW_XX(Unlock,    2, 'B', Unbind)
+            CROW_XX(Unlock,    3, 'I', Unlink)
+#undef CROW_XX
+            default:
+              CROW_SET_ERRNO(CHPE_INVALID_METHOD);
+              goto error;
+          }
+        } else {
+          CROW_SET_ERRNO(CHPE_INVALID_METHOD);
+          goto error;
+        }
+
+        CROW_CALLBACK_NOTIFY_NOADVANCE(method);
+
+        ++parser->index;
+        break;
+      }
+
+      case s_req_spaces_before_url:
+      {
+        if (ch == ' ') break;
+
+        CROW_MARK(url);
+        CROW_MARK(url_start);
+        if (parser->method == (unsigned)HTTPMethod::Connect) {
+          parser->state = s_req_server_start;
+        }
+
+        parser->state = parse_url_char(static_cast<state>(parser->state), ch, parser, url_start_mark, p);
+        if (CROW_UNLIKELY(parser->state == s_dead)) {
+          CROW_SET_ERRNO(CHPE_INVALID_URL);
+          goto error;
+        }
+
+        break;
+      }
+
+      case s_req_schema:
+      case s_req_schema_slash:
+      case s_req_schema_slash_slash:
+      case s_req_server_start:
+      {
+        switch (ch) {
+          /* No whitespace allowed here */
+          case ' ':
+          case cr:
+          case lf:
+            CROW_SET_ERRNO(CHPE_INVALID_URL);
+            goto error;
+          default:
+            parser->state = parse_url_char(static_cast<state>(parser->state), ch, parser, url_start_mark, p);
+            if (CROW_UNLIKELY(parser->state == s_dead)) {
+              CROW_SET_ERRNO(CHPE_INVALID_URL);
+              goto error;
+            }
+        }
+
+        break;
+      }
+
+      case s_req_server:
+      case s_req_server_with_at:
+      case s_req_path:
+      case s_req_query_string_start:
+      case s_req_query_string:
+      {
+        switch (ch) {
+          case ' ':
+            parser->state = s_req_http_start;
+            CROW_CALLBACK_DATA(url);
+            break;
+          case cr: // No space after URL means no HTTP version. Which means the request is using HTTP/0.9
+          case lf:
+            if (CROW_UNLIKELY(parser->method != (unsigned)HTTPMethod::Get)) // HTTP/0.9 doesn't define any method other than GET
+            {
+              parser->state = s_dead;
+              CROW_SET_ERRNO(CHPE_INVALID_VERSION);
+              goto error;
+            }
+            parser->http_major = 0;
+            parser->http_minor = 9;
+            parser->state = (ch == cr) ?
+              s_req_line_almost_done :
+              s_header_field_start;
+            CROW_CALLBACK_DATA(url);
+            break;
+          default:
+            parser->state = parse_url_char(static_cast<state>(parser->state), ch, parser, url_start_mark, p);
+            if (CROW_UNLIKELY(parser->state == s_dead)) {
+              CROW_SET_ERRNO(CHPE_INVALID_URL);
+              goto error;
+            }
+        }
+        break;
+      }
+
+      case s_req_http_start:
+        switch (ch) {
+          case ' ':
+            break;
+          case 'H':
+            parser->state = s_req_http_H;
+            break;
+          case 'I':
+            if (parser->method == (unsigned)HTTPMethod::Source) {
+              parser->state = s_req_http_I;
+              break;
+            }
+            /* fall through */
+          default:
+            CROW_SET_ERRNO(CHPE_INVALID_CONSTANT);
+            goto error;
+        }
+        break;
+
+      case s_req_http_H:
+        CROW_STRICT_CHECK(ch != 'T');
+        parser->state = s_req_http_HT;
+        break;
+
+      case s_req_http_HT:
+        CROW_STRICT_CHECK(ch != 'T');
+        parser->state = s_req_http_HTT;
+        break;
+
+      case s_req_http_HTT:
+        CROW_STRICT_CHECK(ch != 'P');
+        parser->state = s_req_http_HTTP;
+        break;
+
+      case s_req_http_I:
+        CROW_STRICT_CHECK(ch != 'C');
+        parser->state = s_req_http_IC;
+        break;
+
+      case s_req_http_IC:
+        CROW_STRICT_CHECK(ch != 'E');
+        parser->state = s_req_http_HTTP;  /* Treat "ICE" as "HTTP". */
+        break;
+
+      case s_req_http_HTTP:
+        CROW_STRICT_CHECK(ch != '/');
+        parser->state = s_req_http_major;
+        break;
+
+      /* dot */
+      case s_req_http_major:
+        if (CROW_UNLIKELY(!CROW_IS_NUM(ch))) {
+          CROW_SET_ERRNO(CHPE_INVALID_VERSION);
+          goto error;
+        }
+
+        parser->http_major = ch - '0';
+        parser->state = s_req_http_dot;
+        break;
+
+      case s_req_http_dot:
+      {
+        if (CROW_UNLIKELY(ch != '.')) {
+          CROW_SET_ERRNO(CHPE_INVALID_VERSION);
+          goto error;
+        }
+
+        parser->state = s_req_http_minor;
+        break;
+      }
+
+      /* minor HTTP version */
+      case s_req_http_minor:
+        if (CROW_UNLIKELY(!CROW_IS_NUM(ch))) {
+          CROW_SET_ERRNO(CHPE_INVALID_VERSION);
+          goto error;
+        }
+
+        parser->http_minor = ch - '0';
+        parser->state = s_req_http_end;
+        break;
+
+      /* end of request line */
+      case s_req_http_end:
+      {
+        if (ch == cr) {
+          parser->state = s_req_line_almost_done;
+          break;
+        }
+
+        if (ch == lf) {
+          parser->state = s_header_field_start;
+          break;
+        }
+
+        CROW_SET_ERRNO(CHPE_INVALID_VERSION);
+        goto error;
+        break;
+      }
+
+      /* end of request line */
+      case s_req_line_almost_done:
+      {
+        if (CROW_UNLIKELY(ch != lf)) {
+          CROW_SET_ERRNO(CHPE_LF_EXPECTED);
+          goto error;
+        }
+
+        parser->state = s_header_field_start;
+        break;
+      }
+
+      case s_header_field_start:
+      {
+        if (ch == cr) {
+          parser->state = s_headers_almost_done;
+          break;
+        }
+
+        if (ch == lf) {
+          /* they might be just sending \n instead of \r\n so this would be
+           * the second \n to denote the end of headers*/
+          parser->state = s_headers_almost_done;
+          CROW_REEXECUTE();
+        }
+
+        c = CROW_TOKEN(ch);
+
+        if (CROW_UNLIKELY(!c)) {
+          CROW_SET_ERRNO(CHPE_INVALID_HEADER_TOKEN);
+          goto error;
+        }
+
+        CROW_MARK(header_field);
+
+        parser->index = 0;
+        parser->state = s_header_field;
+
+        switch (c) {
+          case 'c':
+            parser->header_state = h_C;
+            break;
+
+          case 'p':
+            parser->header_state = h_matching_proxy_connection;
+            break;
+
+          case 't':
+            parser->header_state = h_matching_transfer_encoding;
+            break;
+
+          case 'u':
+            parser->header_state = h_matching_upgrade;
+            break;
+
+          default:
+            parser->header_state = h_general;
+            break;
+        }
+        break;
+      }
+
+      case s_header_field:
+      {        
+        const char* start = p;
+        for (; p != data + len; p++) {
+          ch = *p;
+          c = CROW_TOKEN(ch);
+
+          if (!c)
+            break;
+          
+          switch (parser->header_state) {
+            case h_general: {
+              size_t left = data + len - p;
+              const char* pe = p + CROW_MIN(left, max_header_size);
+              while (p+1 < pe && CROW_TOKEN(p[1])) {
+                p++;
+              }
+              break;
+            }
+
+            case h_C:
+              parser->index++;
+              parser->header_state = (c == 'o' ? h_CO : h_general);
+              break;
+
+            case h_CO:
+              parser->index++;
+              parser->header_state = (c == 'n' ? h_CON : h_general);
+              break;
+
+            case h_CON:
+              parser->index++;
+              switch (c) {
+                case 'n':
+                  parser->header_state = h_matching_connection;
+                  break;
+                case 't':
+                  parser->header_state = h_matching_content_length;
+                  break;
+                default:
+                  parser->header_state = h_general;
+                  break;
+              }
+              break;
+
+            /* connection */
+
+            case h_matching_connection:
+              parser->index++;
+              if (parser->index > sizeof(CROW_CONNECTION)-1 || c != CROW_CONNECTION[parser->index]) {
+                parser->header_state = h_general;
+              } else if (parser->index == sizeof(CROW_CONNECTION)-2) {
+                parser->header_state = h_connection;
+              }
+              break;
+
+            /* proxy-connection */
+
+            case h_matching_proxy_connection:
+              parser->index++;
+              if (parser->index > sizeof(CROW_PROXY_CONNECTION)-1 || c != CROW_PROXY_CONNECTION[parser->index]) {
+                parser->header_state = h_general;
+              } else if (parser->index == sizeof(CROW_PROXY_CONNECTION)-2) {
+                parser->header_state = h_connection;
+              }
+              break;
+
+            /* content-length */
+
+            case h_matching_content_length:
+              parser->index++;
+              if (parser->index > sizeof(CROW_CONTENT_LENGTH)-1 || c != CROW_CONTENT_LENGTH[parser->index]) {
+                parser->header_state = h_general;
+              } else if (parser->index == sizeof(CROW_CONTENT_LENGTH)-2) {
+                parser->header_state = h_content_length;
+              }
+              break;
+
+            /* transfer-encoding */
+
+            case h_matching_transfer_encoding:
+              parser->index++;
+              if (parser->index > sizeof(CROW_TRANSFER_ENCODING)-1 || c != CROW_TRANSFER_ENCODING[parser->index]) {
+                parser->header_state = h_general;
+              } else if (parser->index == sizeof(CROW_TRANSFER_ENCODING)-2) {
+                parser->header_state = h_transfer_encoding;
+                parser->uses_transfer_encoding = 1;
+              }
+              break;
+
+            /* upgrade */
+
+            case h_matching_upgrade:
+              parser->index++;
+              if (parser->index > sizeof(CROW_UPGRADE)-1 || c != CROW_UPGRADE[parser->index]) {
+                parser->header_state = h_general;
+              } else if (parser->index == sizeof(CROW_UPGRADE)-2) {
+                parser->header_state = h_upgrade;
+              }
+              break;
+
+            case h_connection:
+            case h_content_length:
+            case h_transfer_encoding:
+            case h_upgrade:
+              if (ch != ' ') parser->header_state = h_general;
+              break;
+
+            default:
+              assert(0 && "Unknown header_state");
+              break;
+          }
+        }
+
+        if (p == data + len) {
+          --p;
+          CROW_COUNT_HEADER_SIZE(p - start);
+          break;
+        }
+
+        CROW_COUNT_HEADER_SIZE(p - start);
+
+        if (ch == ':') {
+          parser->state = s_header_value_discard_ws;
+          CROW_CALLBACK_DATA(header_field);
+          break;
+        }
+/* RFC-7230 Sec 3.2.4 expressly forbids line-folding in header field-names.
+        if (ch == cr) {
+          parser->state = s_header_almost_done;
+          CROW_CALLBACK_DATA(header_field);
+          break;
+        }
+
+        if (ch == lf) {
+          parser->state = s_header_field_start;
+          CROW_CALLBACK_DATA(header_field);
+          break;
+        }
+*/
+        CROW_SET_ERRNO(CHPE_INVALID_HEADER_TOKEN);
+        goto error;
+      }
+
+      case s_header_value_discard_ws:
+        if (ch == ' ' || ch == '\t') break;
+
+        if (ch == cr) {
+          parser->state = s_header_value_discard_ws_almost_done;
+          break;
+        }
+
+        if (ch == lf) {
+          parser->state = s_header_value_discard_lws;
+          break;
+        }
+
+        /* fall through */
+
+      case s_header_value_start:
+      {
+        CROW_MARK(header_value);
+
+        parser->state = s_header_value;
+        parser->index = 0;
+
+        c = CROW_LOWER(ch);
+
+        switch (parser->header_state) {
+          case h_upgrade:
+            // Crow does not support HTTP/2 at the moment.
+            // According to the RFC https://datatracker.ietf.org/doc/html/rfc7540#section-3.2
+            // "A server that does not support HTTP/2 can respond to the request as though the Upgrade header field were absent"
+            // => `F_UPGRADE` is not set if the header starts by "h2".
+            // This prevents the parser from skipping the request body.
+            if (ch != 'h' || p+1 == (data + len) || *(p+1) != '2') {
+              parser->flags |= F_UPGRADE;
+            }
+            parser->header_state = h_general;
+            break;
+
+          case h_transfer_encoding:
+            /* looking for 'Transfer-Encoding: chunked' */
+            if ('c' == c) {
+              parser->header_state = h_matching_transfer_encoding_chunked;
+            } else {
+              parser->header_state = h_matching_transfer_encoding_token;
+            }
+            break;
+            
+          /* Multi-value `Transfer-Encoding` header */
+          case h_matching_transfer_encoding_token_start:
+            break;
+
+          case h_content_length:
+            if (CROW_UNLIKELY(!CROW_IS_NUM(ch))) {
+              CROW_SET_ERRNO(CHPE_INVALID_CONTENT_LENGTH);
+              goto error;
+            }
+            
+            if (parser->flags & F_CONTENTLENGTH) {
+              CROW_SET_ERRNO(CHPE_UNEXPECTED_CONTENT_LENGTH);
+              goto error;
+            }
+            parser->flags |= F_CONTENTLENGTH;
+            parser->content_length = ch - '0';
+            parser->header_state = h_content_length_num;
+            break;
+
+          /* when obsolete line folding is encountered for content length
+           * continue to the s_header_value state */
+          case h_content_length_ws:
+            break;
+
+          case h_connection:
+            /* looking for 'Connection: keep-alive' */
+            if (c == 'k') {
+              parser->header_state = h_matching_connection_keep_alive;
+            /* looking for 'Connection: close' */
+            } else if (c == 'c') {
+              parser->header_state = h_matching_connection_close;
+            } else if (c == ' ' || c == '\t') {
+              /* Skip lws */
+            } else {
+              parser->header_state = h_general;
+            }
+            break;
+
+          default:
+            parser->header_state = h_general;
+            break;
+        }
+        break;
+      }
+
+      case s_header_value:
+      {
+        const char* start = p;
+        enum header_states h_state = static_cast<header_states>(parser->header_state);
+        for (; p != data + len; p++) {
+          ch = *p;
+
+          if (ch == cr) {
+            parser->state = s_header_almost_done;
+            parser->header_state = h_state;
+            CROW_CALLBACK_DATA(header_value);
+            break;
+          }
+
+          if (ch == lf) {
+            parser->state = s_header_almost_done;
+            CROW_COUNT_HEADER_SIZE(p - start);
+            parser->header_state = h_state;
+            CROW_CALLBACK_DATA_NOADVANCE(header_value);
+            CROW_REEXECUTE();
+          }
+          
+          if (!lenient && !CROW_IS_HEADER_CHAR(ch)) {
+            CROW_SET_ERRNO(CHPE_INVALID_HEADER_TOKEN);
+            goto error;
+          }
+          
+          c = CROW_LOWER(ch);
+
+          switch (h_state) {
+            case h_general:
+              {
+                size_t left = data + len - p;
+                const char* pe = p + CROW_MIN(left, max_header_size);
+
+                for (; p != pe; p++) {
+                  ch = *p;
+                  if (ch == cr || ch == lf) {
+                    --p;
+                    break;
+                  }
+                  if (!lenient && !CROW_IS_HEADER_CHAR(ch)) {
+                    CROW_SET_ERRNO(CHPE_INVALID_HEADER_TOKEN);
+                    goto error;
+                  }
+                }
+                if (p == data + len)
+                  --p;
+                break;
+              }
+
+            case h_connection:
+            case h_transfer_encoding:
+              assert(0 && "Shouldn't get here.");
+              break;
+
+            case h_content_length:
+              if (ch == ' ') break;
+              h_state = h_content_length_num;
+              /* fall through */
+
+            case h_content_length_num:
+            {
+              uint64_t t;
+
+              if (ch == ' ') {
+                h_state = h_content_length_ws;
+                break;
+              }
+
+              if (CROW_UNLIKELY(!CROW_IS_NUM(ch))) {
+                CROW_SET_ERRNO(CHPE_INVALID_CONTENT_LENGTH);
+                parser->header_state = h_state;
+                goto error;
+              }
+
+              t = parser->content_length;
+              t *= 10;
+              t += ch - '0';
+
+              /* Overflow? Test against a conservative limit for simplicity. */
+              if (CROW_UNLIKELY((CROW_ULLONG_MAX - 10) / 10 < parser->content_length)) {
+                CROW_SET_ERRNO(CHPE_INVALID_CONTENT_LENGTH);
+                parser->header_state = h_state;
+                goto error;
+              }
+
+              parser->content_length = t;
+              break;
+            }
+            
+            case h_content_length_ws:
+              if (ch == ' ') break;
+              CROW_SET_ERRNO(CHPE_INVALID_CONTENT_LENGTH);
+              parser->header_state = h_state;
+              goto error;
+
+            /* Transfer-Encoding: chunked */
+            case h_matching_transfer_encoding_token_start:
+              /* looking for 'Transfer-Encoding: chunked' */
+              if ('c' == c) {
+                h_state = h_matching_transfer_encoding_chunked;
+              } else if (CROW_TOKEN(c)) {
+                /* TODO(indutny): similar code below does this, but why?
+                 * At the very least it seems to be inconsistent given that
+                 * h_matching_transfer_encoding_token does not check for
+                 * `STRICT_TOKEN`
+                 */
+                h_state = h_matching_transfer_encoding_token;
+              } else if (c == ' ' || c == '\t') {
+                /* Skip lws */
+              } else {
+                h_state = h_general;
+              }
+              break;
+
+            case h_matching_transfer_encoding_chunked:
+              parser->index++;
+              if (parser->index > sizeof(CROW_CHUNKED)-1 || c != CROW_CHUNKED[parser->index]) {
+                h_state = h_matching_transfer_encoding_token;
+              } else if (parser->index == sizeof(CROW_CHUNKED)-2) {
+                h_state = h_transfer_encoding_chunked;
+              }
+              break;
+
+            case h_matching_transfer_encoding_token:
+              if (ch == ',') {
+                h_state = h_matching_transfer_encoding_token_start;
+                parser->index = 0;
+              }
+              break;
+
+            /* looking for 'Connection: keep-alive' */
+            case h_matching_connection_keep_alive:
+              parser->index++;
+              if (parser->index > sizeof(CROW_KEEP_ALIVE)-1 || c != CROW_KEEP_ALIVE[parser->index]) {
+                h_state = h_general;
+              } else if (parser->index == sizeof(CROW_KEEP_ALIVE)-2) {
+                h_state = h_connection_keep_alive;
+              }
+              break;
+
+            /* looking for 'Connection: close' */
+            case h_matching_connection_close:
+              parser->index++;
+              if (parser->index > sizeof(CROW_CLOSE)-1 || c != CROW_CLOSE[parser->index]) {
+                h_state = h_general;
+              } else if (parser->index == sizeof(CROW_CLOSE)-2) {
+                h_state = h_connection_close;
+              }
+              break;
+
+              // Edited from original (because of commits that werent included)
+            case h_transfer_encoding_chunked:
+              if (ch != ' ') h_state = h_matching_transfer_encoding_token;
+              break;
+            case h_connection_keep_alive:
+            case h_connection_close:
+              if (ch != ' ') h_state = h_general;
+              break;
+
+            default:
+              parser->state = s_header_value;
+              h_state = h_general;
+              break;
+          }
+        }
+        parser->header_state = h_state;
+        
+        
+        if (p == data + len)
+          --p;
+        
+        CROW_COUNT_HEADER_SIZE(p - start);
+        break;
+      }
+
+      case s_header_almost_done:
+      {
+        if (CROW_UNLIKELY(ch != lf)) {
+          CROW_SET_ERRNO(CHPE_LF_EXPECTED);
+          goto error;
+        }
+
+        parser->state = s_header_value_lws;
+        break;
+      }
+
+      case s_header_value_lws:
+      {
+        if (ch == ' ' || ch == '\t') {
+          if (parser->header_state == h_content_length_num) {
+              /* treat obsolete line folding as space */
+              parser->header_state = h_content_length_ws;
+          }
+          parser->state = s_header_value_start;
+          CROW_REEXECUTE();
+        }
+
+        /* finished the header */
+        switch (parser->header_state) {
+          case h_connection_keep_alive:
+            parser->flags |= F_CONNECTION_KEEP_ALIVE;
+            break;
+          case h_connection_close:
+            parser->flags |= F_CONNECTION_CLOSE;
+            break;
+          case h_transfer_encoding_chunked:
+            parser->flags |= F_CHUNKED;
+            break;
+          default:
+            break;
+        }
+
+        parser->state = s_header_field_start;
+        CROW_REEXECUTE();
+      }
+
+      case s_header_value_discard_ws_almost_done:
+      {
+        CROW_STRICT_CHECK(ch != lf);
+        parser->state = s_header_value_discard_lws;
+        break;
+      }
+
+      case s_header_value_discard_lws:
+      {
+        if (ch == ' ' || ch == '\t') {
+          parser->state = s_header_value_discard_ws;
+          break;
+        } else {
+          /* header value was empty */
+          CROW_MARK(header_value);
+          parser->state = s_header_field_start;
+          CROW_CALLBACK_DATA_NOADVANCE(header_value);
+          CROW_REEXECUTE();
+        }
+      }
+
+      case s_headers_almost_done:
+      {
+        CROW_STRICT_CHECK(ch != lf);
+
+        if (parser->flags & F_TRAILING) {
+          /* End of a chunked request */
+          CROW_CALLBACK_NOTIFY(message_complete);
+          break;
+        }
+        
+        /* Cannot use transfer-encoding and a content-length header together
+           per the HTTP specification. (RFC 7230 Section 3.3.3) */
+        if ((parser->uses_transfer_encoding == 1) &&
+            (parser->flags & F_CONTENTLENGTH)) {
+          /* Allow it for lenient parsing as long as `Transfer-Encoding` is
+           * not `chunked` or allow_length_with_encoding is set
+           */
+          if (parser->flags & F_CHUNKED) {
+            if (!allow_chunked_length) {
+              CROW_SET_ERRNO(CHPE_UNEXPECTED_CONTENT_LENGTH);
+              goto error;
+            }
+          } else if (!lenient) {
+            CROW_SET_ERRNO(CHPE_UNEXPECTED_CONTENT_LENGTH);
+            goto error;
+          }
+        }
+        
+        parser->state = s_headers_done;
+
+        /* Set this here so that on_headers_complete() callbacks can see it */
+        parser->upgrade =
+          (parser->flags & F_UPGRADE || parser->method == (unsigned)HTTPMethod::Connect);
+
+        /* Here we call the headers_complete callback. This is somewhat
+         * different than other callbacks because if the user returns 1, we
+         * will interpret that as saying that this message has no body. This
+         * is needed for the annoying case of recieving a response to a HEAD
+         * request.
+         *
+         * We'd like to use CROW_CALLBACK_NOTIFY_NOADVANCE() here but we cannot, so
+         * we have to simulate it by handling a change in errno below.
+         */
+        if (settings->on_headers_complete) {
+          switch (settings->on_headers_complete(parser)) {
+            case 0:
+              break;
+
+            case 2:
+              parser->upgrade = 1;
+              //break;
+
+            /* fall through */
+            case 1:
+              parser->flags |= F_SKIPBODY;
+              break;
+
+            default:
+              CROW_SET_ERRNO(CHPE_CB_headers_complete);
+              parser->nread = nread;
+              return p - data; /* Error */
+          }
+        }
+
+        if (CROW_HTTP_PARSER_ERRNO(parser) != CHPE_OK) {
+          parser->nread = nread;
+          return p - data;
+        }
+
+        CROW_REEXECUTE();
+      }
+
+      case s_headers_done:
+      {
+        CROW_STRICT_CHECK(ch != lf);
+
+        parser->nread = 0;
+        nread = 0;
+
+        /* Exit, the rest of the connect is in a different protocol. */
+        if (parser->upgrade) {
+          CROW_CALLBACK_NOTIFY(message_complete);
+          parser->nread = nread;
+          return (p - data) + 1;
+        }
+
+        if (parser->flags & F_SKIPBODY) {
+          CROW_CALLBACK_NOTIFY(message_complete);
+        } else if (parser->flags & F_CHUNKED) {
+          /* chunked encoding - ignore Content-Length header,
+           * prepare for a chunk */
+            parser->state = s_chunk_size_start;
+        }
+        else if (parser->uses_transfer_encoding == 1)
+        {
+            if (!lenient)
+            {
+                /* RFC 7230 3.3.3 */
+
+                /* If a Transfer-Encoding header field
+             * is present in a request and the chunked transfer coding is not
+             * the final encoding, the message body length cannot be determined
+             * reliably; the server MUST respond with the 400 (Bad Request)
+             * status code and then close the connection.
+             */
+                CROW_SET_ERRNO(CHPE_INVALID_TRANSFER_ENCODING);
+                parser->nread = nread;
+                return (p - data); /* Error */
+            }
+            else
+            {
+                /* RFC 7230 3.3.3 */
+
+                /* If a Transfer-Encoding header field is present in a response and
+             * the chunked transfer coding is not the final encoding, the
+             * message body length is determined by reading the connection until
+             * it is closed by the server.
+             */
+                parser->state = s_body_identity_eof;
+            }
+        }
+        else
+        {
+            if (parser->content_length == 0)
+            {
+                /* Content-Length header given but zero: Content-Length: 0\r\n */
+                CROW_CALLBACK_NOTIFY(message_complete);
+            }
+            else if (parser->content_length != CROW_ULLONG_MAX)
+            {
+                /* Content-Length header given and non-zero */
+                parser->state = s_body_identity;
+            }
+            else
+            {
+                /* Assume content-length 0 - read the next */
+                CROW_CALLBACK_NOTIFY(message_complete);
+            }
+        }
+
+        break;
+      }
+
+      case s_body_identity:
+      {
+        uint64_t to_read = CROW_MIN(parser->content_length,
+                               (uint64_t) ((data + len) - p));
+
+        assert(parser->content_length != 0
+            && parser->content_length != CROW_ULLONG_MAX);
+
+        /* The difference between advancing content_length and p is because
+         * the latter will automaticaly advance on the next loop iteration.
+         * Further, if content_length ends up at 0, we want to see the last
+         * byte again for our message complete callback.
+         */
+        CROW_MARK(body);
+        parser->content_length -= to_read;
+        p += to_read - 1;
+
+        if (parser->content_length == 0) {
+          parser->state = s_message_done;
+
+          /* Mimic CROW_CALLBACK_DATA_NOADVANCE() but with one extra byte.
+           *
+           * The alternative to doing this is to wait for the next byte to
+           * trigger the data callback, just as in every other case. The
+           * problem with this is that this makes it difficult for the test
+           * harness to distinguish between complete-on-EOF and
+           * complete-on-length. It's not clear that this distinction is
+           * important for applications, but let's keep it for now.
+           */
+          CROW_CALLBACK_DATA_(body, p - body_mark + 1, p - data);
+          CROW_REEXECUTE();
+        }
+
+        break;
+      }
+
+      /* read until EOF */
+      case s_body_identity_eof:
+        CROW_MARK(body);
+        p = data + len - 1;
+
+        break;
+
+      case s_message_done:
+        CROW_CALLBACK_NOTIFY(message_complete);
+        break;
+
+      case s_chunk_size_start:
+      {
+        assert(nread == 1);
+        assert(parser->flags & F_CHUNKED);
+
+        unhex_val = unhex[static_cast<unsigned char>(ch)];
+        if (CROW_UNLIKELY(unhex_val == -1)) {
+          CROW_SET_ERRNO(CHPE_INVALID_CHUNK_SIZE);
+          goto error;
+        }
+
+        parser->content_length = unhex_val;
+        parser->state = s_chunk_size;
+        break;
+      }
+
+      case s_chunk_size:
+      {
+        uint64_t t;
+
+        assert(parser->flags & F_CHUNKED);
+
+        if (ch == cr) {
+          parser->state = s_chunk_size_almost_done;
+          break;
+        }
+
+        unhex_val = unhex[static_cast<unsigned char>(ch)];
+
+        if (unhex_val == -1) {
+          if (ch == ';' || ch == ' ') {
+            parser->state = s_chunk_parameters;
+            break;
+          }
+
+          CROW_SET_ERRNO(CHPE_INVALID_CHUNK_SIZE);
+          goto error;
+        }
+
+        t = parser->content_length;
+        t *= 16;
+        t += unhex_val;
+
+        /* Overflow? Test against a conservative limit for simplicity. */
+        if (CROW_UNLIKELY((CROW_ULLONG_MAX - 16) / 16 < parser->content_length)) {
+          CROW_SET_ERRNO(CHPE_INVALID_CONTENT_LENGTH);
+          goto error;
+        }
+
+        parser->content_length = t;
+        break;
+      }
+
+      case s_chunk_parameters:
+      {
+        assert(parser->flags & F_CHUNKED);
+        /* just ignore this shit. TODO check for overflow */
+        if (ch == cr) {
+          parser->state = s_chunk_size_almost_done;
+          break;
+        }
+        break;
+      }
+
+      case s_chunk_size_almost_done:
+      {
+        assert(parser->flags & F_CHUNKED);
+        CROW_STRICT_CHECK(ch != lf);
+
+        parser->nread = 0;
+        nread = 0;
+
+        if (parser->content_length == 0) {
+          parser->flags |= F_TRAILING;
+          parser->state = s_header_field_start;
+        } else {
+          parser->state = s_chunk_data;
+        }
+        break;
+      }
+
+      case s_chunk_data:
+      {
+        uint64_t to_read = CROW_MIN(parser->content_length,
+                               (uint64_t) ((data + len) - p));
+
+        assert(parser->flags & F_CHUNKED);
+        assert(parser->content_length != 0
+            && parser->content_length != CROW_ULLONG_MAX);
+
+        /* See the explanation in s_body_identity for why the content
+         * length and data pointers are managed this way.
+         */
+        CROW_MARK(body);
+        parser->content_length -= to_read;
+        p += to_read - 1;
+
+        if (parser->content_length == 0) {
+          parser->state = s_chunk_data_almost_done;
+        }
+
+        break;
+      }
+
+      case s_chunk_data_almost_done:
+        assert(parser->flags & F_CHUNKED);
+        assert(parser->content_length == 0);
+        CROW_STRICT_CHECK(ch != cr);
+        parser->state = s_chunk_data_done;
+        CROW_CALLBACK_DATA(body);
+        break;
+
+      case s_chunk_data_done:
+        assert(parser->flags & F_CHUNKED);
+        CROW_STRICT_CHECK(ch != lf);
+        parser->nread = 0;
+        nread = 0;
+        parser->state = s_chunk_size_start;
+        break;
+
+      default:
+        assert(0 && "unhandled state");
+        CROW_SET_ERRNO(CHPE_INVALID_INTERNAL_STATE);
+        goto error;
+    }
+  }
+
+  /* Run callbacks for any marks that we have leftover after we ran out of
+   * bytes. There should be at most one of these set, so it's OK to invoke
+   * them in series (unset marks will not result in callbacks).
+   *
+   * We use the NOADVANCE() variety of callbacks here because 'p' has already
+   * overflowed 'data' and this allows us to correct for the off-by-one that
+   * we'd otherwise have (since CROW_CALLBACK_DATA() is meant to be run with a 'p'
+   * value that's in-bounds).
+   */
+
+  assert(((header_field_mark ? 1 : 0) +
+          (header_value_mark ? 1 : 0) +
+          (url_mark ? 1 : 0)  +
+          (body_mark ? 1 : 0)) <= 1);
+
+  CROW_CALLBACK_DATA_NOADVANCE(header_field);
+  CROW_CALLBACK_DATA_NOADVANCE(header_value);
+  CROW_CALLBACK_DATA_NOADVANCE(url);
+  CROW_CALLBACK_DATA_NOADVANCE(body);
+
+  parser->nread = nread;
+  return len;
+
+error:
+  if (CROW_HTTP_PARSER_ERRNO(parser) == CHPE_OK) {
+    CROW_SET_ERRNO(CHPE_UNKNOWN);
+  }
+
+  parser->nread = nread;
+  return (p - data);
+}
+
+inline void
+  http_parser_init(http_parser* parser)
+{
+  void *data = parser->data; /* preserve application data */
+  memset(parser, 0, sizeof(*parser));
+  parser->data = data;
+  parser->state = s_start_req;
+  parser->http_errno = CHPE_OK;
+}
+
+/* Return a string name of the given error */
+inline const char *
+http_errno_name(enum http_errno err) {
+/* Map errno values to strings for human-readable output */
+#define CROW_HTTP_STRERROR_GEN(n, s) { "CHPE_" #n, s },
+static struct {
+  const char *name;
+  const char *description;
+} http_strerror_tab[] = {
+  CROW_HTTP_ERRNO_MAP(CROW_HTTP_STRERROR_GEN)
+};
+#undef CROW_HTTP_STRERROR_GEN
+  assert(((size_t) err) < CROW_ARRAY_SIZE(http_strerror_tab));
+  return http_strerror_tab[err].name;
+}
+
+/* Return a string description of the given error */
+inline const char *
+http_errno_description(enum http_errno err) {
+/* Map errno values to strings for human-readable output */
+#define CROW_HTTP_STRERROR_GEN(n, s) { "CHPE_" #n, s },
+static struct {
+  const char *name;
+  const char *description;
+} http_strerror_tab[] = {
+  CROW_HTTP_ERRNO_MAP(CROW_HTTP_STRERROR_GEN)
+};
+#undef CROW_HTTP_STRERROR_GEN
+  assert(((size_t) err) < CROW_ARRAY_SIZE(http_strerror_tab));
+  return http_strerror_tab[err].description;
+}
+
+/* Checks if this is the final chunk of the body. */
+inline int
+http_body_is_final(const struct http_parser *parser) {
+    return parser->state == s_message_done;
+}
+
+/* Change the maximum header size provided at compile time. */
+inline void
+http_parser_set_max_header_size(uint32_t size) {
+  max_header_size = size;
+}
+
+#undef CROW_HTTP_ERRNO_MAP
+#undef CROW_SET_ERRNO
+#undef CROW_CALLBACK_NOTIFY_
+#undef CROW_CALLBACK_NOTIFY
+#undef CROW_CALLBACK_NOTIFY_NOADVANCE
+#undef CROW_CALLBACK_DATA_
+#undef CROW_CALLBACK_DATA
+#undef CROW_CALLBACK_DATA_NOADVANCE
+#undef CROW_MARK
+#undef CROW_PROXY_CONNECTION
+#undef CROW_CONNECTION
+#undef CROW_CONTENT_LENGTH
+#undef CROW_TRANSFER_ENCODING
+#undef CROW_UPGRADE
+#undef CROW_CHUNKED
+#undef CROW_KEEP_ALIVE
+#undef CROW_CLOSE
+#undef CROW_PARSING_HEADER
+#undef CROW_LOWER
+#undef CROW_IS_ALPHA
+#undef CROW_IS_NUM
+#undef CROW_IS_ALPHANUM
+//#undef CROW_IS_HEX
+#undef CROW_IS_MARK
+#undef CROW_IS_USERINFO_CHAR
+#undef CROW_TOKEN
+#undef CROW_IS_URL_CHAR
+//#undef CROW_IS_HOST_CHAR
+#undef CROW_STRICT_CHECK
+
+}
+
+// clang-format on
+
+
+#include <string>
+#include <unordered_map>
+#include <algorithm>
+
+
+namespace crow
+{
+    /// A wrapper for `nodejs/http-parser`.
+
+    ///
+    /// Used to generate a \ref crow.request from the TCP socket buffer.
+    template<typename Handler>
+    struct HTTPParser : public http_parser
+    {
+        static int on_message_begin(http_parser*)
+        {
+            return 0;
+        }
+        static int on_method(http_parser* self_)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            self->req.method = static_cast<HTTPMethod>(self->method);
+
+            return 0;
+        }
+        static int on_url(http_parser* self_, const char* at, size_t length)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            self->req.raw_url.insert(self->req.raw_url.end(), at, at + length);
+            self->req.url_params = query_string(self->req.raw_url);
+            self->req.url = self->req.raw_url.substr(0, self->qs_point != 0 ? self->qs_point : std::string::npos);
+
+            self->process_url();
+
+            return 0;
+        }
+        static int on_header_field(http_parser* self_, const char* at, size_t length)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            switch (self->header_building_state)
+            {
+                case 0:
+                    if (!self->header_value.empty())
+                    {
+                        self->req.headers.emplace(std::move(self->header_field), std::move(self->header_value));
+                    }
+                    self->header_field.assign(at, at + length);
+                    self->header_building_state = 1;
+                    break;
+                case 1:
+                    self->header_field.insert(self->header_field.end(), at, at + length);
+                    break;
+            }
+            return 0;
+        }
+        static int on_header_value(http_parser* self_, const char* at, size_t length)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            switch (self->header_building_state)
+            {
+                case 0:
+                    self->header_value.insert(self->header_value.end(), at, at + length);
+                    break;
+                case 1:
+                    self->header_building_state = 0;
+                    self->header_value.assign(at, at + length);
+                    break;
+            }
+            return 0;
+        }
+        static int on_headers_complete(http_parser* self_)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            if (!self->header_field.empty())
+            {
+                self->req.headers.emplace(std::move(self->header_field), std::move(self->header_value));
+            }
+
+            self->set_connection_parameters();
+
+            self->process_header();
+            return 0;
+        }
+        static int on_body(http_parser* self_, const char* at, size_t length)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+            self->req.body.insert(self->req.body.end(), at, at + length);
+            return 0;
+        }
+        static int on_message_complete(http_parser* self_)
+        {
+            HTTPParser* self = static_cast<HTTPParser*>(self_);
+
+            self->message_complete = true;
+            self->process_message();
+            return 0;
+        }
+        HTTPParser(Handler* handler):
+          handler_(handler)
+        {
+            http_parser_init(this);
+        }
+
+        // return false on error
+        /// Parse a buffer into the different sections of an HTTP request.
+        bool feed(const char* buffer, int length)
+        {
+            if (message_complete)
+                return true;
+
+            const static http_parser_settings settings_{
+              on_message_begin,
+              on_method,
+              on_url,
+              on_header_field,
+              on_header_value,
+              on_headers_complete,
+              on_body,
+              on_message_complete,
+            };
+
+            int nparsed = http_parser_execute(this, &settings_, buffer, length);
+            if (http_errno != CHPE_OK)
+            {
+                return false;
+            }
+            return nparsed == length;
+        }
+
+        bool done()
+        {
+            return feed(nullptr, 0);
+        }
+
+        void clear()
+        {
+            req = crow::request();
+            header_field.clear();
+            header_value.clear();
+            header_building_state = 0;
+            qs_point = 0;
+            message_complete = false;
+            state = CROW_NEW_MESSAGE();
+        }
+
+        inline void process_url()
+        {
+            handler_->handle_url();
+        }
+
+        inline void process_header()
+        {
+            handler_->handle_header();
+        }
+
+        inline void process_message()
+        {
+            handler_->handle();
+        }
+
+        inline void set_connection_parameters()
+        {
+            req.http_ver_major = http_major;
+            req.http_ver_minor = http_minor;
+
+            //NOTE(EDev): it seems that the problem is with crow's policy on closing the connection for HTTP_VERSION < 1.0, the behaviour for that in crow is "don't close the connection, but don't send a keep-alive either"
+
+            // HTTP1.1 = always send keep_alive, HTTP1.0 = only send if header exists, HTTP?.? = never send
+            req.keep_alive = (http_major == 1 && http_minor == 0) ?
+                               ((flags & F_CONNECTION_KEEP_ALIVE) ? true : false) :
+                               ((http_major == 1 && http_minor == 1) ? true : false);
+
+            // HTTP1.1 = only close if close header exists, HTTP1.0 = always close unless keep_alive header exists, HTTP?.?= never close
+            req.close_connection = (http_major == 1 && http_minor == 0) ?
+                                     ((flags & F_CONNECTION_KEEP_ALIVE) ? false : true) :
+                                     ((http_major == 1 && http_minor == 1) ? ((flags & F_CONNECTION_CLOSE) ? true : false) : false);
+            req.upgrade = static_cast<bool>(upgrade);
+        }
+
+        /// The final request that this parser outputs.
+        ///
+        /// Data parsed is put directly into this object as soon as the related callback returns. (e.g. the request will have the cooorect method as soon as on_method() returns)
+        request req;
+
+    private:
+        int header_building_state = 0;
+        bool message_complete = false;
+        std::string header_field;
+        std::string header_value;
+
+        Handler* handler_; ///< This is currently an HTTP connection object (\ref crow.Connection).
+    };
+} // namespace crow
+
+#undef CROW_NEW_MESSAGE
+#undef CROW_start_state
+
+
+
+#include <cstdio>
+#include <cstdlib>
+#include <ctime>
+#include <iostream>
+#include <sstream>
+#include <string>
+
+namespace crow
+{
+    enum class LogLevel
+    {
+#ifndef ERROR
+#ifndef DEBUG
+        DEBUG = 0,
+        INFO,
+        WARNING,
+        ERROR,
+        CRITICAL,
+#endif
+#endif
+
+        Debug = 0,
+        Info,
+        Warning,
+        Error,
+        Critical,
+    };
+
+    class ILogHandler
+    {
+    public:
+        virtual ~ILogHandler() = default;
+
+        virtual void log(std::string message, LogLevel level) = 0;
+    };
+
+    class CerrLogHandler : public ILogHandler
+    {
+    public:
+        void log(std::string message, LogLevel level) override
+        {
+            std::string prefix;
+            switch (level)
+            {
+                case LogLevel::Debug:
+                    prefix = "DEBUG   ";
+                    break;
+                case LogLevel::Info:
+                    prefix = "INFO    ";
+                    break;
+                case LogLevel::Warning:
+                    prefix = "WARNING ";
+                    break;
+                case LogLevel::Error:
+                    prefix = "ERROR   ";
+                    break;
+                case LogLevel::Critical:
+                    prefix = "CRITICAL";
+                    break;
+            }
+            std::cerr << std::string("(") + timestamp() + std::string(") [") + prefix + std::string("] ") + message << std::endl;
+        }
+
+    private:
+        static std::string timestamp()
+        {
+            char date[32];
+            time_t t = time(0);
+
+            tm my_tm;
+
+#if defined(_MSC_VER) || defined(__MINGW32__)
+#ifdef CROW_USE_LOCALTIMEZONE
+            localtime_s(&my_tm, &t);
+#else
+            gmtime_s(&my_tm, &t);
+#endif
+#else
+#ifdef CROW_USE_LOCALTIMEZONE
+            localtime_r(&t, &my_tm);
+#else
+            gmtime_r(&t, &my_tm);
+#endif
+#endif
+
+            size_t sz = strftime(date, sizeof(date), "%Y-%m-%d %H:%M:%S", &my_tm);
+            return std::string(date, date + sz);
+        }
+    };
+
+    class logger
+    {
+    public:
+        logger(LogLevel level):
+          level_(level)
+        {}
+        ~logger()
+        {
+#ifdef CROW_ENABLE_LOGGING
+            if (level_ >= get_current_log_level())
+            {
+                get_handler_ref()->log(stringstream_.str(), level_);
+            }
+#endif
+        }
+
+        //
+        template<typename T>
+        logger& operator<<(T const& value)
+        {
+#ifdef CROW_ENABLE_LOGGING
+            if (level_ >= get_current_log_level())
+            {
+                stringstream_ << value;
+            }
+#endif
+            return *this;
+        }
+
+        //
+        static void setLogLevel(LogLevel level) { get_log_level_ref() = level; }
+
+        static void setHandler(ILogHandler* handler) { get_handler_ref() = handler; }
+
+        static LogLevel get_current_log_level() { return get_log_level_ref(); }
+
+    private:
+        //
+        static LogLevel& get_log_level_ref()
+        {
+            static LogLevel current_level = static_cast<LogLevel>(CROW_LOG_LEVEL);
+            return current_level;
+        }
+        static ILogHandler*& get_handler_ref()
+        {
+            static CerrLogHandler default_handler;
+            static ILogHandler* current_handler = &default_handler;
+            return current_handler;
+        }
+
+        //
+        std::ostringstream stringstream_;
+        LogLevel level_;
+    };
+} // namespace crow
+
+#define CROW_LOG_CRITICAL                                                  \
+    if (crow::logger::get_current_log_level() <= crow::LogLevel::Critical) \
+    crow::logger(crow::LogLevel::Critical)
+#define CROW_LOG_ERROR                                                  \
+    if (crow::logger::get_current_log_level() <= crow::LogLevel::Error) \
+    crow::logger(crow::LogLevel::Error)
+#define CROW_LOG_WARNING                                                  \
+    if (crow::logger::get_current_log_level() <= crow::LogLevel::Warning) \
+    crow::logger(crow::LogLevel::Warning)
+#define CROW_LOG_INFO                                                  \
+    if (crow::logger::get_current_log_level() <= crow::LogLevel::Info) \
+    crow::logger(crow::LogLevel::Info)
+#define CROW_LOG_DEBUG                                                  \
+    if (crow::logger::get_current_log_level() <= crow::LogLevel::Debug) \
+    crow::logger(crow::LogLevel::Debug)
+
+
+//#define CROW_JSON_NO_ERROR_CHECK
+//#define CROW_JSON_USE_MAP
+
+#include <string>
+#ifdef CROW_JSON_USE_MAP
+#include <map>
+#else
+#include <unordered_map>
+#endif
+#include <iostream>
+#include <algorithm>
+#include <memory>
+#include <vector>
+#include <cmath>
+#include <cfloat>
+
+
+using std::isinf;
+using std::isnan;
+
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+    namespace mustache
+    {
+        class template_t;
+    }
+
+    namespace json
+    {
+        inline void escape(const std::string& str, std::string& ret)
+        {
+            ret.reserve(ret.size() + str.size() + str.size() / 4);
+            for (auto c : str)
+            {
+                switch (c)
+                {
+                    case '"': ret += "\\\""; break;
+                    case '\\': ret += "\\\\"; break;
+                    case '\n': ret += "\\n"; break;
+                    case '\b': ret += "\\b"; break;
+                    case '\f': ret += "\\f"; break;
+                    case '\r': ret += "\\r"; break;
+                    case '\t': ret += "\\t"; break;
+                    default:
+                        if (c >= 0 && c < 0x20)
+                        {
+                            ret += "\\u00";
+                            auto to_hex = [](char c) {
+                                c = c & 0xf;
+                                if (c < 10)
+                                    return '0' + c;
+                                return 'a' + c - 10;
+                            };
+                            ret += to_hex(c / 16);
+                            ret += to_hex(c % 16);
+                        }
+                        else
+                            ret += c;
+                        break;
+                }
+            }
+        }
+        inline std::string escape(const std::string& str)
+        {
+            std::string ret;
+            escape(str, ret);
+            return ret;
+        }
+
+        enum class type : char
+        {
+            Null,
+            False,
+            True,
+            Number,
+            String,
+            List,
+            Object,
+            Function
+        };
+
+        inline const char* get_type_str(type t)
+        {
+            switch (t)
+            {
+                case type::Number: return "Number";
+                case type::False: return "False";
+                case type::True: return "True";
+                case type::List: return "List";
+                case type::String: return "String";
+                case type::Object: return "Object";
+                case type::Function: return "Function";
+                default: return "Unknown";
+            }
+        }
+
+        enum class num_type : char
+        {
+            Signed_integer,
+            Unsigned_integer,
+            Floating_point,
+            Null,
+            Double_precision_floating_point
+        };
+
+        class rvalue;
+        rvalue load(const char* data, size_t size);
+
+        namespace detail
+        {
+            /// A read string implementation with comparison functionality.
+            struct r_string
+            {
+                r_string(){};
+                r_string(char* s, char* e):
+                  s_(s), e_(e){};
+                ~r_string()
+                {
+                    if (owned_)
+                        delete[] s_;
+                }
+
+                r_string(const r_string& r)
+                {
+                    *this = r;
+                }
+
+                r_string(r_string&& r)
+                {
+                    *this = r;
+                }
+
+                r_string& operator=(r_string&& r)
+                {
+                    s_ = r.s_;
+                    e_ = r.e_;
+                    owned_ = r.owned_;
+                    if (r.owned_)
+                        r.owned_ = 0;
+                    return *this;
+                }
+
+                r_string& operator=(const r_string& r)
+                {
+                    s_ = r.s_;
+                    e_ = r.e_;
+                    owned_ = 0;
+                    return *this;
+                }
+
+                operator std::string() const
+                {
+                    return std::string(s_, e_);
+                }
+
+
+                const char* begin() const { return s_; }
+                const char* end() const { return e_; }
+                size_t size() const { return end() - begin(); }
+
+                using iterator = const char*;
+                using const_iterator = const char*;
+
+                char* s_;         ///< Start.
+                mutable char* e_; ///< End.
+                uint8_t owned_{0};
+                friend std::ostream& operator<<(std::ostream& os, const r_string& s)
+                {
+                    os << static_cast<std::string>(s);
+                    return os;
+                }
+
+            private:
+                void force(char* s, uint32_t length)
+                {
+                    s_ = s;
+                    e_ = s_ + length;
+                    owned_ = 1;
+                }
+                friend rvalue crow::json::load(const char* data, size_t size);
+
+                friend bool operator==(const r_string& l, const r_string& r);
+                friend bool operator==(const std::string& l, const r_string& r);
+                friend bool operator==(const r_string& l, const std::string& r);
+
+                template<typename T, typename U>
+                inline static bool equals(const T& l, const U& r)
+                {
+                    if (l.size() != r.size())
+                        return false;
+
+                    for (size_t i = 0; i < l.size(); i++)
+                    {
+                        if (*(l.begin() + i) != *(r.begin() + i))
+                            return false;
+                    }
+
+                    return true;
+                }
+            };
+
+            inline bool operator<(const r_string& l, const r_string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator<(const r_string& l, const std::string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator<(const std::string& l, const r_string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator>(const r_string& l, const r_string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator>(const r_string& l, const std::string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator>(const std::string& l, const r_string& r)
+            {
+                return std::lexicographical_compare(l.begin(), l.end(), r.begin(), r.end());
+            }
+
+            inline bool operator==(const r_string& l, const r_string& r)
+            {
+                return r_string::equals(l, r);
+            }
+
+            inline bool operator==(const r_string& l, const std::string& r)
+            {
+                return r_string::equals(l, r);
+            }
+
+            inline bool operator==(const std::string& l, const r_string& r)
+            {
+                return r_string::equals(l, r);
+            }
+
+            inline bool operator!=(const r_string& l, const r_string& r)
+            {
+                return !(l == r);
+            }
+
+            inline bool operator!=(const r_string& l, const std::string& r)
+            {
+                return !(l == r);
+            }
+
+            inline bool operator!=(const std::string& l, const r_string& r)
+            {
+                return !(l == r);
+            }
+        } // namespace detail
+
+        /// JSON read value.
+
+        ///
+        /// Value can mean any json value, including a JSON object.
+        /// Read means this class is used to primarily read strings into a JSON value.
+        class rvalue
+        {
+            static const int cached_bit = 2;
+            static const int error_bit = 4;
+
+        public:
+            rvalue() noexcept:
+              option_{error_bit}
+            {
+            }
+            rvalue(type t) noexcept:
+              lsize_{}, lremain_{}, t_{t}
+            {
+            }
+            rvalue(type t, char* s, char* e) noexcept:
+              start_{s}, end_{e}, t_{t}
+            {
+                determine_num_type();
+            }
+
+            rvalue(const rvalue& r):
+              start_(r.start_), end_(r.end_), key_(r.key_), t_(r.t_), nt_(r.nt_), option_(r.option_)
+            {
+                copy_l(r);
+            }
+
+            rvalue(rvalue&& r) noexcept
+            {
+                *this = std::move(r);
+            }
+
+            rvalue& operator=(const rvalue& r)
+            {
+                start_ = r.start_;
+                end_ = r.end_;
+                key_ = r.key_;
+                t_ = r.t_;
+                nt_ = r.nt_;
+                option_ = r.option_;
+                copy_l(r);
+                return *this;
+            }
+            rvalue& operator=(rvalue&& r) noexcept
+            {
+                start_ = r.start_;
+                end_ = r.end_;
+                key_ = std::move(r.key_);
+                l_ = std::move(r.l_);
+                lsize_ = r.lsize_;
+                lremain_ = r.lremain_;
+                t_ = r.t_;
+                nt_ = r.nt_;
+                option_ = r.option_;
+                return *this;
+            }
+
+            explicit operator bool() const noexcept
+            {
+                return (option_ & error_bit) == 0;
+            }
+
+            explicit operator int64_t() const
+            {
+                return i();
+            }
+
+            explicit operator uint64_t() const
+            {
+                return u();
+            }
+
+            explicit operator int() const
+            {
+                return static_cast<int>(i());
+            }
+
+            /// Return any json value (not object or list) as a string.
+            explicit operator std::string() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() == type::Object || t() == type::List)
+                    throw std::runtime_error("json type container");
+#endif
+                switch (t())
+                {
+                    case type::String:
+                        return std::string(s());
+                    case type::Null:
+                        return std::string("null");
+                    case type::True:
+                        return std::string("true");
+                    case type::False:
+                        return std::string("false");
+                    default:
+                        return std::string(start_, end_ - start_);
+                }
+            }
+
+            /// The type of the JSON value.
+            type t() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (option_ & error_bit)
+                {
+                    throw std::runtime_error("invalid json object");
+                }
+#endif
+                return t_;
+            }
+
+            /// The number type of the JSON value.
+            num_type nt() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (option_ & error_bit)
+                {
+                    throw std::runtime_error("invalid json object");
+                }
+#endif
+                return nt_;
+            }
+
+            /// The integer value.
+            int64_t i() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                switch (t())
+                {
+                    case type::Number:
+                    case type::String:
+                        return utility::lexical_cast<int64_t>(start_, end_ - start_);
+                    default:
+                        const std::string msg = "expected number, got: " + std::string(get_type_str(t()));
+                        throw std::runtime_error(msg);
+                }
+#endif
+                return utility::lexical_cast<int64_t>(start_, end_ - start_);
+            }
+
+            /// The unsigned integer value.
+            uint64_t u() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                switch (t())
+                {
+                    case type::Number:
+                    case type::String:
+                        return utility::lexical_cast<uint64_t>(start_, end_ - start_);
+                    default:
+                        throw std::runtime_error(std::string("expected number, got: ") + get_type_str(t()));
+                }
+#endif
+                return utility::lexical_cast<uint64_t>(start_, end_ - start_);
+            }
+
+            /// The double precision floating-point number value.
+            double d() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Number)
+                    throw std::runtime_error("value is not number");
+#endif
+                return utility::lexical_cast<double>(start_, end_ - start_);
+            }
+
+            /// The boolean value.
+            bool b() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::True && t() != type::False)
+                    throw std::runtime_error("value is not boolean");
+#endif
+                return t() == type::True;
+            }
+
+            /// The string value.
+            detail::r_string s() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::String)
+                    throw std::runtime_error("value is not string");
+#endif
+                unescape();
+                return detail::r_string{start_, end_};
+            }
+
+            /// The list or object value
+            std::vector<rvalue> lo()
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object && t() != type::List)
+                    throw std::runtime_error("value is not a container");
+#endif
+                std::vector<rvalue> ret;
+                ret.reserve(lsize_);
+                for (uint32_t i = 0; i < lsize_; i++)
+                {
+                    ret.emplace_back(l_[i]);
+                }
+                return ret;
+            }
+
+            /// Convert escaped string character to their original form ("\\n" -> '\n').
+            void unescape() const
+            {
+                if (*(start_ - 1))
+                {
+                    char* head = start_;
+                    char* tail = start_;
+                    while (head != end_)
+                    {
+                        if (*head == '\\')
+                        {
+                            switch (*++head)
+                            {
+                                case '"': *tail++ = '"'; break;
+                                case '\\': *tail++ = '\\'; break;
+                                case '/': *tail++ = '/'; break;
+                                case 'b': *tail++ = '\b'; break;
+                                case 'f': *tail++ = '\f'; break;
+                                case 'n': *tail++ = '\n'; break;
+                                case 'r': *tail++ = '\r'; break;
+                                case 't': *tail++ = '\t'; break;
+                                case 'u':
+                                {
+                                    auto from_hex = [](char c) {
+                                        if (c >= 'a')
+                                            return c - 'a' + 10;
+                                        if (c >= 'A')
+                                            return c - 'A' + 10;
+                                        return c - '0';
+                                    };
+                                    unsigned int code =
+                                      (from_hex(head[1]) << 12) +
+                                      (from_hex(head[2]) << 8) +
+                                      (from_hex(head[3]) << 4) +
+                                      from_hex(head[4]);
+                                    if (code >= 0x800)
+                                    {
+                                        *tail++ = 0xE0 | (code >> 12);
+                                        *tail++ = 0x80 | ((code >> 6) & 0x3F);
+                                        *tail++ = 0x80 | (code & 0x3F);
+                                    }
+                                    else if (code >= 0x80)
+                                    {
+                                        *tail++ = 0xC0 | (code >> 6);
+                                        *tail++ = 0x80 | (code & 0x3F);
+                                    }
+                                    else
+                                    {
+                                        *tail++ = code;
+                                    }
+                                    head += 4;
+                                }
+                                break;
+                            }
+                        }
+                        else
+                            *tail++ = *head;
+                        head++;
+                    }
+                    end_ = tail;
+                    *end_ = 0;
+                    *(start_ - 1) = 0;
+                }
+            }
+
+            /// Check if the json object has the passed string as a key.
+            bool has(const char* str) const
+            {
+                return has(std::string(str));
+            }
+
+            bool has(const std::string& str) const
+            {
+                struct Pred
+                {
+                    bool operator()(const rvalue& l, const rvalue& r) const
+                    {
+                        return l.key_ < r.key_;
+                    };
+                    bool operator()(const rvalue& l, const std::string& r) const
+                    {
+                        return l.key_ < r;
+                    };
+                    bool operator()(const std::string& l, const rvalue& r) const
+                    {
+                        return l < r.key_;
+                    };
+                };
+                if (!is_cached())
+                {
+                    std::sort(begin(), end(), Pred());
+                    set_cached();
+                }
+                auto it = lower_bound(begin(), end(), str, Pred());
+                return it != end() && it->key_ == str;
+            }
+
+            int count(const std::string& str)
+            {
+                return has(str) ? 1 : 0;
+            }
+
+            rvalue* begin() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object && t() != type::List)
+                    throw std::runtime_error("value is not a container");
+#endif
+                return l_.get();
+            }
+            rvalue* end() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object && t() != type::List)
+                    throw std::runtime_error("value is not a container");
+#endif
+                return l_.get() + lsize_;
+            }
+
+            const detail::r_string& key() const
+            {
+                return key_;
+            }
+
+            size_t size() const
+            {
+                if (t() == type::String)
+                    return s().size();
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object && t() != type::List)
+                    throw std::runtime_error("value is not a container");
+#endif
+                return lsize_;
+            }
+
+            const rvalue& operator[](int index) const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::List)
+                    throw std::runtime_error("value is not a list");
+                if (index >= static_cast<int>(lsize_) || index < 0)
+                    throw std::runtime_error("list out of bound");
+#endif
+                return l_[index];
+            }
+
+            const rvalue& operator[](size_t index) const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::List)
+                    throw std::runtime_error("value is not a list");
+                if (index >= lsize_)
+                    throw std::runtime_error("list out of bound");
+#endif
+                return l_[index];
+            }
+
+            const rvalue& operator[](const char* str) const
+            {
+                return this->operator[](std::string(str));
+            }
+
+            const rvalue& operator[](const std::string& str) const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object)
+                    throw std::runtime_error("value is not an object");
+#endif
+                struct Pred
+                {
+                    bool operator()(const rvalue& l, const rvalue& r) const
+                    {
+                        return l.key_ < r.key_;
+                    };
+                    bool operator()(const rvalue& l, const std::string& r) const
+                    {
+                        return l.key_ < r;
+                    };
+                    bool operator()(const std::string& l, const rvalue& r) const
+                    {
+                        return l < r.key_;
+                    };
+                };
+                if (!is_cached())
+                {
+                    std::sort(begin(), end(), Pred());
+                    set_cached();
+                }
+                auto it = lower_bound(begin(), end(), str, Pred());
+                if (it != end() && it->key_ == str)
+                    return *it;
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                throw std::runtime_error("cannot find key");
+#else
+                static rvalue nullValue;
+                return nullValue;
+#endif
+            }
+
+            void set_error()
+            {
+                option_ |= error_bit;
+            }
+
+            bool error() const
+            {
+                return (option_ & error_bit) != 0;
+            }
+
+            std::vector<std::string> keys() const
+            {
+#ifndef CROW_JSON_NO_ERROR_CHECK
+                if (t() != type::Object)
+                    throw std::runtime_error("value is not an object");
+#endif
+                std::vector<std::string> ret;
+                ret.reserve(lsize_);
+                for (uint32_t i = 0; i < lsize_; i++)
+                {
+                    ret.emplace_back(std::string(l_[i].key()));
+                }
+                return ret;
+            }
+
+        private:
+            bool is_cached() const
+            {
+                return (option_ & cached_bit) != 0;
+            }
+            void set_cached() const
+            {
+                option_ |= cached_bit;
+            }
+            void copy_l(const rvalue& r)
+            {
+                if (r.t() != type::Object && r.t() != type::List)
+                    return;
+                lsize_ = r.lsize_;
+                lremain_ = 0;
+                l_.reset(new rvalue[lsize_]);
+                std::copy(r.begin(), r.end(), begin());
+            }
+
+            void emplace_back(rvalue&& v)
+            {
+                if (!lremain_)
+                {
+                    int new_size = lsize_ + lsize_;
+                    if (new_size - lsize_ > 60000)
+                        new_size = lsize_ + 60000;
+                    if (new_size < 4)
+                        new_size = 4;
+                    rvalue* p = new rvalue[new_size];
+                    rvalue* p2 = p;
+                    for (auto& x : *this)
+                        *p2++ = std::move(x);
+                    l_.reset(p);
+                    lremain_ = new_size - lsize_;
+                }
+                l_[lsize_++] = std::move(v);
+                lremain_--;
+            }
+
+            /// Determines num_type from the string.
+            void determine_num_type()
+            {
+                if (t_ != type::Number)
+                {
+                    nt_ = num_type::Null;
+                    return;
+                }
+
+                const std::size_t len = end_ - start_;
+                const bool has_minus = std::memchr(start_, '-', len) != nullptr;
+                const bool has_e = std::memchr(start_, 'e', len) != nullptr || std::memchr(start_, 'E', len) != nullptr;
+                const bool has_dec_sep = std::memchr(start_, '.', len) != nullptr;
+                if (has_dec_sep || has_e)
+                    nt_ = num_type::Floating_point;
+                else if (has_minus)
+                    nt_ = num_type::Signed_integer;
+                else
+                    nt_ = num_type::Unsigned_integer;
+            }
+
+            mutable char* start_;
+            mutable char* end_;
+            detail::r_string key_;
+            std::unique_ptr<rvalue[]> l_;
+            uint32_t lsize_;
+            uint16_t lremain_;
+            type t_;
+            num_type nt_{num_type::Null};
+            mutable uint8_t option_{0};
+
+            friend rvalue load_nocopy_internal(char* data, size_t size);
+            friend rvalue load(const char* data, size_t size);
+            friend std::ostream& operator<<(std::ostream& os, const rvalue& r)
+            {
+                switch (r.t_)
+                {
+
+                    case type::Null: os << "null"; break;
+                    case type::False: os << "false"; break;
+                    case type::True: os << "true"; break;
+                    case type::Number:
+                    {
+                        switch (r.nt())
+                        {
+                            case num_type::Floating_point: os << r.d(); break;
+                            case num_type::Double_precision_floating_point: os << r.d(); break;
+                            case num_type::Signed_integer: os << r.i(); break;
+                            case num_type::Unsigned_integer: os << r.u(); break;
+                            case num_type::Null: throw std::runtime_error("Number with num_type Null");
+                        }
+                    }
+                    break;
+                    case type::String: os << '"' << r.s() << '"'; break;
+                    case type::List:
+                    {
+                        os << '[';
+                        bool first = true;
+                        for (auto& x : r)
+                        {
+                            if (!first)
+                                os << ',';
+                            first = false;
+                            os << x;
+                        }
+                        os << ']';
+                    }
+                    break;
+                    case type::Object:
+                    {
+                        os << '{';
+                        bool first = true;
+                        for (auto& x : r)
+                        {
+                            if (!first)
+                                os << ',';
+                            os << '"' << escape(x.key_) << "\":";
+                            first = false;
+                            os << x;
+                        }
+                        os << '}';
+                    }
+                    break;
+                    case type::Function: os << "custom function"; break;
+                }
+                return os;
+            }
+        };
+        namespace detail
+        {
+        }
+
+        inline bool operator==(const rvalue& l, const std::string& r)
+        {
+            return l.s() == r;
+        }
+
+        inline bool operator==(const std::string& l, const rvalue& r)
+        {
+            return l == r.s();
+        }
+
+        inline bool operator!=(const rvalue& l, const std::string& r)
+        {
+            return l.s() != r;
+        }
+
+        inline bool operator!=(const std::string& l, const rvalue& r)
+        {
+            return l != r.s();
+        }
+
+        inline bool operator==(const rvalue& l, double r)
+        {
+            return l.d() == r;
+        }
+
+        inline bool operator==(double l, const rvalue& r)
+        {
+            return l == r.d();
+        }
+
+        inline bool operator!=(const rvalue& l, double r)
+        {
+            return l.d() != r;
+        }
+
+        inline bool operator!=(double l, const rvalue& r)
+        {
+            return l != r.d();
+        }
+
+
+        inline rvalue load_nocopy_internal(char* data, size_t size)
+        {
+            // Defend against excessive recursion
+            static constexpr unsigned max_depth = 10000;
+
+            //static const char* escaped = "\"\\/\b\f\n\r\t";
+            struct Parser
+            {
+                Parser(char* data, size_t /*size*/):
+                  data(data)
+                {
+                }
+
+                bool consume(char c)
+                {
+                    if (CROW_UNLIKELY(*data != c))
+                        return false;
+                    data++;
+                    return true;
+                }
+
+                void ws_skip()
+                {
+                    while (*data == ' ' || *data == '\t' || *data == '\r' || *data == '\n')
+                        ++data;
+                };
+
+                rvalue decode_string()
+                {
+                    if (CROW_UNLIKELY(!consume('"')))
+                        return {};
+                    char* start = data;
+                    uint8_t has_escaping = 0;
+                    while (1)
+                    {
+                        if (CROW_LIKELY(*data != '"' && *data != '\\' && *data != '\0'))
+                        {
+                            data++;
+                        }
+                        else if (*data == '"')
+                        {
+                            *data = 0;
+                            *(start - 1) = has_escaping;
+                            data++;
+                            return {type::String, start, data - 1};
+                        }
+                        else if (*data == '\\')
+                        {
+                            has_escaping = 1;
+                            data++;
+                            switch (*data)
+                            {
+                                case 'u':
+                                {
+                                    auto check = [](char c) {
+                                        return ('0' <= c && c <= '9') ||
+                                               ('a' <= c && c <= 'f') ||
+                                               ('A' <= c && c <= 'F');
+                                    };
+                                    if (!(check(*(data + 1)) &&
+                                          check(*(data + 2)) &&
+                                          check(*(data + 3)) &&
+                                          check(*(data + 4))))
+                                        return {};
+                                }
+                                    data += 5;
+                                    break;
+                                case '"':
+                                case '\\':
+                                case '/':
+                                case 'b':
+                                case 'f':
+                                case 'n':
+                                case 'r':
+                                case 't':
+                                    data++;
+                                    break;
+                                default:
+                                    return {};
+                            }
+                        }
+                        else
+                            return {};
+                    }
+                    return {};
+                }
+
+                rvalue decode_list(unsigned depth)
+                {
+                    rvalue ret(type::List);
+                    if (CROW_UNLIKELY(!consume('[')) || CROW_UNLIKELY(depth > max_depth))
+                    {
+                        ret.set_error();
+                        return ret;
+                    }
+                    ws_skip();
+                    if (CROW_UNLIKELY(*data == ']'))
+                    {
+                        data++;
+                        return ret;
+                    }
+
+                    while (1)
+                    {
+                        auto v = decode_value(depth + 1);
+                        if (CROW_UNLIKELY(!v))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+                        ws_skip();
+                        ret.emplace_back(std::move(v));
+                        if (*data == ']')
+                        {
+                            data++;
+                            break;
+                        }
+                        if (CROW_UNLIKELY(!consume(',')))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+                        ws_skip();
+                    }
+                    return ret;
+                }
+
+                rvalue decode_number()
+                {
+                    char* start = data;
+
+                    enum NumberParsingState
+                    {
+                        Minus,
+                        AfterMinus,
+                        ZeroFirst,
+                        Digits,
+                        DigitsAfterPoints,
+                        E,
+                        DigitsAfterE,
+                        Invalid,
+                    } state{Minus};
+                    while (CROW_LIKELY(state != Invalid))
+                    {
+                        switch (*data)
+                        {
+                            case '0':
+                                state = static_cast<NumberParsingState>("\2\2\7\3\4\6\6"[state]);
+                                /*if (state == NumberParsingState::Minus || state == NumberParsingState::AfterMinus)
+                                {
+                                    state = NumberParsingState::ZeroFirst;
+                                }
+                                else if (state == NumberParsingState::Digits ||
+                                    state == NumberParsingState::DigitsAfterE ||
+                                    state == NumberParsingState::DigitsAfterPoints)
+                                {
+                                    // ok; pass
+                                }
+                                else if (state == NumberParsingState::E)
+                                {
+                                    state = NumberParsingState::DigitsAfterE;
+                                }
+                                else
+                                    return {};*/
+                                break;
+                            case '1':
+                            case '2':
+                            case '3':
+                            case '4':
+                            case '5':
+                            case '6':
+                            case '7':
+                            case '8':
+                            case '9':
+                                state = static_cast<NumberParsingState>("\3\3\7\3\4\6\6"[state]);
+                                while (*(data + 1) >= '0' && *(data + 1) <= '9')
+                                    data++;
+                                /*if (state == NumberParsingState::Minus || state == NumberParsingState::AfterMinus)
+                                {
+                                    state = NumberParsingState::Digits;
+                                }
+                                else if (state == NumberParsingState::Digits ||
+                                    state == NumberParsingState::DigitsAfterE ||
+                                    state == NumberParsingState::DigitsAfterPoints)
+                                {
+                                    // ok; pass
+                                }
+                                else if (state == NumberParsingState::E)
+                                {
+                                    state = NumberParsingState::DigitsAfterE;
+                                }
+                                else
+                                    return {};*/
+                                break;
+                            case '.':
+                                state = static_cast<NumberParsingState>("\7\7\4\4\7\7\7"[state]);
+                                /*
+                                if (state == NumberParsingState::Digits || state == NumberParsingState::ZeroFirst)
+                                {
+                                    state = NumberParsingState::DigitsAfterPoints;
+                                }
+                                else
+                                    return {};
+                                */
+                                break;
+                            case '-':
+                                state = static_cast<NumberParsingState>("\1\7\7\7\7\6\7"[state]);
+                                /*if (state == NumberParsingState::Minus)
+                                {
+                                    state = NumberParsingState::AfterMinus;
+                                }
+                                else if (state == NumberParsingState::E)
+                                {
+                                    state = NumberParsingState::DigitsAfterE;
+                                }
+                                else
+                                    return {};*/
+                                break;
+                            case '+':
+                                state = static_cast<NumberParsingState>("\7\7\7\7\7\6\7"[state]);
+                                /*if (state == NumberParsingState::E)
+                                {
+                                    state = NumberParsingState::DigitsAfterE;
+                                }
+                                else
+                                    return {};*/
+                                break;
+                            case 'e':
+                            case 'E':
+                                state = static_cast<NumberParsingState>("\7\7\7\5\5\7\7"[state]);
+                                /*if (state == NumberParsingState::Digits ||
+                                    state == NumberParsingState::DigitsAfterPoints)
+                                {
+                                    state = NumberParsingState::E;
+                                }
+                                else
+                                    return {};*/
+                                break;
+                            default:
+                                if (CROW_LIKELY(state == NumberParsingState::ZeroFirst ||
+                                                state == NumberParsingState::Digits ||
+                                                state == NumberParsingState::DigitsAfterPoints ||
+                                                state == NumberParsingState::DigitsAfterE))
+                                    return {type::Number, start, data};
+                                else
+                                    return {};
+                        }
+                        data++;
+                    }
+
+                    return {};
+                }
+
+
+                rvalue decode_value(unsigned depth)
+                {
+                    switch (*data)
+                    {
+                        case '[':
+                            return decode_list(depth + 1);
+                        case '{':
+                            return decode_object(depth + 1);
+                        case '"':
+                            return decode_string();
+                        case 't':
+                            if ( //e-data >= 4 &&
+                              data[1] == 'r' &&
+                              data[2] == 'u' &&
+                              data[3] == 'e')
+                            {
+                                data += 4;
+                                return {type::True};
+                            }
+                            else
+                                return {};
+                        case 'f':
+                            if ( //e-data >= 5 &&
+                              data[1] == 'a' &&
+                              data[2] == 'l' &&
+                              data[3] == 's' &&
+                              data[4] == 'e')
+                            {
+                                data += 5;
+                                return {type::False};
+                            }
+                            else
+                                return {};
+                        case 'n':
+                            if ( //e-data >= 4 &&
+                              data[1] == 'u' &&
+                              data[2] == 'l' &&
+                              data[3] == 'l')
+                            {
+                                data += 4;
+                                return {type::Null};
+                            }
+                            else
+                                return {};
+                        //case '1': case '2': case '3':
+                        //case '4': case '5': case '6':
+                        //case '7': case '8': case '9':
+                        //case '0': case '-':
+                        default:
+                            return decode_number();
+                    }
+                    return {};
+                }
+
+                rvalue decode_object(unsigned depth)
+                {
+                    rvalue ret(type::Object);
+                    if (CROW_UNLIKELY(!consume('{')) || CROW_UNLIKELY(depth > max_depth))
+                    {
+                        ret.set_error();
+                        return ret;
+                    }
+
+                    ws_skip();
+
+                    if (CROW_UNLIKELY(*data == '}'))
+                    {
+                        data++;
+                        return ret;
+                    }
+
+                    while (1)
+                    {
+                        auto t = decode_string();
+                        if (CROW_UNLIKELY(!t))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+
+                        ws_skip();
+                        if (CROW_UNLIKELY(!consume(':')))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+
+                        // TODO(ipkn) caching key to speed up (flyweight?)
+                        // I have no idea how flyweight could apply here, but maybe some speedup can happen if we stopped checking type since decode_string returns a string anyway
+                        auto key = t.s();
+
+                        ws_skip();
+                        auto v = decode_value(depth + 1);
+                        if (CROW_UNLIKELY(!v))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+                        ws_skip();
+
+                        v.key_ = std::move(key);
+                        ret.emplace_back(std::move(v));
+                        if (CROW_UNLIKELY(*data == '}'))
+                        {
+                            data++;
+                            break;
+                        }
+                        if (CROW_UNLIKELY(!consume(',')))
+                        {
+                            ret.set_error();
+                            break;
+                        }
+                        ws_skip();
+                    }
+                    return ret;
+                }
+
+                rvalue parse()
+                {
+                    ws_skip();
+                    auto ret = decode_value(0); // or decode object?
+                    ws_skip();
+                    if (ret && *data != '\0')
+                        ret.set_error();
+                    return ret;
+                }
+
+                char* data;
+            };
+            return Parser(data, size).parse();
+        }
+        inline rvalue load(const char* data, size_t size)
+        {
+            char* s = new char[size + 1];
+            memcpy(s, data, size);
+            s[size] = 0;
+            auto ret = load_nocopy_internal(s, size);
+            if (ret)
+                ret.key_.force(s, size);
+            else
+                delete[] s;
+            return ret;
+        }
+
+        inline rvalue load(const char* data)
+        {
+            return load(data, strlen(data));
+        }
+
+        inline rvalue load(const std::string& str)
+        {
+            return load(str.data(), str.size());
+        }
+
+        struct wvalue_reader;
+
+        /// JSON write value.
+
+        ///
+        /// Value can mean any json value, including a JSON object.<br>
+        /// Write means this class is used to primarily assemble JSON objects using keys and values and export those into a string.
+        class wvalue : public returnable
+        {
+            friend class crow::mustache::template_t;
+            friend struct wvalue_reader;
+
+        public:
+            using object =
+#ifdef CROW_JSON_USE_MAP
+              std::map<std::string, wvalue>;
+#else
+              std::unordered_map<std::string, wvalue>;
+#endif
+
+            using list = std::vector<wvalue>;
+
+            type t() const { return t_; }
+
+            /// Create an empty json value (outputs "{}" instead of a "null" string)
+            static crow::json::wvalue empty_object() { return crow::json::wvalue::object(); }
+
+        private:
+            type t_{type::Null};         ///< The type of the value.
+            num_type nt{num_type::Null}; ///< The specific type of the number if \ref t_ is a number.
+            union number
+            {
+                double d;
+                int64_t si;
+                uint64_t ui;
+
+            public:
+                constexpr number() noexcept:
+                  ui() {} /* default constructor initializes unsigned integer. */
+                constexpr number(std::uint64_t value) noexcept:
+                  ui(value) {}
+                constexpr number(std::int64_t value) noexcept:
+                  si(value) {}
+                explicit constexpr number(double value) noexcept:
+                  d(value) {}
+                explicit constexpr number(float value) noexcept:
+                  d(value) {}
+            } num;                                      ///< Value if type is a number.
+            std::string s;                              ///< Value if type is a string.
+            std::unique_ptr<list> l;                    ///< Value if type is a list.
+            std::unique_ptr<object> o;                  ///< Value if type is a JSON object.
+            std::function<std::string(std::string&)> f; ///< Value if type is a function (C++ lambda)
+
+        public:
+            wvalue():
+              returnable("application/json") {}
+
+            wvalue(std::nullptr_t):
+              returnable("application/json"), t_(type::Null) {}
+
+            wvalue(bool value):
+              returnable("application/json"), t_(value ? type::True : type::False) {}
+
+            wvalue(std::uint8_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Unsigned_integer), num(static_cast<std::uint64_t>(value)) {}
+            wvalue(std::uint16_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Unsigned_integer), num(static_cast<std::uint64_t>(value)) {}
+            wvalue(std::uint32_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Unsigned_integer), num(static_cast<std::uint64_t>(value)) {}
+            wvalue(std::uint64_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Unsigned_integer), num(static_cast<std::uint64_t>(value)) {}
+
+            wvalue(std::int8_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Signed_integer), num(static_cast<std::int64_t>(value)) {}
+            wvalue(std::int16_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Signed_integer), num(static_cast<std::int64_t>(value)) {}
+            wvalue(std::int32_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Signed_integer), num(static_cast<std::int64_t>(value)) {}
+            wvalue(std::int64_t value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Signed_integer), num(static_cast<std::int64_t>(value)) {}
+
+            wvalue(float value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Floating_point), num(static_cast<double>(value)) {}
+            wvalue(double value):
+              returnable("application/json"), t_(type::Number), nt(num_type::Double_precision_floating_point), num(static_cast<double>(value)) {}
+
+            wvalue(char const* value):
+              returnable("application/json"), t_(type::String), s(value) {}
+
+            wvalue(std::string const& value):
+              returnable("application/json"), t_(type::String), s(value) {}
+            wvalue(std::string&& value):
+              returnable("application/json"), t_(type::String), s(std::move(value)) {}
+
+            wvalue(std::initializer_list<std::pair<std::string const, wvalue>> initializer_list):
+              returnable("application/json"), t_(type::Object), o(new object(initializer_list)) {}
+
+            wvalue(object const& value):
+              returnable("application/json"), t_(type::Object), o(new object(value)) {}
+            wvalue(object&& value):
+              returnable("application/json"), t_(type::Object), o(new object(std::move(value))) {}
+
+            wvalue(const list& r):
+              returnable("application/json")
+            {
+                t_ = type::List;
+                l = std::unique_ptr<list>(new list{});
+                l->reserve(r.size());
+                for (auto it = r.begin(); it != r.end(); ++it)
+                    l->emplace_back(*it);
+            }
+            wvalue(list& r):
+              returnable("application/json")
+            {
+                t_ = type::List;
+                l = std::unique_ptr<list>(new list{});
+                l->reserve(r.size());
+                for (auto it = r.begin(); it != r.end(); ++it)
+                    l->emplace_back(*it);
+            }
+
+            /// Create a write value from a read value (useful for editing JSON strings).
+            wvalue(const rvalue& r):
+              returnable("application/json")
+            {
+                t_ = r.t();
+                switch (r.t())
+                {
+                    case type::Null:
+                    case type::False:
+                    case type::True:
+                    case type::Function:
+                        return;
+                    case type::Number:
+                        nt = r.nt();
+                        if (nt == num_type::Floating_point || nt == num_type::Double_precision_floating_point)
+                            num.d = r.d();
+                        else if (nt == num_type::Signed_integer)
+                            num.si = r.i();
+                        else
+                            num.ui = r.u();
+                        return;
+                    case type::String:
+                        s = r.s();
+                        return;
+                    case type::List:
+                        l = std::unique_ptr<list>(new list{});
+                        l->reserve(r.size());
+                        for (auto it = r.begin(); it != r.end(); ++it)
+                            l->emplace_back(*it);
+                        return;
+                    case type::Object:
+                        o = std::unique_ptr<object>(new object{});
+                        for (auto it = r.begin(); it != r.end(); ++it)
+                            o->emplace(it->key(), *it);
+                        return;
+                }
+            }
+
+            wvalue(const wvalue& r):
+              returnable("application/json")
+            {
+                t_ = r.t();
+                switch (r.t())
+                {
+                    case type::Null:
+                    case type::False:
+                    case type::True:
+                        return;
+                    case type::Number:
+                        nt = r.nt;
+                        if (nt == num_type::Floating_point || nt == num_type::Double_precision_floating_point)
+                            num.d = r.num.d;
+                        else if (nt == num_type::Signed_integer)
+                            num.si = r.num.si;
+                        else
+                            num.ui = r.num.ui;
+                        return;
+                    case type::String:
+                        s = r.s;
+                        return;
+                    case type::List:
+                        l = std::unique_ptr<list>(new list{});
+                        l->reserve(r.size());
+                        for (auto it = r.l->begin(); it != r.l->end(); ++it)
+                            l->emplace_back(*it);
+                        return;
+                    case type::Object:
+                        o = std::unique_ptr<object>(new object{});
+                        o->insert(r.o->begin(), r.o->end());
+                        return;
+                    case type::Function:
+                        f = r.f;
+                }
+            }
+
+            wvalue(wvalue&& r):
+              returnable("application/json")
+            {
+                *this = std::move(r);
+            }
+
+            wvalue& operator=(wvalue&& r)
+            {
+                t_ = r.t_;
+                nt = r.nt;
+                num = r.num;
+                s = std::move(r.s);
+                l = std::move(r.l);
+                o = std::move(r.o);
+                return *this;
+            }
+
+            /// Used for compatibility, same as \ref reset()
+            void clear()
+            {
+                reset();
+            }
+
+            void reset()
+            {
+                t_ = type::Null;
+                l.reset();
+                o.reset();
+            }
+
+            wvalue& operator=(std::nullptr_t)
+            {
+                reset();
+                return *this;
+            }
+            wvalue& operator=(bool value)
+            {
+                reset();
+                if (value)
+                    t_ = type::True;
+                else
+                    t_ = type::False;
+                return *this;
+            }
+
+            wvalue& operator=(float value)
+            {
+                reset();
+                t_ = type::Number;
+                num.d = value;
+                nt = num_type::Floating_point;
+                return *this;
+            }
+
+            wvalue& operator=(double value)
+            {
+                reset();
+                t_ = type::Number;
+                num.d = value;
+                nt = num_type::Double_precision_floating_point;
+                return *this;
+            }
+
+            wvalue& operator=(unsigned short value)
+            {
+                reset();
+                t_ = type::Number;
+                num.ui = value;
+                nt = num_type::Unsigned_integer;
+                return *this;
+            }
+
+            wvalue& operator=(short value)
+            {
+                reset();
+                t_ = type::Number;
+                num.si = value;
+                nt = num_type::Signed_integer;
+                return *this;
+            }
+
+            wvalue& operator=(long long value)
+            {
+                reset();
+                t_ = type::Number;
+                num.si = value;
+                nt = num_type::Signed_integer;
+                return *this;
+            }
+
+            wvalue& operator=(long value)
+            {
+                reset();
+                t_ = type::Number;
+                num.si = value;
+                nt = num_type::Signed_integer;
+                return *this;
+            }
+
+            wvalue& operator=(int value)
+            {
+                reset();
+                t_ = type::Number;
+                num.si = value;
+                nt = num_type::Signed_integer;
+                return *this;
+            }
+
+            wvalue& operator=(unsigned long long value)
+            {
+                reset();
+                t_ = type::Number;
+                num.ui = value;
+                nt = num_type::Unsigned_integer;
+                return *this;
+            }
+
+            wvalue& operator=(unsigned long value)
+            {
+                reset();
+                t_ = type::Number;
+                num.ui = value;
+                nt = num_type::Unsigned_integer;
+                return *this;
+            }
+
+            wvalue& operator=(unsigned int value)
+            {
+                reset();
+                t_ = type::Number;
+                num.ui = value;
+                nt = num_type::Unsigned_integer;
+                return *this;
+            }
+
+            wvalue& operator=(const char* str)
+            {
+                reset();
+                t_ = type::String;
+                s = str;
+                return *this;
+            }
+
+            wvalue& operator=(const std::string& str)
+            {
+                reset();
+                t_ = type::String;
+                s = str;
+                return *this;
+            }
+
+            wvalue& operator=(list&& v)
+            {
+                if (t_ != type::List)
+                    reset();
+                t_ = type::List;
+                if (!l)
+                    l = std::unique_ptr<list>(new list{});
+                l->clear();
+                l->resize(v.size());
+                size_t idx = 0;
+                for (auto& x : v)
+                {
+                    (*l)[idx++] = std::move(x);
+                }
+                return *this;
+            }
+
+            template<typename T>
+            wvalue& operator=(const std::vector<T>& v)
+            {
+                if (t_ != type::List)
+                    reset();
+                t_ = type::List;
+                if (!l)
+                    l = std::unique_ptr<list>(new list{});
+                l->clear();
+                l->resize(v.size());
+                size_t idx = 0;
+                for (auto& x : v)
+                {
+                    (*l)[idx++] = x;
+                }
+                return *this;
+            }
+
+            wvalue& operator=(std::initializer_list<std::pair<std::string const, wvalue>> initializer_list)
+            {
+                if (t_ != type::Object)
+                {
+                    reset();
+                    t_ = type::Object;
+                    o = std::unique_ptr<object>(new object(initializer_list));
+                }
+                else
+                {
+#if defined(__APPLE__) || defined(__MACH__) || defined(__FreeBSD__) || defined(__ANDROID__) || defined(_LIBCPP_VERSION)
+                    o = std::unique_ptr<object>(new object(initializer_list));
+#else
+                    (*o) = initializer_list;
+#endif
+                }
+                return *this;
+            }
+
+            wvalue& operator=(object const& value)
+            {
+                if (t_ != type::Object)
+                {
+                    reset();
+                    t_ = type::Object;
+                    o = std::unique_ptr<object>(new object(value));
+                }
+                else
+                {
+#if defined(__APPLE__) || defined(__MACH__) || defined(__FreeBSD__) || defined(__ANDROID__) || defined(_LIBCPP_VERSION)
+                    o = std::unique_ptr<object>(new object(value));
+#else
+                    (*o) = value;
+#endif
+                }
+                return *this;
+            }
+
+            wvalue& operator=(object&& value)
+            {
+                if (t_ != type::Object)
+                {
+                    reset();
+                    t_ = type::Object;
+                    o = std::unique_ptr<object>(new object(std::move(value)));
+                }
+                else
+                {
+                    (*o) = std::move(value);
+                }
+                return *this;
+            }
+
+            wvalue& operator=(std::function<std::string(std::string&)>&& func)
+            {
+                reset();
+                t_ = type::Function;
+                f = std::move(func);
+                return *this;
+            }
+
+            wvalue& operator[](unsigned index)
+            {
+                if (t_ != type::List)
+                    reset();
+                t_ = type::List;
+                if (!l)
+                    l = std::unique_ptr<list>(new list{});
+                if (l->size() < index + 1)
+                    l->resize(index + 1);
+                return (*l)[index];
+            }
+
+            const wvalue& operator[](unsigned index) const
+            {
+                return const_cast<wvalue*>(this)->operator[](index);
+            }
+
+            int count(const std::string& str) const
+            {
+                if (t_ != type::Object)
+                    return 0;
+                if (!o)
+                    return 0;
+                return o->count(str);
+            }
+
+            wvalue& operator[](const std::string& str)
+            {
+                if (t_ != type::Object)
+                    reset();
+                t_ = type::Object;
+                if (!o)
+                    o = std::unique_ptr<object>(new object{});
+                return (*o)[str];
+            }
+
+            const wvalue& operator[](const std::string& str) const
+            {
+                return const_cast<wvalue*>(this)->operator[](str);
+            }
+
+            std::vector<std::string> keys() const
+            {
+                if (t_ != type::Object)
+                    return {};
+                std::vector<std::string> result;
+                for (auto& kv : *o)
+                {
+                    result.push_back(kv.first);
+                }
+                return result;
+            }
+
+            std::string execute(std::string txt = "") const //Not using reference because it cannot be used with a default rvalue
+            {
+                if (t_ != type::Function)
+                    return "";
+                return f(txt);
+            }
+
+            /// If the wvalue is a list, it returns the length of the list, otherwise it returns 1.
+            std::size_t size() const
+            {
+                if (t_ != type::List)
+                    return 1;
+                return l->size();
+            }
+
+            /// Returns an estimated size of the value in bytes.
+            size_t estimate_length() const
+            {
+                switch (t_)
+                {
+                    case type::Null: return 4;
+                    case type::False: return 5;
+                    case type::True: return 4;
+                    case type::Number: return 30;
+                    case type::String: return 2 + s.size() + s.size() / 2;
+                    case type::List:
+                    {
+                        size_t sum{};
+                        if (l)
+                        {
+                            for (auto& x : *l)
+                            {
+                                sum += 1;
+                                sum += x.estimate_length();
+                            }
+                        }
+                        return sum + 2;
+                    }
+                    case type::Object:
+                    {
+                        size_t sum{};
+                        if (o)
+                        {
+                            for (auto& kv : *o)
+                            {
+                                sum += 2;
+                                sum += 2 + kv.first.size() + kv.first.size() / 2;
+                                sum += kv.second.estimate_length();
+                            }
+                        }
+                        return sum + 2;
+                    }
+                    case type::Function:
+                        return 0;
+                }
+                return 1;
+            }
+
+        private:
+            inline void dump_string(const std::string& str, std::string& out) const
+            {
+                out.push_back('"');
+                escape(str, out);
+                out.push_back('"');
+            }
+
+            inline void dump_indentation_part(std::string& out, const int indent, const char separator, const int indent_level) const
+            {
+                out.push_back('\n');
+                out.append(indent_level * indent, separator);
+            }
+
+
+            inline void dump_internal(const wvalue& v, std::string& out, const int indent, const char separator, const int indent_level = 0) const
+            {
+                switch (v.t_)
+                {
+                    case type::Null: out += "null"; break;
+                    case type::False: out += "false"; break;
+                    case type::True: out += "true"; break;
+                    case type::Number:
+                    {
+                        if (v.nt == num_type::Floating_point || v.nt == num_type::Double_precision_floating_point)
+                        {
+                            if (isnan(v.num.d) || isinf(v.num.d))
+                            {
+                                out += "null";
+                                CROW_LOG_WARNING << "Invalid JSON value detected (" << v.num.d << "), value set to null";
+                                break;
+                            }
+                            enum
+                            {
+                                start,
+                                decp, // Decimal point
+                                zero
+                            } f_state;
+                            char outbuf[128];
+                            if (v.nt == num_type::Double_precision_floating_point)
+                            {
+#ifdef _MSC_VER
+                                sprintf_s(outbuf, sizeof(outbuf), "%.*g", DECIMAL_DIG, v.num.d);
+#else
+                                snprintf(outbuf, sizeof(outbuf), "%.*g", DECIMAL_DIG, v.num.d);
+#endif
+                            }
+                            else
+                            {
+#ifdef _MSC_VER
+                                sprintf_s(outbuf, sizeof(outbuf), "%f", v.num.d);
+#else
+                                snprintf(outbuf, sizeof(outbuf), "%f", v.num.d);
+#endif 
+                            }
+                            char *p = &outbuf[0], *o = nullptr; // o is the position of the first trailing 0
+                            f_state = start;
+                            while (*p != '\0')
+                            {
+                                //std::cout << *p << std::endl;
+                                char ch = *p;
+                                switch (f_state)
+                                {
+                                    case start: // Loop and lookahead until a decimal point is found
+                                        if (ch == '.')
+                                        {
+                                            char fch = *(p + 1);
+                                            // if the first character is 0, leave it be (this is so that "1.00000" becomes "1.0" and not "1.")
+                                            if (fch != '\0' && fch == '0') p++;
+                                            f_state = decp;
+                                        }
+                                        p++;
+                                        break;
+                                    case decp: // Loop until a 0 is found, if found, record its position
+                                        if (ch == '0')
+                                        {
+                                            f_state = zero;
+                                            o = p;
+                                        }
+                                        p++;
+                                        break;
+                                    case zero: // if a non 0 is found (e.g. 1.00004) remove the earlier recorded 0 position and look for more trailing 0s
+                                        if (ch != '0')
+                                        {
+                                            o = nullptr;
+                                            f_state = decp;
+                                        }
+                                        p++;
+                                        break;
+                                }
+                            }
+                            if (o != nullptr) // if any trailing 0s are found, terminate the string where they begin
+                                *o = '\0';
+                            out += outbuf;
+                        }
+                        else if (v.nt == num_type::Signed_integer)
+                        {
+                            out += std::to_string(v.num.si);
+                        }
+                        else
+                        {
+                            out += std::to_string(v.num.ui);
+                        }
+                    }
+                    break;
+                    case type::String: dump_string(v.s, out); break;
+                    case type::List:
+                    {
+                        out.push_back('[');
+
+                        if (indent >= 0)
+                        {
+                            dump_indentation_part(out, indent, separator, indent_level + 1);
+                        }
+
+                        if (v.l)
+                        {
+                            bool first = true;
+                            for (auto& x : *v.l)
+                            {
+                                if (!first)
+                                {
+                                    out.push_back(',');
+
+                                    if (indent >= 0)
+                                    {
+                                        dump_indentation_part(out, indent, separator, indent_level + 1);
+                                    }
+                                }
+                                first = false;
+                                dump_internal(x, out, indent, separator, indent_level + 1);
+                            }
+                        }
+
+                        if (indent >= 0)
+                        {
+                            dump_indentation_part(out, indent, separator, indent_level);
+                        }
+
+                        out.push_back(']');
+                    }
+                    break;
+                    case type::Object:
+                    {
+                        out.push_back('{');
+
+                        if (indent >= 0)
+                        {
+                            dump_indentation_part(out, indent, separator, indent_level + 1);
+                        }
+
+                        if (v.o)
+                        {
+                            bool first = true;
+                            for (auto& kv : *v.o)
+                            {
+                                if (!first)
+                                {
+                                    out.push_back(',');
+                                    if (indent >= 0)
+                                    {
+                                        dump_indentation_part(out, indent, separator, indent_level + 1);
+                                    }
+                                }
+                                first = false;
+                                dump_string(kv.first, out);
+                                out.push_back(':');
+
+                                if (indent >= 0)
+                                {
+                                    out.push_back(' ');
+                                }
+
+                                dump_internal(kv.second, out, indent, separator, indent_level + 1);
+                            }
+                        }
+
+                        if (indent >= 0)
+                        {
+                            dump_indentation_part(out, indent, separator, indent_level);
+                        }
+
+                        out.push_back('}');
+                    }
+                    break;
+
+                    case type::Function:
+                        out += "custom function";
+                        break;
+                }
+            }
+
+        public:
+            std::string dump(const int indent, const char separator = ' ') const
+            {
+                std::string ret;
+                ret.reserve(estimate_length());
+                dump_internal(*this, ret, indent, separator);
+                return ret;
+            }
+
+            std::string dump() const
+            {
+                static constexpr int DontIndent = -1;
+
+                return dump(DontIndent);
+            }
+        };
+
+        // Used for accessing the internals of a wvalue
+        struct wvalue_reader
+        {
+            int64_t get(int64_t fallback)
+            {
+                if (ref.t() != type::Number || ref.nt == num_type::Floating_point || 
+                    ref.nt == num_type::Double_precision_floating_point)
+                    return fallback;
+                return ref.num.si;
+            }
+
+            double get(double fallback)
+            {
+                if (ref.t() != type::Number || ref.nt != num_type::Floating_point ||
+                    ref.nt == num_type::Double_precision_floating_point)
+                    return fallback;
+                return ref.num.d;
+            }
+
+            bool get(bool fallback)
+            {
+                if (ref.t() == type::True) return true;
+                if (ref.t() == type::False) return false;
+                return fallback;
+            }
+
+            std::string get(const std::string& fallback)
+            {
+                if (ref.t() != type::String) return fallback;
+                return ref.s;
+            }
+
+            const wvalue& ref;
+        };
+
+        //std::vector<asio::const_buffer> dump_ref(wvalue& v)
+        //{
+        //}
+    } // namespace json
+} // namespace crow
+
+#include <string>
+#include <unordered_map>
+#include <ios>
+#include <fstream>
+#include <sstream>
+// S_ISREG is not defined for windows
+// This defines it like suggested in https://stackoverflow.com/a/62371749
+#if defined(_MSC_VER)
+#define _CRT_INTERNAL_NONSTDC_NAMES 1
+#endif
+#include <sys/stat.h>
+#if !defined(S_ISREG) && defined(S_IFMT) && defined(S_IFREG)
+#define S_ISREG(m) (((m)&S_IFMT) == S_IFREG)
+#endif
+
+
+
+namespace crow
+{
+    template<typename Adaptor, typename Handler, typename... Middlewares>
+    class Connection;
+
+    class Router;
+
+    /// HTTP response
+    struct response
+    {
+        template<typename Adaptor, typename Handler, typename... Middlewares>
+        friend class crow::Connection;
+
+        friend class Router;
+
+        int code{200};    ///< The Status code for the response.
+        std::string body; ///< The actual payload containing the response data.
+        ci_map headers;   ///< HTTP headers.
+
+#ifdef CROW_ENABLE_COMPRESSION
+        bool compressed = true; ///< If compression is enabled and this is false, the individual response will not be compressed.
+#endif
+        bool skip_body = false;            ///< Whether this is a response to a HEAD request.
+        bool manual_length_header = false; ///< Whether Crow should automatically add a "Content-Length" header.
+
+        /// Set the value of an existing header in the response.
+        void set_header(std::string key, std::string value)
+        {
+            headers.erase(key);
+            headers.emplace(std::move(key), std::move(value));
+        }
+
+        /// Add a new header to the response.
+        void add_header(std::string key, std::string value)
+        {
+            headers.emplace(std::move(key), std::move(value));
+        }
+
+        const std::string& get_header_value(const std::string& key)
+        {
+            return crow::get_header_value(headers, key);
+        }
+
+        // naive validation of a mime-type string
+        static bool validate_mime_type(const std::string& candidate) noexcept
+        {
+            // Here we simply check that the candidate type starts with
+            // a valid parent type, and has at least one character afterwards.
+            std::array<std::string, 10> valid_parent_types = {
+              "application/", "audio/", "font/", "example/",
+              "image/", "message/", "model/", "multipart/",
+              "text/", "video/"};
+            for (const std::string& parent : valid_parent_types)
+            {
+                // ensure the candidate is *longer* than the parent,
+                // to avoid unnecessary string comparison and to
+                // reject zero-length subtypes.
+                if (candidate.size() <= parent.size())
+                {
+                    continue;
+                }
+                // strncmp is used rather than substr to avoid allocation,
+                // but a string_view approach would be better if Crow
+                // migrates to C++17.
+                if (strncmp(parent.c_str(), candidate.c_str(), parent.size()) == 0)
+                {
+                    return true;
+                }
+            }
+            return false;
+        }
+
+        // Find the mime type from the content type either by lookup,
+        // or by the content type itself, if it is a valid a mime type.
+        // Defaults to text/plain.
+        static std::string get_mime_type(const std::string& contentType)
+        {
+            const auto mimeTypeIterator = mime_types.find(contentType);
+            if (mimeTypeIterator != mime_types.end())
+            {
+                return mimeTypeIterator->second;
+            }
+            else if (validate_mime_type(contentType))
+            {
+                return contentType;
+            }
+            else
+            {
+                CROW_LOG_WARNING << "Unable to interpret mime type for content type '" << contentType << "'. Defaulting to text/plain.";
+                return "text/plain";
+            }
+        }
+
+
+        // clang-format off
+        response() {}
+        explicit response(int code) : code(code) {}
+        response(std::string body) : body(std::move(body)) {}
+        response(int code, std::string body) : code(code), body(std::move(body)) {}
+        // clang-format on
+        response(returnable&& value)
+        {
+            body = value.dump();
+            set_header("Content-Type", value.content_type);
+        }
+        response(returnable& value)
+        {
+            body = value.dump();
+            set_header("Content-Type", value.content_type);
+        }
+        response(int code, returnable& value):
+          code(code)
+        {
+            body = value.dump();
+            set_header("Content-Type", value.content_type);
+        }
+        response(int code, returnable&& value):
+          code(code), body(value.dump())
+        {
+            set_header("Content-Type", std::move(value.content_type));
+        }
+
+        response(response&& r)
+        {
+            *this = std::move(r);
+        }
+
+        response(std::string contentType, std::string body):
+          body(std::move(body))
+        {
+            set_header("Content-Type", get_mime_type(contentType));
+        }
+
+        response(int code, std::string contentType, std::string body):
+          code(code), body(std::move(body))
+        {
+            set_header("Content-Type", get_mime_type(contentType));
+        }
+
+        response& operator=(const response& r) = delete;
+
+        response& operator=(response&& r) noexcept
+        {
+            body = std::move(r.body);
+            code = r.code;
+            headers = std::move(r.headers);
+            completed_ = r.completed_;
+            file_info = std::move(r.file_info);
+            return *this;
+        }
+
+        /// Check if the response has completed (whether response.end() has been called)
+        bool is_completed() const noexcept
+        {
+            return completed_;
+        }
+
+        void clear()
+        {
+            body.clear();
+            code = 200;
+            headers.clear();
+            completed_ = false;
+            file_info = static_file_info{};
+        }
+
+        /// Return a "Temporary Redirect" response.
+
+        ///
+        /// Location can either be a route or a full URL.
+        void redirect(const std::string& location)
+        {
+            code = 307;
+            set_header("Location", location);
+        }
+
+        /// Return a "Permanent Redirect" response.
+
+        ///
+        /// Location can either be a route or a full URL.
+        void redirect_perm(const std::string& location)
+        {
+            code = 308;
+            set_header("Location", location);
+        }
+
+        /// Return a "Found (Moved Temporarily)" response.
+
+        ///
+        /// Location can either be a route or a full URL.
+        void moved(const std::string& location)
+        {
+            code = 302;
+            set_header("Location", location);
+        }
+
+        /// Return a "Moved Permanently" response.
+
+        ///
+        /// Location can either be a route or a full URL.
+        void moved_perm(const std::string& location)
+        {
+            code = 301;
+            set_header("Location", location);
+        }
+
+        void write(const std::string& body_part)
+        {
+            body += body_part;
+        }
+
+        /// Set the response completion flag and call the handler (to send the response).
+        void end()
+        {
+            if (!completed_)
+            {
+                completed_ = true;
+                if (skip_body)
+                {
+                    set_header("Content-Length", std::to_string(body.size()));
+                    body = "";
+                    manual_length_header = true;
+                }
+                if (complete_request_handler_)
+                {
+                    complete_request_handler_();
+                    manual_length_header = false;
+                    skip_body = false;
+                }
+            }
+        }
+
+        /// Same as end() except it adds a body part right before ending.
+        void end(const std::string& body_part)
+        {
+            body += body_part;
+            end();
+        }
+
+        /// Check if the connection is still alive (usually by checking the socket status).
+        bool is_alive()
+        {
+            return is_alive_helper_ && is_alive_helper_();
+        }
+
+        /// Check whether the response has a static file defined.
+        bool is_static_type()
+        {
+            return file_info.path.size();
+        }
+
+        /// This constains metadata (coming from the `stat` command) related to any static files associated with this response.
+
+        ///
+        /// Either a static file or a string body can be returned as 1 response.
+        struct static_file_info
+        {
+            std::string path = "";
+            struct stat statbuf;
+            int statResult;
+        };
+
+        /// Return a static file as the response body
+        void set_static_file_info(std::string path)
+        {
+            utility::sanitize_filename(path);
+            set_static_file_info_unsafe(path);
+        }
+
+        /// Return a static file as the response body without sanitizing the path (use set_static_file_info instead)
+        void set_static_file_info_unsafe(std::string path)
+        {
+            file_info.path = path;
+            file_info.statResult = stat(file_info.path.c_str(), &file_info.statbuf);
+#ifdef CROW_ENABLE_COMPRESSION
+            compressed = false;
+#endif
+            if (file_info.statResult == 0 && S_ISREG(file_info.statbuf.st_mode))
+            {
+                std::size_t last_dot = path.find_last_of(".");
+                std::string extension = path.substr(last_dot + 1);
+                code = 200;
+                this->add_header("Content-Length", std::to_string(file_info.statbuf.st_size));
+
+                if (!extension.empty())
+                {
+                    this->add_header("Content-Type", get_mime_type(extension));
+                }
+            }
+            else
+            {
+                code = 404;
+                file_info.path.clear();
+            }
+        }
+
+    private:
+        bool completed_{};
+        std::function<void()> complete_request_handler_;
+        std::function<bool()> is_alive_helper_;
+        static_file_info file_info;
+    };
+} // namespace crow
+
+
+namespace crow
+{
+
+    struct UTF8
+    {
+        struct context
+        {};
+
+        void before_handle(request& /*req*/, response& /*res*/, context& /*ctx*/)
+        {}
+
+        void after_handle(request& /*req*/, response& res, context& /*ctx*/)
+        {
+            if (get_header_value(res.headers, "Content-Type").empty())
+            {
+                res.set_header("Content-Type", "text/plain; charset=utf-8");
+            }
+        }
+    };
+
+} // namespace crow
+
+#include <iomanip>
+#include <memory>
+
+namespace crow
+{
+    // Any middleware requires following 3 members:
+
+    // struct context;
+    //      storing data for the middleware; can be read from another middleware or handlers
+
+    // before_handle
+    //      called before handling the request.
+    //      if res.end() is called, the operation is halted.
+    //      (still call after_handle of this middleware)
+    //      2 signatures:
+    //      void before_handle(request& req, response& res, context& ctx)
+    //          if you only need to access this middlewares context.
+    //      template <typename AllContext>
+    //      void before_handle(request& req, response& res, context& ctx, AllContext& all_ctx)
+    //          you can access another middlewares' context by calling `all_ctx.template get<MW>()'
+    //          ctx == all_ctx.template get<CurrentMiddleware>()
+
+    // after_handle
+    //      called after handling the request.
+    //      void after_handle(request& req, response& res, context& ctx)
+    //      template <typename AllContext>
+    //      void after_handle(request& req, response& res, context& ctx, AllContext& all_ctx)
+
+    struct CookieParser
+    {
+        // Cookie stores key, value and attributes
+        struct Cookie
+        {
+            enum class SameSitePolicy
+            {
+                Strict,
+                Lax,
+                None
+            };
+
+            template<typename U>
+            Cookie(const std::string& key, U&& value):
+              Cookie()
+            {
+                key_ = key;
+                value_ = std::forward<U>(value);
+            }
+
+            Cookie(const std::string& key):
+              Cookie(key, "") {}
+
+            // format cookie to HTTP header format
+            std::string dump() const
+            {
+                const static char* HTTP_DATE_FORMAT = "%a, %d %b %Y %H:%M:%S GMT";
+
+                std::stringstream ss;
+                ss << key_ << '=';
+                ss << (value_.empty() ? "\"\"" : value_);
+                dumpString(ss, !domain_.empty(), "Domain=", domain_);
+                dumpString(ss, !path_.empty(), "Path=", path_);
+                dumpString(ss, secure_, "Secure");
+                dumpString(ss, httponly_, "HttpOnly");
+                if (expires_at_)
+                {
+                    ss << DIVIDER << "Expires="
+                       << std::put_time(expires_at_.get(), HTTP_DATE_FORMAT);
+                }
+                if (max_age_)
+                {
+                    ss << DIVIDER << "Max-Age=" << *max_age_;
+                }
+                if (same_site_)
+                {
+                    ss << DIVIDER << "SameSite=";
+                    switch (*same_site_)
+                    {
+                        case SameSitePolicy::Strict:
+                            ss << "Strict";
+                            break;
+                        case SameSitePolicy::Lax:
+                            ss << "Lax";
+                            break;
+                        case SameSitePolicy::None:
+                            ss << "None";
+                            break;
+                    }
+                }
+                return ss.str();
+            }
+
+            const std::string& name()
+            {
+                return key_;
+            }
+
+            template<typename U>
+            Cookie& value(U&& value)
+            {
+                value_ = std::forward<U>(value);
+                return *this;
+            }
+
+            // Expires attribute
+            Cookie& expires(const std::tm& time)
+            {
+                expires_at_ = std::unique_ptr<std::tm>(new std::tm(time));
+                return *this;
+            }
+
+            // Max-Age attribute
+            Cookie& max_age(long long seconds)
+            {
+                max_age_ = std::unique_ptr<long long>(new long long(seconds));
+                return *this;
+            }
+
+            // Domain attribute
+            Cookie& domain(const std::string& name)
+            {
+                domain_ = name;
+                return *this;
+            }
+
+            // Path attribute
+            Cookie& path(const std::string& path)
+            {
+                path_ = path;
+                return *this;
+            }
+
+            // Secured attribute
+            Cookie& secure()
+            {
+                secure_ = true;
+                return *this;
+            }
+
+            // HttpOnly attribute
+            Cookie& httponly()
+            {
+                httponly_ = true;
+                return *this;
+            }
+
+            // SameSite attribute
+            Cookie& same_site(SameSitePolicy ssp)
+            {
+                same_site_ = std::unique_ptr<SameSitePolicy>(new SameSitePolicy(ssp));
+                return *this;
+            }
+
+            Cookie(const Cookie& c):
+              key_(c.key_),
+              value_(c.value_),
+              domain_(c.domain_),
+              path_(c.path_),
+              secure_(c.secure_),
+              httponly_(c.httponly_)
+            {
+                if (c.max_age_)
+                    max_age_ = std::unique_ptr<long long>(new long long(*c.max_age_));
+
+                if (c.expires_at_)
+                    expires_at_ = std::unique_ptr<std::tm>(new std::tm(*c.expires_at_));
+
+                if (c.same_site_)
+                    same_site_ = std::unique_ptr<SameSitePolicy>(new SameSitePolicy(*c.same_site_));
+            }
+
+        private:
+            Cookie() = default;
+
+            static void dumpString(std::stringstream& ss, bool cond, const char* prefix,
+                                   const std::string& value = "")
+            {
+                if (cond)
+                {
+                    ss << DIVIDER << prefix << value;
+                }
+            }
+
+        private:
+            std::string key_;
+            std::string value_;
+            std::unique_ptr<long long> max_age_{};
+            std::string domain_ = "";
+            std::string path_ = "";
+            bool secure_ = false;
+            bool httponly_ = false;
+            std::unique_ptr<std::tm> expires_at_{};
+            std::unique_ptr<SameSitePolicy> same_site_{};
+
+            static constexpr const char* DIVIDER = "; ";
+        };
+
+
+        struct context
+        {
+            std::unordered_map<std::string, std::string> jar;
+
+            std::string get_cookie(const std::string& key) const
+            {
+                auto cookie = jar.find(key);
+                if (cookie != jar.end())
+                    return cookie->second;
+                return {};
+            }
+
+            template<typename U>
+            Cookie& set_cookie(const std::string& key, U&& value)
+            {
+                cookies_to_add.emplace_back(key, std::forward<U>(value));
+                return cookies_to_add.back();
+            }
+
+            Cookie& set_cookie(Cookie cookie)
+            {
+                cookies_to_add.push_back(std::move(cookie));
+                return cookies_to_add.back();
+            }
+
+        private:
+            friend struct CookieParser;
+            std::vector<Cookie> cookies_to_add;
+        };
+
+        void before_handle(request& req, response& res, context& ctx)
+        {
+            // TODO(dranikpg): remove copies, use string_view with c++17
+            int count = req.headers.count("Cookie");
+            if (!count)
+                return;
+            if (count > 1)
+            {
+                res.code = 400;
+                res.end();
+                return;
+            }
+            std::string cookies = req.get_header_value("Cookie");
+            size_t pos = 0;
+            while (pos < cookies.size())
+            {
+                size_t pos_equal = cookies.find('=', pos);
+                if (pos_equal == cookies.npos)
+                    break;
+                std::string name = cookies.substr(pos, pos_equal - pos);
+                name = utility::trim(name);
+                pos = pos_equal + 1;
+                if (pos == cookies.size())
+                    break;
+
+                size_t pos_semicolon = cookies.find(';', pos);
+                std::string value = cookies.substr(pos, pos_semicolon - pos);
+
+                value = utility::trim(value);
+                if (value[0] == '"' && value[value.size() - 1] == '"')
+                {
+                    value = value.substr(1, value.size() - 2);
+                }
+
+                ctx.jar.emplace(std::move(name), std::move(value));
+
+                pos = pos_semicolon;
+                if (pos == cookies.npos)
+                    break;
+                pos++;
+            }
+        }
+
+        void after_handle(request& /*req*/, response& res, context& ctx)
+        {
+            for (const auto& cookie : ctx.cookies_to_add)
+            {
+                res.add_header("Set-Cookie", cookie.dump());
+            }
+        }
+    };
+
+    /*
+    App<CookieParser, AnotherJarMW> app;
+    A B C
+    A::context
+        int aa;
+
+    ctx1 : public A::context
+    ctx2 : public ctx1, public B::context
+    ctx3 : public ctx2, public C::context
+
+    C depends on A
+
+    C::handle
+        context.aaa
+
+    App::context : private CookieParser::context, ...
+    {
+        jar
+
+    }
+
+    SimpleApp
+    */
+} // namespace crow
+
+
+
+#include <unordered_map>
+#include <unordered_set>
+#include <set>
+#include <queue>
+
+#include <memory>
+#include <string>
+#include <cstdio>
+#include <mutex>
+
+#include <fstream>
+#include <sstream>
+
+#include <type_traits>
+#include <functional>
+#include <chrono>
+
+#ifdef CROW_CAN_USE_CPP17
+#include <variant>
+#endif
+
+namespace
+{
+    // convert all integer values to int64_t
+    template<typename T>
+    using wrap_integral_t = typename std::conditional<
+      std::is_integral<T>::value && !std::is_same<bool, T>::value
+        // except for uint64_t because that could lead to overflow on conversion
+        && !std::is_same<uint64_t, T>::value,
+      int64_t, T>::type;
+
+    // convert char[]/char* to std::string
+    template<typename T>
+    using wrap_char_t = typename std::conditional<
+      std::is_same<typename std::decay<T>::type, char*>::value,
+      std::string, T>::type;
+
+    // Upgrade to correct type for multi_variant use
+    template<typename T>
+    using wrap_mv_t = wrap_char_t<wrap_integral_t<T>>;
+} // namespace
+
+namespace crow
+{
+    namespace session
+    {
+
+#ifdef CROW_CAN_USE_CPP17
+        using multi_value_types = black_magic::S<bool, int64_t, double, std::string>;
+
+        /// A multi_value is a safe variant wrapper with json conversion support
+        struct multi_value
+        {
+            json::wvalue json() const
+            {
+                // clang-format off
+                return std::visit([](auto arg) {
+                    return json::wvalue(arg);
+                }, v_);
+                // clang-format on
+            }
+
+            static multi_value from_json(const json::rvalue&);
+
+            std::string string() const
+            {
+                // clang-format off
+                return std::visit([](auto arg) {
+                    if constexpr (std::is_same_v<decltype(arg), std::string>)
+                        return arg;
+                    else
+                        return std::to_string(arg);
+                }, v_);
+                // clang-format on
+            }
+
+            template<typename T, typename RT = wrap_mv_t<T>>
+            RT get(const T& fallback)
+            {
+                if (const RT* val = std::get_if<RT>(&v_)) return *val;
+                return fallback;
+            }
+
+            template<typename T, typename RT = wrap_mv_t<T>>
+            void set(T val)
+            {
+                v_ = RT(std::move(val));
+            }
+
+            typename multi_value_types::rebind<std::variant> v_;
+        };
+
+        inline multi_value multi_value::from_json(const json::rvalue& rv)
+        {
+            using namespace json;
+            switch (rv.t())
+            {
+                case type::Number:
+                {
+                    if (rv.nt() == num_type::Floating_point || rv.nt() == num_type::Double_precision_floating_point)
+                        return multi_value{rv.d()};
+                    else if (rv.nt() == num_type::Unsigned_integer)
+                        return multi_value{int64_t(rv.u())};
+                    else
+                        return multi_value{rv.i()};
+                }
+                case type::False: return multi_value{false};
+                case type::True: return multi_value{true};
+                case type::String: return multi_value{std::string(rv)};
+                default: return multi_value{false};
+            }
+        }
+#else
+        // Fallback for C++11/14 that uses a raw json::wvalue internally.
+        // This implementation consumes significantly more memory
+        // than the variant-based version
+        struct multi_value
+        {
+            json::wvalue json() const { return v_; }
+
+            static multi_value from_json(const json::rvalue&);
+
+            std::string string() const { return v_.dump(); }
+
+            template<typename T, typename RT = wrap_mv_t<T>>
+            RT get(const T& fallback)
+            {
+                return json::wvalue_reader{v_}.get((const RT&)(fallback));
+            }
+
+            template<typename T, typename RT = wrap_mv_t<T>>
+            void set(T val)
+            {
+                v_ = RT(std::move(val));
+            }
+
+            json::wvalue v_;
+        };
+
+        inline multi_value multi_value::from_json(const json::rvalue& rv)
+        {
+            return {rv};
+        }
+#endif
+
+        /// Expiration tracker keeps track of soonest-to-expire keys
+        struct ExpirationTracker
+        {
+            using DataPair = std::pair<uint64_t /*time*/, std::string /*key*/>;
+
+            /// Add key with time to tracker.
+            /// If the key is already present, it will be updated
+            void add(std::string key, uint64_t time)
+            {
+                auto it = times_.find(key);
+                if (it != times_.end()) remove(key);
+                times_[key] = time;
+                queue_.insert({time, std::move(key)});
+            }
+
+            void remove(const std::string& key)
+            {
+                auto it = times_.find(key);
+                if (it != times_.end())
+                {
+                    queue_.erase({it->second, key});
+                    times_.erase(it);
+                }
+            }
+
+            /// Get expiration time of soonest-to-expire entry
+            uint64_t peek_first() const
+            {
+                if (queue_.empty()) return std::numeric_limits<uint64_t>::max();
+                return queue_.begin()->first;
+            }
+
+            std::string pop_first()
+            {
+                auto it = times_.find(queue_.begin()->second);
+                auto key = it->first;
+                times_.erase(it);
+                queue_.erase(queue_.begin());
+                return key;
+            }
+
+            using iterator = typename std::set<DataPair>::const_iterator;
+
+            iterator begin() const { return queue_.cbegin(); }
+
+            iterator end() const { return queue_.cend(); }
+
+        private:
+            std::set<DataPair> queue_;
+            std::unordered_map<std::string, uint64_t> times_;
+        };
+
+        /// CachedSessions are shared across requests
+        struct CachedSession
+        {
+            std::string session_id;
+            std::string requested_session_id; // session hasn't been created yet, but a key was requested
+
+            std::unordered_map<std::string, multi_value> entries;
+            std::unordered_set<std::string> dirty; // values that were changed after last load
+
+            void* store_data;
+            bool requested_refresh;
+
+            // number of references held - used for correctly destroying the cache.
+            // No need to be atomic, all SessionMiddleware accesses are synchronized
+            int referrers;
+            std::recursive_mutex mutex;
+        };
+    } // namespace session
+
+    // SessionMiddleware allows storing securely and easily small snippets of user information
+    template<typename Store>
+    struct SessionMiddleware
+    {
+#ifdef CROW_CAN_USE_CPP17
+        using lock = std::scoped_lock<std::mutex>;
+        using rc_lock = std::scoped_lock<std::recursive_mutex>;
+#else
+        using lock = std::lock_guard<std::mutex>;
+        using rc_lock = std::lock_guard<std::recursive_mutex>;
+#endif
+
+        struct context
+        {
+            // Get a mutex for locking this session
+            std::recursive_mutex& mutex()
+            {
+                check_node();
+                return node->mutex;
+            }
+
+            // Check whether this session is already present
+            bool exists() { return bool(node); }
+
+            // Get a value by key or fallback if it doesn't exist or is of another type
+            template<typename F>
+            auto get(const std::string& key, const F& fallback = F())
+              // This trick lets the multi_value deduce the return type from the fallback
+              // which allows both:
+              //   context.get<std::string>("key")
+              //   context.get("key", "") -> char[] is transformed into string by multivalue
+              // to return a string
+              -> decltype(std::declval<session::multi_value>().get<F>(std::declval<F>()))
+            {
+                if (!node) return fallback;
+                rc_lock l(node->mutex);
+
+                auto it = node->entries.find(key);
+                if (it != node->entries.end()) return it->second.get<F>(fallback);
+                return fallback;
+            }
+
+            // Set a value by key
+            template<typename T>
+            void set(const std::string& key, T value)
+            {
+                check_node();
+                rc_lock l(node->mutex);
+
+                node->dirty.insert(key);
+                node->entries[key].set(std::move(value));
+            }
+
+            bool contains(const std::string& key)
+            {
+                if (!node) return false;
+                return node->entries.find(key) != node->entries.end();
+            }
+
+            // Atomically mutate a value with a function
+            template<typename Func>
+            void apply(const std::string& key, const Func& f)
+            {
+                using traits = utility::function_traits<Func>;
+                using arg = typename std::decay<typename traits::template arg<0>>::type;
+                using retv = typename std::decay<typename traits::result_type>::type;
+                check_node();
+                rc_lock l(node->mutex);
+                node->dirty.insert(key);
+                node->entries[key].set<retv>(f(node->entries[key].get(arg{})));
+            }
+
+            // Remove a value from the session
+            void remove(const std::string& key)
+            {
+                if (!node) return;
+                rc_lock l(node->mutex);
+                node->dirty.insert(key);
+                node->entries.erase(key);
+            }
+
+            // Format value by key as a string
+            std::string string(const std::string& key)
+            {
+                if (!node) return "";
+                rc_lock l(node->mutex);
+
+                auto it = node->entries.find(key);
+                if (it != node->entries.end()) return it->second.string();
+                return "";
+            }
+
+            // Get a list of keys present in session
+            std::vector<std::string> keys()
+            {
+                if (!node) return {};
+                rc_lock l(node->mutex);
+
+                std::vector<std::string> out;
+                for (const auto& p : node->entries)
+                    out.push_back(p.first);
+                return out;
+            }
+
+            // Delay expiration by issuing another cookie with an updated expiration time
+            // and notifying the store
+            void refresh_expiration()
+            {
+                if (!node) return;
+                node->requested_refresh = true;
+            }
+
+        private:
+            friend struct SessionMiddleware;
+
+            void check_node()
+            {
+                if (!node) node = std::make_shared<session::CachedSession>();
+            }
+
+            std::shared_ptr<session::CachedSession> node;
+        };
+
+        template<typename... Ts>
+        SessionMiddleware(
+          CookieParser::Cookie cookie,
+          int id_length,
+          Ts... ts):
+          id_length_(id_length),
+          cookie_(cookie),
+          store_(std::forward<Ts>(ts)...), mutex_(new std::mutex{})
+        {}
+
+        template<typename... Ts>
+        SessionMiddleware(Ts... ts):
+          SessionMiddleware(
+            CookieParser::Cookie("session").path("/").max_age(/*month*/ 30 * 24 * 60 * 60),
+            /*id_length */ 20, // around 10^34 possible combinations, but small enough to fit into SSO
+            std::forward<Ts>(ts)...)
+        {}
+
+        template<typename AllContext>
+        void before_handle(request& /*req*/, response& /*res*/, context& ctx, AllContext& all_ctx)
+        {
+            lock l(*mutex_);
+
+            auto& cookies = all_ctx.template get<CookieParser>();
+            auto session_id = load_id(cookies);
+            if (session_id == "") return;
+
+            // search entry in cache
+            auto it = cache_.find(session_id);
+            if (it != cache_.end())
+            {
+                it->second->referrers++;
+                ctx.node = it->second;
+                return;
+            }
+
+            // check this is a valid entry before loading
+            if (!store_.contains(session_id)) return;
+
+            auto node = std::make_shared<session::CachedSession>();
+            node->session_id = session_id;
+            node->referrers = 1;
+
+            try
+            {
+                store_.load(*node);
+            }
+            catch (...)
+            {
+                CROW_LOG_ERROR << "Exception occurred during session load";
+                return;
+            }
+
+            ctx.node = node;
+            cache_[session_id] = node;
+        }
+
+        template<typename AllContext>
+        void after_handle(request& /*req*/, response& /*res*/, context& ctx, AllContext& all_ctx)
+        {
+            lock l(*mutex_);
+            if (!ctx.node || --ctx.node->referrers > 0) return;
+            ctx.node->requested_refresh |= ctx.node->session_id == "";
+
+            // generate new id
+            if (ctx.node->session_id == "")
+            {
+                // check for requested id
+                ctx.node->session_id = std::move(ctx.node->requested_session_id);
+                if (ctx.node->session_id == "")
+                {
+                    ctx.node->session_id = utility::random_alphanum(id_length_);
+                }
+            }
+            else
+            {
+                cache_.erase(ctx.node->session_id);
+            }
+
+            if (ctx.node->requested_refresh)
+            {
+                auto& cookies = all_ctx.template get<CookieParser>();
+                store_id(cookies, ctx.node->session_id);
+            }
+
+            try
+            {
+                store_.save(*ctx.node);
+            }
+            catch (...)
+            {
+                CROW_LOG_ERROR << "Exception occurred during session save";
+                return;
+            }
+        }
+
+    private:
+        std::string next_id()
+        {
+            std::string id;
+            do
+            {
+                id = utility::random_alphanum(id_length_);
+            } while (store_.contains(id));
+            return id;
+        }
+
+        std::string load_id(const CookieParser::context& cookies)
+        {
+            return cookies.get_cookie(cookie_.name());
+        }
+
+        void store_id(CookieParser::context& cookies, const std::string& session_id)
+        {
+            cookie_.value(session_id);
+            cookies.set_cookie(cookie_);
+        }
+
+    private:
+        int id_length_;
+
+        // prototype for cookie
+        CookieParser::Cookie cookie_;
+
+        Store store_;
+
+        // mutexes are immovable
+        std::unique_ptr<std::mutex> mutex_;
+        std::unordered_map<std::string, std::shared_ptr<session::CachedSession>> cache_;
+    };
+
+    /// InMemoryStore stores all entries in memory
+    struct InMemoryStore
+    {
+        // Load a value into the session cache.
+        // A load is always followed by a save, no loads happen consecutively
+        void load(session::CachedSession& cn)
+        {
+            // load & stores happen sequentially, so moving is safe
+            cn.entries = std::move(entries[cn.session_id]);
+        }
+
+        // Persist session data
+        void save(session::CachedSession& cn)
+        {
+            entries[cn.session_id] = std::move(cn.entries);
+            // cn.dirty is a list of changed keys since the last load
+        }
+
+        bool contains(const std::string& key)
+        {
+            return entries.count(key) > 0;
+        }
+
+        std::unordered_map<std::string, std::unordered_map<std::string, session::multi_value>> entries;
+    };
+
+    // FileStore stores all data as json files in a folder.
+    // Files are deleted after expiration. Expiration refreshes are automatically picked up.
+    struct FileStore
+    {
+        FileStore(const std::string& folder, uint64_t expiration_seconds = /*month*/ 30 * 24 * 60 * 60):
+          path_(folder), expiration_seconds_(expiration_seconds)
+        {
+            std::ifstream ifs(get_filename(".expirations", false));
+
+            auto current_ts = chrono_time();
+            std::string key;
+            uint64_t time;
+            while (ifs >> key >> time)
+            {
+                if (current_ts > time)
+                {
+                    evict(key);
+                }
+                else if (contains(key))
+                {
+                    expirations_.add(key, time);
+                }
+            }
+        }
+
+        ~FileStore()
+        {
+            std::ofstream ofs(get_filename(".expirations", false), std::ios::trunc);
+            for (const auto& p : expirations_)
+                ofs << p.second << " " << p.first << "\n";
+        }
+
+        // Delete expired entries
+        // At most 3 to prevent freezes
+        void handle_expired()
+        {
+            int deleted = 0;
+            auto current_ts = chrono_time();
+            while (current_ts > expirations_.peek_first() && deleted < 3)
+            {
+                evict(expirations_.pop_first());
+                deleted++;
+            }
+        }
+
+        void load(session::CachedSession& cn)
+        {
+            handle_expired();
+
+            std::ifstream file(get_filename(cn.session_id));
+
+            std::stringstream buffer;
+            buffer << file.rdbuf() << std::endl;
+
+            for (const auto& p : json::load(buffer.str()))
+                cn.entries[p.key()] = session::multi_value::from_json(p);
+        }
+
+        void save(session::CachedSession& cn)
+        {
+            if (cn.requested_refresh)
+                expirations_.add(cn.session_id, chrono_time() + expiration_seconds_);
+            if (cn.dirty.empty()) return;
+
+            std::ofstream file(get_filename(cn.session_id));
+            json::wvalue jw;
+            for (const auto& p : cn.entries)
+                jw[p.first] = p.second.json();
+            file << jw.dump() << std::flush;
+        }
+
+        std::string get_filename(const std::string& key, bool suffix = true)
+        {
+            return utility::join_path(path_, key + (suffix ? ".json" : ""));
+        }
+
+        bool contains(const std::string& key)
+        {
+            std::ifstream file(get_filename(key));
+            return file.good();
+        }
+
+        void evict(const std::string& key)
+        {
+            std::remove(get_filename(key).c_str());
+        }
+
+        uint64_t chrono_time() const
+        {
+            return std::chrono::duration_cast<std::chrono::seconds>(
+                     std::chrono::system_clock::now().time_since_epoch())
+              .count();
+        }
+
+        std::string path_;
+        uint64_t expiration_seconds_;
+        session::ExpirationTracker expirations_;
+    };
+
+} // namespace crow
+
+
+
+#include <tuple>
+#include <type_traits>
+#include <iostream>
+#include <utility>
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+
+    /// Local middleware should extend ILocalMiddleware
+    struct ILocalMiddleware
+    {
+        using call_global = std::false_type;
+    };
+
+    namespace detail
+    {
+        template<typename MW>
+        struct check_before_handle_arity_3_const
+        {
+            template<typename T, void (T::*)(request&, response&, typename MW::context&) const = &T::before_handle>
+            struct get
+            {};
+        };
+
+        template<typename MW>
+        struct check_before_handle_arity_3
+        {
+            template<typename T, void (T::*)(request&, response&, typename MW::context&) = &T::before_handle>
+            struct get
+            {};
+        };
+
+        template<typename MW>
+        struct check_after_handle_arity_3_const
+        {
+            template<typename T, void (T::*)(request&, response&, typename MW::context&) const = &T::after_handle>
+            struct get
+            {};
+        };
+
+        template<typename MW>
+        struct check_after_handle_arity_3
+        {
+            template<typename T, void (T::*)(request&, response&, typename MW::context&) = &T::after_handle>
+            struct get
+            {};
+        };
+
+        template<typename MW>
+        struct check_global_call_false
+        {
+            template<typename T, typename std::enable_if<T::call_global::value == false, bool>::type = true>
+            struct get
+            {};
+        };
+
+        template<typename T>
+        struct is_before_handle_arity_3_impl
+        {
+            template<typename C>
+            static std::true_type f(typename check_before_handle_arity_3_const<T>::template get<C>*);
+
+            template<typename C>
+            static std::true_type f(typename check_before_handle_arity_3<T>::template get<C>*);
+
+            template<typename C>
+            static std::false_type f(...);
+
+        public:
+            static const bool value = decltype(f<T>(nullptr))::value;
+        };
+
+        template<typename T>
+        struct is_after_handle_arity_3_impl
+        {
+            template<typename C>
+            static std::true_type f(typename check_after_handle_arity_3_const<T>::template get<C>*);
+
+            template<typename C>
+            static std::true_type f(typename check_after_handle_arity_3<T>::template get<C>*);
+
+            template<typename C>
+            static std::false_type f(...);
+
+        public:
+            static constexpr bool value = decltype(f<T>(nullptr))::value;
+        };
+
+        template<typename MW>
+        struct is_middleware_global
+        {
+            template<typename C>
+            static std::false_type f(typename check_global_call_false<MW>::template get<C>*);
+
+            template<typename C>
+            static std::true_type f(...);
+
+            static const bool value = decltype(f<MW>(nullptr))::value;
+        };
+
+        template<typename MW, typename Context, typename ParentContext>
+        typename std::enable_if<!is_before_handle_arity_3_impl<MW>::value>::type
+          before_handler_call(MW& mw, request& req, response& res, Context& ctx, ParentContext& /*parent_ctx*/)
+        {
+            mw.before_handle(req, res, ctx.template get<MW>(), ctx);
+        }
+
+        template<typename MW, typename Context, typename ParentContext>
+        typename std::enable_if<is_before_handle_arity_3_impl<MW>::value>::type
+          before_handler_call(MW& mw, request& req, response& res, Context& ctx, ParentContext& /*parent_ctx*/)
+        {
+            mw.before_handle(req, res, ctx.template get<MW>());
+        }
+
+        template<typename MW, typename Context, typename ParentContext>
+        typename std::enable_if<!is_after_handle_arity_3_impl<MW>::value>::type
+          after_handler_call(MW& mw, request& req, response& res, Context& ctx, ParentContext& /*parent_ctx*/)
+        {
+            mw.after_handle(req, res, ctx.template get<MW>(), ctx);
+        }
+
+        template<typename MW, typename Context, typename ParentContext>
+        typename std::enable_if<is_after_handle_arity_3_impl<MW>::value>::type
+          after_handler_call(MW& mw, request& req, response& res, Context& ctx, ParentContext& /*parent_ctx*/)
+        {
+            mw.after_handle(req, res, ctx.template get<MW>());
+        }
+
+
+        template<typename CallCriteria,
+                 int N, typename Context, typename Container>
+        typename std::enable_if<(N < std::tuple_size<typename std::remove_reference<Container>::type>::value), bool>::type
+          middleware_call_helper(const CallCriteria& cc, Container& middlewares, request& req, response& res, Context& ctx)
+        {
+
+            using CurrentMW = typename std::tuple_element<N, typename std::remove_reference<Container>::type>::type;
+
+            if (!cc.template enabled<CurrentMW>(N))
+            {
+                return middleware_call_helper<CallCriteria, N + 1, Context, Container>(cc, middlewares, req, res, ctx);
+            }
+
+            using parent_context_t = typename Context::template partial<N - 1>;
+            before_handler_call<CurrentMW, Context, parent_context_t>(std::get<N>(middlewares), req, res, ctx, static_cast<parent_context_t&>(ctx));
+            if (res.is_completed())
+            {
+                after_handler_call<CurrentMW, Context, parent_context_t>(std::get<N>(middlewares), req, res, ctx, static_cast<parent_context_t&>(ctx));
+                return true;
+            }
+
+            if (middleware_call_helper<CallCriteria, N + 1, Context, Container>(cc, middlewares, req, res, ctx))
+            {
+                after_handler_call<CurrentMW, Context, parent_context_t>(std::get<N>(middlewares), req, res, ctx, static_cast<parent_context_t&>(ctx));
+                return true;
+            }
+
+            return false;
+        }
+
+        template<typename CallCriteria, int N, typename Context, typename Container>
+        typename std::enable_if<(N >= std::tuple_size<typename std::remove_reference<Container>::type>::value), bool>::type
+          middleware_call_helper(const CallCriteria& /*cc*/, Container& /*middlewares*/, request& /*req*/, response& /*res*/, Context& /*ctx*/)
+        {
+            return false;
+        }
+
+        template<typename CallCriteria, int N, typename Context, typename Container>
+        typename std::enable_if<(N < 0)>::type
+          after_handlers_call_helper(const CallCriteria& /*cc*/, Container& /*middlewares*/, Context& /*context*/, request& /*req*/, response& /*res*/)
+        {
+        }
+
+        template<typename CallCriteria, int N, typename Context, typename Container>
+        typename std::enable_if<(N == 0)>::type after_handlers_call_helper(const CallCriteria& cc, Container& middlewares, Context& ctx, request& req, response& res)
+        {
+            using parent_context_t = typename Context::template partial<N - 1>;
+            using CurrentMW = typename std::tuple_element<N, typename std::remove_reference<Container>::type>::type;
+            if (cc.template enabled<CurrentMW>(N))
+            {
+                after_handler_call<CurrentMW, Context, parent_context_t>(std::get<N>(middlewares), req, res, ctx, static_cast<parent_context_t&>(ctx));
+            }
+        }
+
+        template<typename CallCriteria, int N, typename Context, typename Container>
+        typename std::enable_if<(N > 0)>::type after_handlers_call_helper(const CallCriteria& cc, Container& middlewares, Context& ctx, request& req, response& res)
+        {
+            using parent_context_t = typename Context::template partial<N - 1>;
+            using CurrentMW = typename std::tuple_element<N, typename std::remove_reference<Container>::type>::type;
+            if (cc.template enabled<CurrentMW>(N))
+            {
+                after_handler_call<CurrentMW, Context, parent_context_t>(std::get<N>(middlewares), req, res, ctx, static_cast<parent_context_t&>(ctx));
+            }
+            after_handlers_call_helper<CallCriteria, N - 1, Context, Container>(cc, middlewares, ctx, req, res);
+        }
+
+        // A CallCriteria that accepts only global middleware
+        struct middleware_call_criteria_only_global
+        {
+            template<typename MW>
+            constexpr bool enabled(int) const
+            {
+                return is_middleware_global<MW>::value;
+            }
+        };
+
+        template<typename F, typename... Args>
+        typename std::enable_if<black_magic::CallHelper<F, black_magic::S<Args...>>::value, void>::type
+          wrapped_handler_call(crow::request& /*req*/, crow::response& res, const F& f, Args&&... args)
+        {
+            static_assert(!std::is_same<void, decltype(f(std::declval<Args>()...))>::value,
+                          "Handler function cannot have void return type; valid return types: string, int, crow::response, crow::returnable");
+
+            res = crow::response(f(std::forward<Args>(args)...));
+            res.end();
+        }
+
+        template<typename F, typename... Args>
+        typename std::enable_if<
+          !black_magic::CallHelper<F, black_magic::S<Args...>>::value &&
+            black_magic::CallHelper<F, black_magic::S<crow::request&, Args...>>::value,
+          void>::type
+          wrapped_handler_call(crow::request& req, crow::response& res, const F& f, Args&&... args)
+        {
+            static_assert(!std::is_same<void, decltype(f(std::declval<crow::request>(), std::declval<Args>()...))>::value,
+                          "Handler function cannot have void return type; valid return types: string, int, crow::response, crow::returnable");
+
+            res = crow::response(f(req, std::forward<Args>(args)...));
+            res.end();
+        }
+
+        template<typename F, typename... Args>
+        typename std::enable_if<
+          !black_magic::CallHelper<F, black_magic::S<Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<crow::request&, Args...>>::value &&
+            black_magic::CallHelper<F, black_magic::S<crow::response&, Args...>>::value,
+          void>::type
+          wrapped_handler_call(crow::request& /*req*/, crow::response& res, const F& f, Args&&... args)
+        {
+            static_assert(std::is_same<void, decltype(f(std::declval<crow::response&>(), std::declval<Args>()...))>::value,
+                          "Handler function with response argument should have void return type");
+
+            f(res, std::forward<Args>(args)...);
+        }
+
+        template<typename F, typename... Args>
+        typename std::enable_if<
+          !black_magic::CallHelper<F, black_magic::S<Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<crow::request&, Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<crow::response&, Args...>>::value &&
+            black_magic::CallHelper<F, black_magic::S<const crow::request&, crow::response&, Args...>>::value,
+          void>::type
+          wrapped_handler_call(crow::request& req, crow::response& res, const F& f, Args&&... args)
+        {
+            static_assert(std::is_same<void, decltype(f(std::declval<crow::request&>(), std::declval<crow::response&>(), std::declval<Args>()...))>::value,
+                          "Handler function with response argument should have void return type");
+
+            f(req, res, std::forward<Args>(args)...);
+        }
+
+        // wrapped_handler_call transparently wraps a handler call behind (req, res, args...)
+        template<typename F, typename... Args>
+        typename std::enable_if<
+          !black_magic::CallHelper<F, black_magic::S<Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<crow::request&, Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<crow::response&, Args...>>::value &&
+            !black_magic::CallHelper<F, black_magic::S<const crow::request&, crow::response&, Args...>>::value,
+          void>::type
+          wrapped_handler_call(crow::request& req, crow::response& res, const F& f, Args&&... args)
+        {
+            static_assert(std::is_same<void, decltype(f(std::declval<crow::request&>(), std::declval<crow::response&>(), std::declval<Args>()...))>::value,
+                          "Handler function with response argument should have void return type");
+
+            f(req, res, std::forward<Args>(args)...);
+        }
+
+        template<bool Reversed>
+        struct middleware_call_criteria_dynamic
+        {};
+
+        template<>
+        struct middleware_call_criteria_dynamic<false>
+        {
+            middleware_call_criteria_dynamic(const std::vector<int>& indices):
+              indices(indices), slider(0) {}
+
+            template<typename>
+            bool enabled(int mw_index) const
+            {
+                if (slider < int(indices.size()) && indices[slider] == mw_index)
+                {
+                    slider++;
+                    return true;
+                }
+                return false;
+            }
+
+        private:
+            const std::vector<int>& indices;
+            mutable int slider;
+        };
+
+        template<>
+        struct middleware_call_criteria_dynamic<true>
+        {
+            middleware_call_criteria_dynamic(const std::vector<int>& indices):
+              indices(indices), slider(int(indices.size()) - 1) {}
+
+            template<typename>
+            bool enabled(int mw_index) const
+            {
+                if (slider >= 0 && indices[slider] == mw_index)
+                {
+                    slider--;
+                    return true;
+                }
+                return false;
+            }
+
+        private:
+            const std::vector<int>& indices;
+            mutable int slider;
+        };
+
+    } // namespace detail
+} // namespace crow
+
+
+
+namespace crow
+{
+    namespace detail
+    {
+
+
+        template<typename... Middlewares>
+        struct partial_context : public black_magic::pop_back<Middlewares...>::template rebind<partial_context>, public black_magic::last_element_type<Middlewares...>::type::context
+        {
+            using parent_context = typename black_magic::pop_back<Middlewares...>::template rebind<::crow::detail::partial_context>;
+            template<int N>
+            using partial = typename std::conditional<N == sizeof...(Middlewares) - 1, partial_context, typename parent_context::template partial<N>>::type;
+
+            template<typename T>
+            typename T::context& get()
+            {
+                return static_cast<typename T::context&>(*this);
+            }
+        };
+
+
+
+        template<>
+        struct partial_context<>
+        {
+            template<int>
+            using partial = partial_context;
+        };
+
+
+        template<typename... Middlewares>
+        struct context : private partial_context<Middlewares...>
+        //struct context : private Middlewares::context... // simple but less type-safe
+        {
+            template<typename CallCriteria, int N, typename Context, typename Container>
+            friend typename std::enable_if<(N == 0)>::type after_handlers_call_helper(const CallCriteria& cc, Container& middlewares, Context& ctx, request& req, response& res);
+            template<typename CallCriteria, int N, typename Context, typename Container>
+            friend typename std::enable_if<(N > 0)>::type after_handlers_call_helper(const CallCriteria& cc, Container& middlewares, Context& ctx, request& req, response& res);
+
+            template<typename CallCriteria, int N, typename Context, typename Container>
+            friend typename std::enable_if<(N < std::tuple_size<typename std::remove_reference<Container>::type>::value), bool>::type
+              middleware_call_helper(const CallCriteria& cc, Container& middlewares, request& req, response& res, Context& ctx);
+
+            template<typename T>
+            typename T::context& get()
+            {
+                return static_cast<typename T::context&>(*this);
+            }
+
+            template<int N>
+            using partial = typename partial_context<Middlewares...>::template partial<N>;
+        };
+    } // namespace detail
+} // namespace crow
+
+
+#ifdef CROW_USE_BOOST
+#include <boost/asio.hpp>
+#include <boost/asio/basic_waitable_timer.hpp>
+#else
+#ifndef ASIO_STANDALONE
+#define ASIO_STANDALONE
+#endif
+#include <asio.hpp>
+#include <asio/basic_waitable_timer.hpp>
+#endif
+
+#include <chrono>
+#include <functional>
+#include <map>
+#include <vector>
+
+
+namespace crow
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+    using error_code = boost::system::error_code;
+#else
+    using error_code = asio::error_code;
+#endif
+    namespace detail
+    {
+
+        /// A class for scheduling functions to be called after a specific amount of ticks. A tick is equal to 1 second.
+        class task_timer
+        {
+        public:
+            using task_type = std::function<void()>;
+            using identifier_type = size_t;
+
+        private:
+            using clock_type = std::chrono::steady_clock;
+            using time_type = clock_type::time_point;
+
+        public:
+            task_timer(asio::io_service& io_service):
+              io_service_(io_service), timer_(io_service_)
+            {
+                timer_.expires_after(std::chrono::seconds(1));
+                timer_.async_wait(
+                  std::bind(&task_timer::tick_handler, this, std::placeholders::_1));
+            }
+
+            ~task_timer() { timer_.cancel(); }
+
+            void cancel(identifier_type id)
+            {
+                tasks_.erase(id);
+                CROW_LOG_DEBUG << "task_timer cancelled: " << this << ' ' << id;
+            }
+
+            /// Schedule the given task to be executed after the default amount of ticks.
+
+            ///
+            /// \return identifier_type Used to cancel the thread.
+            /// It is not bound to this task_timer instance and in some cases could lead to
+            /// undefined behavior if used with other task_timer objects or after the task
+            /// has been successfully executed.
+            identifier_type schedule(const task_type& task)
+            {
+                tasks_.insert(
+                  {++highest_id_,
+                   {clock_type::now() + std::chrono::seconds(get_default_timeout()),
+                    task}});
+                CROW_LOG_DEBUG << "task_timer scheduled: " << this << ' ' << highest_id_;
+                return highest_id_;
+            }
+
+            /// Schedule the given task to be executed after the given time.
+
+            ///
+            /// \param timeout The amount of ticks (seconds) to wait before execution.
+            ///
+            /// \return identifier_type Used to cancel the thread.
+            /// It is not bound to this task_timer instance and in some cases could lead to
+            /// undefined behavior if used with other task_timer objects or after the task
+            /// has been successfully executed.
+            identifier_type schedule(const task_type& task, std::uint8_t timeout)
+            {
+                tasks_.insert({++highest_id_,
+                               {clock_type::now() + std::chrono::seconds(timeout), task}});
+                CROW_LOG_DEBUG << "task_timer scheduled: " << this << ' ' << highest_id_;
+                return highest_id_;
+            }
+
+            /// Set the default timeout for this task_timer instance. (Default: 5)
+
+            ///
+            /// \param timeout The amount of ticks (seconds) to wait before execution.
+            void set_default_timeout(std::uint8_t timeout) { default_timeout_ = timeout; }
+
+            /// Get the default timeout. (Default: 5)
+            std::uint8_t get_default_timeout() const { return default_timeout_; }
+
+        private:
+            void process_tasks()
+            {
+                time_type current_time = clock_type::now();
+                std::vector<identifier_type> finished_tasks;
+
+                for (const auto& task : tasks_)
+                {
+                    if (task.second.first < current_time)
+                    {
+                        (task.second.second)();
+                        finished_tasks.push_back(task.first);
+                        CROW_LOG_DEBUG << "task_timer called: " << this << ' ' << task.first;
+                    }
+                }
+
+                for (const auto& task : finished_tasks)
+                    tasks_.erase(task);
+
+                // If no task is currently scheduled, reset the issued ids back to 0.
+                if (tasks_.empty()) highest_id_ = 0;
+            }
+
+            void tick_handler(const error_code& ec)
+            {
+                if (ec) return;
+
+                process_tasks();
+
+                timer_.expires_after(std::chrono::seconds(1));
+                timer_.async_wait(
+                  std::bind(&task_timer::tick_handler, this, std::placeholders::_1));
+            }
+
+        private:
+            std::uint8_t default_timeout_{5};
+            asio::io_service& io_service_;
+            asio::basic_waitable_timer<clock_type> timer_;
+            std::map<identifier_type, std::pair<time_type, task_type>> tasks_;
+
+            // A continuosly increasing number to be issued to threads to identify them.
+            // If no tasks are scheduled, it will be reset to 0.
+            identifier_type highest_id_{0};
+        };
+    } // namespace detail
+} // namespace crow
+
+
+#ifdef CROW_USE_BOOST
+#include <boost/asio.hpp>
+#else
+#ifndef ASIO_STANDALONE
+#define ASIO_STANDALONE
+#endif
+#include <asio.hpp>
+#endif
+
+#include <algorithm>
+#include <atomic>
+#include <chrono>
+#include <memory>
+#include <vector>
+
+
+namespace crow
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+    using error_code = boost::system::error_code;
+#else
+    using error_code = asio::error_code;
+#endif
+    using tcp = asio::ip::tcp;
+
+#ifdef CROW_ENABLE_DEBUG
+    static std::atomic<int> connectionCount;
+#endif
+
+    /// An HTTP connection.
+    template<typename Adaptor, typename Handler, typename... Middlewares>
+    class Connection: public std::enable_shared_from_this<Connection<Adaptor, Handler, Middlewares...>>
+    {
+        friend struct crow::response;
+
+    public:
+        Connection(
+          asio::io_service& io_service,
+          Handler* handler,
+          const std::string& server_name,
+          std::tuple<Middlewares...>* middlewares,
+          std::function<std::string()>& get_cached_date_str_f,
+          detail::task_timer& task_timer,
+          typename Adaptor::context* adaptor_ctx_,
+          std::atomic<unsigned int>& queue_length):
+          adaptor_(io_service, adaptor_ctx_),
+          handler_(handler),
+          parser_(this),
+          req_(parser_.req),
+          server_name_(server_name),
+          middlewares_(middlewares),
+          get_cached_date_str(get_cached_date_str_f),
+          task_timer_(task_timer),
+          res_stream_threshold_(handler->stream_threshold()),
+          queue_length_(queue_length)
+        {
+#ifdef CROW_ENABLE_DEBUG
+            connectionCount++;
+            CROW_LOG_DEBUG << "Connection (" << this << ") allocated, total: " << connectionCount;
+#endif
+        }
+
+        ~Connection()
+        {
+#ifdef CROW_ENABLE_DEBUG
+            connectionCount--;
+            CROW_LOG_DEBUG << "Connection (" << this << ") freed, total: " << connectionCount;
+#endif
+        }
+
+        /// The TCP socket on top of which the connection is established.
+        decltype(std::declval<Adaptor>().raw_socket())& socket()
+        {
+            return adaptor_.raw_socket();
+        }
+
+        void start()
+        {
+            auto self = this->shared_from_this();
+            adaptor_.start([self](const error_code& ec) {
+                if (!ec)
+                {
+                    self->start_deadline();
+                    self->parser_.clear();
+
+                    self->do_read();
+                }
+                else
+                {
+                    CROW_LOG_ERROR << "Could not start adaptor: " << ec.message();
+                }
+            });
+        }
+
+        void handle_url()
+        {
+            routing_handle_result_ = handler_->handle_initial(req_, res);
+            // if no route is found for the request method, return the response without parsing or processing anything further.
+            if (!routing_handle_result_->rule_index)
+            {
+                parser_.done();
+                need_to_call_after_handlers_ = true;
+                complete_request();
+            }
+        }
+
+        void handle_header()
+        {
+            // HTTP 1.1 Expect: 100-continue
+            if (req_.http_ver_major == 1 && req_.http_ver_minor == 1 && get_header_value(req_.headers, "expect") == "100-continue")
+            {
+                continue_requested = true;
+                buffers_.clear();
+                static std::string expect_100_continue = "HTTP/1.1 100 Continue\r\n\r\n";
+                buffers_.emplace_back(expect_100_continue.data(), expect_100_continue.size());
+                do_write();
+            }
+        }
+
+        void handle()
+        {
+            // TODO(EDev): cancel_deadline_timer should be looked into, it might be a good idea to add it to handle_url() and then restart the timer once everything passes
+            cancel_deadline_timer();
+            bool is_invalid_request = false;
+            add_keep_alive_ = false;
+
+            // Create context
+            ctx_ = detail::context<Middlewares...>();
+            req_.middleware_context = static_cast<void*>(&ctx_);
+            req_.middleware_container = static_cast<void*>(middlewares_);
+            req_.io_service = &adaptor_.get_io_service();
+            
+            req_.remote_ip_address = adaptor_.remote_endpoint().address().to_string();
+
+            add_keep_alive_ = req_.keep_alive;
+            close_connection_ = req_.close_connection;
+
+            if (req_.check_version(1, 1)) // HTTP/1.1
+            {
+                if (!req_.headers.count("host"))
+                {
+                    is_invalid_request = true;
+                    res = response(400);
+                }
+                else if (req_.upgrade)
+                {
+                    // h2 or h2c headers
+                    if (req_.get_header_value("upgrade").substr(0, 2) == "h2")
+                    {
+                        // TODO(ipkn): HTTP/2
+                        // currently, ignore upgrade header
+                    }
+                    else
+                    {
+                
+                        detail::middleware_call_helper<detail::middleware_call_criteria_only_global,
+                                                       0, decltype(ctx_), decltype(*middlewares_)>({}, *middlewares_, req_, res, ctx_);
+                        close_connection_ = true;
+                        handler_->handle_upgrade(req_, res, std::move(adaptor_));
+                        return;
+                    }
+                }
+            }
+
+            CROW_LOG_INFO << "Request: " << utility::lexical_cast<std::string>(adaptor_.remote_endpoint()) << " " << this << " HTTP/" << (char)(req_.http_ver_major + '0') << "." << (char)(req_.http_ver_minor + '0') << ' ' << method_name(req_.method) << " " << req_.url;
+
+
+            need_to_call_after_handlers_ = false;
+            if (!is_invalid_request)
+            {
+                res.complete_request_handler_ = nullptr;
+                auto self = this->shared_from_this();
+                res.is_alive_helper_ = [self]() -> bool {
+                    return self->adaptor_.is_open();
+                };
+                
+                detail::middleware_call_helper<detail::middleware_call_criteria_only_global,
+                                               0, decltype(ctx_), decltype(*middlewares_)>({}, *middlewares_, req_, res, ctx_);
+
+                if (!res.completed_)
+                {
+                    auto self = this->shared_from_this();
+                    res.complete_request_handler_ = [self] {
+                        self->complete_request();
+                    };
+                    need_to_call_after_handlers_ = true;
+                    handler_->handle(req_, res, routing_handle_result_);
+                    if (add_keep_alive_)
+                        res.set_header("connection", "Keep-Alive");
+                }
+                else
+                {
+                    complete_request();
+                }
+            }
+            else
+            {
+                complete_request();
+            }
+        }
+
+        /// Call the after handle middleware and send the write the response to the connection.
+        void complete_request()
+        {
+            CROW_LOG_INFO << "Response: " << this << ' ' << req_.raw_url << ' ' << res.code << ' ' << close_connection_;
+            res.is_alive_helper_ = nullptr;
+
+            if (need_to_call_after_handlers_)
+            {
+                need_to_call_after_handlers_ = false;
+
+                // call all after_handler of middlewares
+                detail::after_handlers_call_helper<
+                  detail::middleware_call_criteria_only_global,
+                  (static_cast<int>(sizeof...(Middlewares)) - 1),
+                  decltype(ctx_),
+                  decltype(*middlewares_)>({}, *middlewares_, ctx_, req_, res);
+            }
+#ifdef CROW_ENABLE_COMPRESSION
+            if (handler_->compression_used())
+            {
+                std::string accept_encoding = req_.get_header_value("Accept-Encoding");
+                if (!accept_encoding.empty() && res.compressed)
+                {
+                    switch (handler_->compression_algorithm())
+                    {
+                        case compression::DEFLATE:
+                            if (accept_encoding.find("deflate") != std::string::npos)
+                            {
+                                res.body = compression::compress_string(res.body, compression::algorithm::DEFLATE);
+                                res.set_header("Content-Encoding", "deflate");
+                            }
+                            break;
+                        case compression::GZIP:
+                            if (accept_encoding.find("gzip") != std::string::npos)
+                            {
+                                res.body = compression::compress_string(res.body, compression::algorithm::GZIP);
+                                res.set_header("Content-Encoding", "gzip");
+                            }
+                            break;
+                        default:
+                            break;
+                    }
+                }
+            }
+#endif
+            //if there is a redirection with a partial URL, treat the URL as a route.
+            std::string location = res.get_header_value("Location");
+            if (!location.empty() && location.find("://", 0) == std::string::npos)
+            {
+#ifdef CROW_ENABLE_SSL
+                if (handler_->ssl_used())
+                    location.insert(0, "https://" + req_.get_header_value("Host"));
+                else
+#endif
+                    location.insert(0, "http://" + req_.get_header_value("Host"));
+                res.set_header("location", location);
+            }
+
+            prepare_buffers();
+
+            if (res.is_static_type())
+            {
+                do_write_static();
+            }
+            else
+            {
+                do_write_general();
+            }
+        }
+
+    private:
+        void prepare_buffers()
+        {
+            res.complete_request_handler_ = nullptr;
+            res.is_alive_helper_ = nullptr;
+
+            if (!adaptor_.is_open())
+            {
+                //CROW_LOG_DEBUG << this << " delete (socket is closed) " << is_reading << ' ' << is_writing;
+                //delete this;
+                return;
+            }
+            // TODO(EDev): HTTP version in status codes should be dynamic
+            // Keep in sync with common.h/status
+            static std::unordered_map<int, std::string> statusCodes = {
+              {status::CONTINUE, "HTTP/1.1 100 Continue\r\n"},
+              {status::SWITCHING_PROTOCOLS, "HTTP/1.1 101 Switching Protocols\r\n"},
+
+              {status::OK, "HTTP/1.1 200 OK\r\n"},
+              {status::CREATED, "HTTP/1.1 201 Created\r\n"},
+              {status::ACCEPTED, "HTTP/1.1 202 Accepted\r\n"},
+              {status::NON_AUTHORITATIVE_INFORMATION, "HTTP/1.1 203 Non-Authoritative Information\r\n"},
+              {status::NO_CONTENT, "HTTP/1.1 204 No Content\r\n"},
+              {status::RESET_CONTENT, "HTTP/1.1 205 Reset Content\r\n"},
+              {status::PARTIAL_CONTENT, "HTTP/1.1 206 Partial Content\r\n"},
+
+              {status::MULTIPLE_CHOICES, "HTTP/1.1 300 Multiple Choices\r\n"},
+              {status::MOVED_PERMANENTLY, "HTTP/1.1 301 Moved Permanently\r\n"},
+              {status::FOUND, "HTTP/1.1 302 Found\r\n"},
+              {status::SEE_OTHER, "HTTP/1.1 303 See Other\r\n"},
+              {status::NOT_MODIFIED, "HTTP/1.1 304 Not Modified\r\n"},
+              {status::TEMPORARY_REDIRECT, "HTTP/1.1 307 Temporary Redirect\r\n"},
+              {status::PERMANENT_REDIRECT, "HTTP/1.1 308 Permanent Redirect\r\n"},
+
+              {status::BAD_REQUEST, "HTTP/1.1 400 Bad Request\r\n"},
+              {status::UNAUTHORIZED, "HTTP/1.1 401 Unauthorized\r\n"},
+              {status::FORBIDDEN, "HTTP/1.1 403 Forbidden\r\n"},
+              {status::NOT_FOUND, "HTTP/1.1 404 Not Found\r\n"},
+              {status::METHOD_NOT_ALLOWED, "HTTP/1.1 405 Method Not Allowed\r\n"},
+              {status::NOT_ACCEPTABLE, "HTTP/1.1 406 Not Acceptable\r\n"},
+              {status::PROXY_AUTHENTICATION_REQUIRED, "HTTP/1.1 407 Proxy Authentication Required\r\n"},
+              {status::CONFLICT, "HTTP/1.1 409 Conflict\r\n"},
+              {status::GONE, "HTTP/1.1 410 Gone\r\n"},
+              {status::PAYLOAD_TOO_LARGE, "HTTP/1.1 413 Payload Too Large\r\n"},
+              {status::UNSUPPORTED_MEDIA_TYPE, "HTTP/1.1 415 Unsupported Media Type\r\n"},
+              {status::RANGE_NOT_SATISFIABLE, "HTTP/1.1 416 Range Not Satisfiable\r\n"},
+              {status::EXPECTATION_FAILED, "HTTP/1.1 417 Expectation Failed\r\n"},
+              {status::PRECONDITION_REQUIRED, "HTTP/1.1 428 Precondition Required\r\n"},
+              {status::TOO_MANY_REQUESTS, "HTTP/1.1 429 Too Many Requests\r\n"},
+              {status::UNAVAILABLE_FOR_LEGAL_REASONS, "HTTP/1.1 451 Unavailable For Legal Reasons\r\n"},
+
+              {status::INTERNAL_SERVER_ERROR, "HTTP/1.1 500 Internal Server Error\r\n"},
+              {status::NOT_IMPLEMENTED, "HTTP/1.1 501 Not Implemented\r\n"},
+              {status::BAD_GATEWAY, "HTTP/1.1 502 Bad Gateway\r\n"},
+              {status::SERVICE_UNAVAILABLE, "HTTP/1.1 503 Service Unavailable\r\n"},
+              {status::GATEWAY_TIMEOUT, "HTTP/1.1 504 Gateway Timeout\r\n"},
+              {status::VARIANT_ALSO_NEGOTIATES, "HTTP/1.1 506 Variant Also Negotiates\r\n"},
+            };
+
+            static const std::string seperator = ": ";
+
+            buffers_.clear();
+            buffers_.reserve(4 * (res.headers.size() + 5) + 3);
+
+            if (!statusCodes.count(res.code))
+            {
+                CROW_LOG_WARNING << this << " status code "
+                                 << "(" << res.code << ")"
+                                 << " not defined, returning 500 instead";
+                res.code = 500;
+            }
+
+            auto& status = statusCodes.find(res.code)->second;
+            buffers_.emplace_back(status.data(), status.size());
+
+            if (res.code >= 400 && res.body.empty())
+                res.body = statusCodes[res.code].substr(9);
+
+            for (auto& kv : res.headers)
+            {
+                buffers_.emplace_back(kv.first.data(), kv.first.size());
+                buffers_.emplace_back(seperator.data(), seperator.size());
+                buffers_.emplace_back(kv.second.data(), kv.second.size());
+                buffers_.emplace_back(crlf.data(), crlf.size());
+            }
+
+            if (!res.manual_length_header && !res.headers.count("content-length"))
+            {
+                content_length_ = std::to_string(res.body.size());
+                static std::string content_length_tag = "Content-Length: ";
+                buffers_.emplace_back(content_length_tag.data(), content_length_tag.size());
+                buffers_.emplace_back(content_length_.data(), content_length_.size());
+                buffers_.emplace_back(crlf.data(), crlf.size());
+            }
+            if (!res.headers.count("server"))
+            {
+                static std::string server_tag = "Server: ";
+                buffers_.emplace_back(server_tag.data(), server_tag.size());
+                buffers_.emplace_back(server_name_.data(), server_name_.size());
+                buffers_.emplace_back(crlf.data(), crlf.size());
+            }
+            if (!res.headers.count("date"))
+            {
+                static std::string date_tag = "Date: ";
+                date_str_ = get_cached_date_str();
+                buffers_.emplace_back(date_tag.data(), date_tag.size());
+                buffers_.emplace_back(date_str_.data(), date_str_.size());
+                buffers_.emplace_back(crlf.data(), crlf.size());
+            }
+            if (add_keep_alive_)
+            {
+                static std::string keep_alive_tag = "Connection: Keep-Alive";
+                buffers_.emplace_back(keep_alive_tag.data(), keep_alive_tag.size());
+                buffers_.emplace_back(crlf.data(), crlf.size());
+            }
+
+            buffers_.emplace_back(crlf.data(), crlf.size());
+        }
+
+        void do_write_static()
+        {
+            asio::write(adaptor_.socket(), buffers_);
+
+            if (res.file_info.statResult == 0)
+            {
+                std::ifstream is(res.file_info.path.c_str(), std::ios::in | std::ios::binary);
+                std::vector<asio::const_buffer> buffers{1};
+                char buf[16384];
+                is.read(buf, sizeof(buf));
+                while (is.gcount() > 0)
+                {
+                    buffers[0] = asio::buffer(buf, is.gcount());
+                    do_write_sync(buffers);
+                    is.read(buf, sizeof(buf));
+                }
+            }
+            if (close_connection_)
+            {
+                adaptor_.shutdown_readwrite();
+                adaptor_.close();
+                CROW_LOG_DEBUG << this << " from write (static)";
+            }
+
+            res.end();
+            res.clear();
+            buffers_.clear();
+            parser_.clear();
+        }
+
+        void do_write_general()
+        {
+            if (res.body.length() < res_stream_threshold_)
+            {
+                res_body_copy_.swap(res.body);
+                buffers_.emplace_back(res_body_copy_.data(), res_body_copy_.size());
+
+                do_write();
+
+                if (need_to_start_read_after_complete_)
+                {
+                    need_to_start_read_after_complete_ = false;
+                    start_deadline();
+                    do_read();
+                }
+            }
+            else
+            {
+                asio::write(adaptor_.socket(), buffers_); // Write the response start / headers
+                cancel_deadline_timer();
+                if (res.body.length() > 0)
+                {
+                    std::vector<asio::const_buffer> buffers{1};
+                    const uint8_t *data = reinterpret_cast<const uint8_t*>(res.body.data());
+                    size_t length = res.body.length();
+                    for(size_t transferred = 0; transferred < length;)
+                    {
+                        size_t to_transfer = CROW_MIN(16384UL, length-transferred);
+                        buffers[0] = asio::const_buffer(data+transferred, to_transfer);
+                        do_write_sync(buffers);
+                        transferred += to_transfer;
+                    }
+                }
+                if (close_connection_)
+                {
+                    adaptor_.shutdown_readwrite();
+                    adaptor_.close();
+                    CROW_LOG_DEBUG << this << " from write (res_stream)";
+                }
+
+                res.end();
+                res.clear();
+                buffers_.clear();
+                parser_.clear();
+            }
+        }
+
+        void do_read()
+        {
+            auto self = this->shared_from_this();
+            adaptor_.socket().async_read_some(
+              asio::buffer(buffer_),
+              [self](const error_code& ec, std::size_t bytes_transferred) {
+                  bool error_while_reading = true;
+                  if (!ec)
+                  {
+                      bool ret = self->parser_.feed(self->buffer_.data(), bytes_transferred);
+                      if (ret && self->adaptor_.is_open())
+                      {
+                          error_while_reading = false;
+                      }
+                  }
+
+                  if (error_while_reading)
+                  {
+                      self->cancel_deadline_timer();
+                      self->parser_.done();
+                      self->adaptor_.shutdown_read();
+                      self->adaptor_.close();
+                      CROW_LOG_DEBUG << self << " from read(1) with description: \"" << http_errno_description(static_cast<http_errno>(self->parser_.http_errno)) << '\"';
+                  }
+                  else if (self->close_connection_)
+                  {
+                      self->cancel_deadline_timer();
+                      self->parser_.done();
+                      // adaptor will close after write
+                  }
+                  else if (!self->need_to_call_after_handlers_)
+                  {
+                      self->start_deadline();
+                      self->do_read();
+                  }
+                  else
+                  {
+                      // res will be completed later by user
+                      self->need_to_start_read_after_complete_ = true;
+                  }
+              });
+        }
+
+        void do_write()
+        {
+            auto self = this->shared_from_this();
+            asio::async_write(
+              adaptor_.socket(), buffers_,
+              [self](const error_code& ec, std::size_t /*bytes_transferred*/) {
+                  self->res.clear();
+                  self->res_body_copy_.clear();                  
+                  if (!self->continue_requested)
+                  {
+                      self->parser_.clear();
+                  }
+                  else
+                  {
+                      self->continue_requested = false;
+                  }
+                  
+                  if (!ec)
+                  {
+                      if (self->close_connection_)
+                      {
+                          self->adaptor_.shutdown_write();
+                          self->adaptor_.close();
+                          CROW_LOG_DEBUG << self << " from write(1)";
+                      }
+                  }
+                  else
+                  {
+                      CROW_LOG_DEBUG << self << " from write(2)";
+                  }
+              });
+        }
+
+        inline void do_write_sync(std::vector<asio::const_buffer>& buffers)
+        {
+
+            asio::write(adaptor_.socket(), buffers, [&](error_code ec, std::size_t) {
+                if (!ec)
+                {
+                    return false;
+                }
+                else
+                {
+                    CROW_LOG_ERROR << ec << " - happened while sending buffers";
+                    CROW_LOG_DEBUG << this << " from write (sync)(2)";
+                    return true;
+                }
+            });
+        }
+
+        void cancel_deadline_timer()
+        {
+            CROW_LOG_DEBUG << this << " timer cancelled: " << &task_timer_ << ' ' << task_id_;
+            task_timer_.cancel(task_id_);
+        }
+
+        void start_deadline(/*int timeout = 5*/)
+        {
+            cancel_deadline_timer();
+
+            auto self = this->shared_from_this();
+            task_id_ = task_timer_.schedule([self] {
+                if (!self->adaptor_.is_open())
+                {
+                    return;
+                }
+                self->adaptor_.shutdown_readwrite();
+                self->adaptor_.close();
+            });
+            CROW_LOG_DEBUG << this << " timer added: " << &task_timer_ << ' ' << task_id_;
+        }
+
+    private:
+        Adaptor adaptor_;
+        Handler* handler_;
+
+        std::array<char, 4096> buffer_;
+
+        HTTPParser<Connection> parser_;
+        std::unique_ptr<routing_handle_result> routing_handle_result_;
+        request& req_;
+        response res;
+
+        bool close_connection_ = false;
+
+        const std::string& server_name_;
+        std::vector<asio::const_buffer> buffers_;
+
+        std::string content_length_;
+        std::string date_str_;
+        std::string res_body_copy_;
+
+        detail::task_timer::identifier_type task_id_{};
+
+        bool continue_requested{};
+        bool need_to_call_after_handlers_{};
+        bool need_to_start_read_after_complete_{};
+        bool add_keep_alive_{};
+
+        std::tuple<Middlewares...>* middlewares_;
+        detail::context<Middlewares...> ctx_;
+
+        std::function<std::string()>& get_cached_date_str;
+        detail::task_timer& task_timer_;
+
+        size_t res_stream_threshold_;
+
+        std::atomic<unsigned int>& queue_length_;
+    };
+
+} // namespace crow
+
+#include <array>
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+    using error_code = boost::system::error_code;
+#else
+    using error_code = asio::error_code;
+#endif
+
+    /**
+     * \namespace crow::websocket
+     * \brief Namespace that includes the \ref Connection class
+     * and \ref connection struct. Useful for WebSockets connection.
+     *
+     * Used specially in crow/websocket.h, crow/app.h and crow/routing.h
+     */
+    namespace websocket
+    {
+        enum class WebSocketReadState
+        {
+            MiniHeader,
+            Len16,
+            Len64,
+            Mask,
+            Payload,
+        };
+
+        /// A base class for websocket connection.
+        struct connection
+        {
+            virtual void send_binary(std::string msg) = 0;
+            virtual void send_text(std::string msg) = 0;
+            virtual void send_ping(std::string msg) = 0;
+            virtual void send_pong(std::string msg) = 0;
+            virtual void close(std::string const& msg = "quit") = 0;
+            virtual std::string get_remote_ip() = 0;
+            virtual ~connection() = default;
+
+            void userdata(void* u) { userdata_ = u; }
+            void* userdata() { return userdata_; }
+
+        private:
+            void* userdata_;
+        };
+
+        // Modified version of the illustration in RFC6455 Section-5.2
+        //
+        //
+        //  0               1               2               3               -byte
+        //  0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 -bit
+        // +-+-+-+-+-------+-+-------------+-------------------------------+
+        // |F|R|R|R| opcode|M| Payload len |    Extended payload length    |
+        // |I|S|S|S|  (4)  |A|     (7)     |             (16/64)           |
+        // |N|V|V|V|       |S|             |   (if payload len==126/127)   |
+        // | |1|2|3|       |K|             |                               |
+        // +-+-+-+-+-------+-+-------------+ - - - - - - - - - - - - - - - +
+        // |     Extended payload length continued, if payload len == 127  |
+        // + - - - - - - - - - - - - - - - +-------------------------------+
+        // |                               |Masking-key, if MASK set to 1  |
+        // +-------------------------------+-------------------------------+
+        // | Masking-key (continued)       |          Payload Data         |
+        // +-------------------------------- - - - - - - - - - - - - - - - +
+        // :                     Payload Data continued ...                :
+        // + - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - +
+        // |                     Payload Data continued ...                |
+        // +---------------------------------------------------------------+
+        //
+
+        /// A websocket connection.
+
+        template<typename Adaptor, typename Handler>
+        class Connection : public connection
+        {
+        public:
+            /// Constructor for a connection.
+
+            ///
+            /// Requires a request with an "Upgrade: websocket" header.<br>
+            /// Automatically handles the handshake.
+            Connection(const crow::request& req, Adaptor&& adaptor, Handler* handler, uint64_t max_payload,
+                       std::function<void(crow::websocket::connection&)> open_handler,
+                       std::function<void(crow::websocket::connection&, const std::string&, bool)> message_handler,
+                       std::function<void(crow::websocket::connection&, const std::string&)> close_handler,
+                       std::function<void(crow::websocket::connection&, const std::string&)> error_handler,
+                       std::function<bool(const crow::request&, void**)> accept_handler):
+              adaptor_(std::move(adaptor)),
+              handler_(handler),
+              max_payload_bytes_(max_payload),
+              open_handler_(std::move(open_handler)),
+              message_handler_(std::move(message_handler)),
+              close_handler_(std::move(close_handler)),
+              error_handler_(std::move(error_handler)),
+              accept_handler_(std::move(accept_handler))
+            {
+                if (!utility::string_equals(req.get_header_value("upgrade"), "websocket"))
+                {
+                    adaptor_.close();
+                    handler_->remove_websocket(this);
+                    delete this;
+                    return;
+                }
+
+                if (accept_handler_)
+                {
+                    void* ud = nullptr;
+                    if (!accept_handler_(req, &ud))
+                    {
+                        adaptor_.close();
+                        handler_->remove_websocket(this);
+                        delete this;
+                        return;
+                    }
+                    userdata(ud);
+                }
+
+                // Sec-WebSocket-Key: dGhlIHNhbXBsZSBub25jZQ==
+                // Sec-WebSocket-Version: 13
+                std::string magic = req.get_header_value("Sec-WebSocket-Key") + "258EAFA5-E914-47DA-95CA-C5AB0DC85B11";
+                sha1::SHA1 s;
+                s.processBytes(magic.data(), magic.size());
+                uint8_t digest[20];
+                s.getDigestBytes(digest);
+
+                start(crow::utility::base64encode((unsigned char*)digest, 20));
+            }
+
+            ~Connection() noexcept override
+            {
+                // Do not modify anchor_ here since writing shared_ptr is not atomic.
+                auto watch = std::weak_ptr<void>{anchor_};
+
+                // Wait until all unhandled asynchronous operations to join.
+                // As the deletion occurs inside 'check_destroy()', which already locks
+                //  anchor, use count can be 1 on valid deletion context.
+                while (watch.use_count() > 2) // 1 for 'check_destroy() routine', 1 for 'this->anchor_'
+                {
+                    std::this_thread::yield();
+                }
+            }
+
+            template<typename Callable>
+            struct WeakWrappedMessage
+            {
+                Callable callable;
+                std::weak_ptr<void> watch;
+
+                void operator()()
+                {
+                    if (auto anchor = watch.lock())
+                    {
+                        std::move(callable)();
+                    }
+                }
+            };
+
+            /// Send data through the socket.
+            template<typename CompletionHandler>
+            void dispatch(CompletionHandler&& handler)
+            {
+                asio::dispatch(adaptor_.get_io_service(),
+                               WeakWrappedMessage<typename std::decay<CompletionHandler>::type>{
+                                 std::forward<CompletionHandler>(handler), anchor_});
+            }
+
+            /// Send data through the socket and return immediately.
+            template<typename CompletionHandler>
+            void post(CompletionHandler&& handler)
+            {
+                asio::post(adaptor_.get_io_service(),
+                           WeakWrappedMessage<typename std::decay<CompletionHandler>::type>{
+                             std::forward<CompletionHandler>(handler), anchor_});
+            }
+
+            /// Send a "Ping" message.
+
+            ///
+            /// Usually invoked to check if the other point is still online.
+            void send_ping(std::string msg) override
+            {
+                send_data(0x9, std::move(msg));
+            }
+
+            /// Send a "Pong" message.
+
+            ///
+            /// Usually automatically invoked as a response to a "Ping" message.
+            void send_pong(std::string msg) override
+            {
+                send_data(0xA, std::move(msg));
+            }
+
+            /// Send a binary encoded message.
+            void send_binary(std::string msg) override
+            {
+                send_data(0x2, std::move(msg));
+            }
+
+            /// Send a plaintext message.
+            void send_text(std::string msg) override
+            {
+                send_data(0x1, std::move(msg));
+            }
+
+            /// Send a close signal.
+
+            ///
+            /// Sets a flag to destroy the object once the message is sent.
+            void close(std::string const& msg) override
+            {
+                dispatch([this, msg]() mutable {
+                    has_sent_close_ = true;
+                    if (has_recv_close_ && !is_close_handler_called_)
+                    {
+                        is_close_handler_called_ = true;
+                        if (close_handler_)
+                            close_handler_(*this, msg);
+                    }
+                    auto header = build_header(0x8, msg.size());
+                    write_buffers_.emplace_back(std::move(header));
+                    write_buffers_.emplace_back(msg);
+                    do_write();
+                });
+            }
+
+            std::string get_remote_ip() override
+            {
+                return adaptor_.remote_endpoint().address().to_string();
+            }
+
+            void set_max_payload_size(uint64_t payload)
+            {
+                max_payload_bytes_ = payload;
+            }
+
+        protected:
+            /// Generate the websocket headers using an opcode and the message size (in bytes).
+            std::string build_header(int opcode, size_t size)
+            {
+                char buf[2 + 8] = "\x80\x00";
+                buf[0] += opcode;
+                if (size < 126)
+                {
+                    buf[1] += static_cast<char>(size);
+                    return {buf, buf + 2};
+                }
+                else if (size < 0x10000)
+                {
+                    buf[1] += 126;
+                    *(uint16_t*)(buf + 2) = htons(static_cast<uint16_t>(size));
+                    return {buf, buf + 4};
+                }
+                else
+                {
+                    buf[1] += 127;
+                    *reinterpret_cast<uint64_t*>(buf + 2) = ((1 == htonl(1)) ? static_cast<uint64_t>(size) : (static_cast<uint64_t>(htonl((size)&0xFFFFFFFF)) << 32) | htonl(static_cast<uint64_t>(size) >> 32));
+                    return {buf, buf + 10};
+                }
+            }
+
+            /// Send the HTTP upgrade response.
+
+            ///
+            /// Finishes the handshake process, then starts reading messages from the socket.
+            void start(std::string&& hello)
+            {
+                static const std::string header =
+                  "HTTP/1.1 101 Switching Protocols\r\n"
+                  "Upgrade: websocket\r\n"
+                  "Connection: Upgrade\r\n"
+                  "Sec-WebSocket-Accept: ";
+                write_buffers_.emplace_back(header);
+                write_buffers_.emplace_back(std::move(hello));
+                write_buffers_.emplace_back(crlf);
+                write_buffers_.emplace_back(crlf);
+                do_write();
+                if (open_handler_)
+                    open_handler_(*this);
+                do_read();
+            }
+
+            /// Read a websocket message.
+
+            ///
+            /// Involves:<br>
+            /// Handling headers (opcodes, size).<br>
+            /// Unmasking the payload.<br>
+            /// Reading the actual payload.<br>
+            void do_read()
+            {
+                if (has_sent_close_ && has_recv_close_)
+                {
+                    close_connection_ = true;
+                    adaptor_.shutdown_readwrite();
+                    adaptor_.close();
+                    check_destroy();
+                    return;
+                }
+
+                is_reading = true;
+                switch (state_)
+                {
+                    case WebSocketReadState::MiniHeader:
+                    {
+                        mini_header_ = 0;
+                        //asio::async_read(adaptor_.socket(), asio::buffer(&mini_header_, 1),
+                        adaptor_.socket().async_read_some(
+                          asio::buffer(&mini_header_, 2),
+                          [this](const error_code& ec, std::size_t
+#ifdef CROW_ENABLE_DEBUG
+                                                               bytes_transferred
+#endif
+                          )
+
+                          {
+                              is_reading = false;
+                              mini_header_ = ntohs(mini_header_);
+#ifdef CROW_ENABLE_DEBUG
+
+                              if (!ec && bytes_transferred != 2)
+                              {
+                                  throw std::runtime_error("WebSocket:MiniHeader:async_read fail:asio bug?");
+                              }
+#endif
+
+                              if (!ec)
+                              {
+                                  if ((mini_header_ & 0x80) == 0x80)
+                                      has_mask_ = true;
+                                  else //if the websocket specification is enforced and the message isn't masked, terminate the connection
+                                  {
+#ifndef CROW_ENFORCE_WS_SPEC
+                                      has_mask_ = false;
+#else
+                                      close_connection_ = true;
+                                      adaptor_.shutdown_readwrite();
+                                      adaptor_.close();
+                                      if (error_handler_)
+                                          error_handler_(*this, "Client connection not masked.");
+                                      check_destroy();
+#endif
+                                  }
+
+                                  if ((mini_header_ & 0x7f) == 127)
+                                  {
+                                      state_ = WebSocketReadState::Len64;
+                                  }
+                                  else if ((mini_header_ & 0x7f) == 126)
+                                  {
+                                      state_ = WebSocketReadState::Len16;
+                                  }
+                                  else
+                                  {
+                                      remaining_length_ = mini_header_ & 0x7f;
+                                      state_ = WebSocketReadState::Mask;
+                                  }
+                                  do_read();
+                              }
+                              else
+                              {
+                                  close_connection_ = true;
+                                  adaptor_.shutdown_readwrite();
+                                  adaptor_.close();
+                                  if (error_handler_)
+                                      error_handler_(*this, ec.message());
+                                  check_destroy();
+                              }
+                          });
+                    }
+                    break;
+                    case WebSocketReadState::Len16:
+                    {
+                        remaining_length_ = 0;
+                        remaining_length16_ = 0;
+                        asio::async_read(
+                          adaptor_.socket(), asio::buffer(&remaining_length16_, 2),
+                          [this](const error_code& ec, std::size_t
+#ifdef CROW_ENABLE_DEBUG
+                                                               bytes_transferred
+#endif
+                          ) {
+                              is_reading = false;
+                              remaining_length16_ = ntohs(remaining_length16_);
+                              remaining_length_ = remaining_length16_;
+#ifdef CROW_ENABLE_DEBUG
+                              if (!ec && bytes_transferred != 2)
+                              {
+                                  throw std::runtime_error("WebSocket:Len16:async_read fail:asio bug?");
+                              }
+#endif
+
+                              if (!ec)
+                              {
+                                  state_ = WebSocketReadState::Mask;
+                                  do_read();
+                              }
+                              else
+                              {
+                                  close_connection_ = true;
+                                  adaptor_.shutdown_readwrite();
+                                  adaptor_.close();
+                                  if (error_handler_)
+                                      error_handler_(*this, ec.message());
+                                  check_destroy();
+                              }
+                          });
+                    }
+                    break;
+                    case WebSocketReadState::Len64:
+                    {
+                        asio::async_read(
+                          adaptor_.socket(), asio::buffer(&remaining_length_, 8),
+                          [this](const error_code& ec, std::size_t
+#ifdef CROW_ENABLE_DEBUG
+                                                               bytes_transferred
+#endif
+                          ) {
+                              is_reading = false;
+                              remaining_length_ = ((1 == ntohl(1)) ? (remaining_length_) : (static_cast<uint64_t>(ntohl((remaining_length_)&0xFFFFFFFF)) << 32) | ntohl((remaining_length_) >> 32));
+#ifdef CROW_ENABLE_DEBUG
+                              if (!ec && bytes_transferred != 8)
+                              {
+                                  throw std::runtime_error("WebSocket:Len16:async_read fail:asio bug?");
+                              }
+#endif
+
+                              if (!ec)
+                              {
+                                  state_ = WebSocketReadState::Mask;
+                                  do_read();
+                              }
+                              else
+                              {
+                                  close_connection_ = true;
+                                  adaptor_.shutdown_readwrite();
+                                  adaptor_.close();
+                                  if (error_handler_)
+                                      error_handler_(*this, ec.message());
+                                  check_destroy();
+                              }
+                          });
+                    }
+                    break;
+                    case WebSocketReadState::Mask:
+                        if (remaining_length_ > max_payload_bytes_)
+                        {
+                            close_connection_ = true;
+                            adaptor_.close();
+                            if (error_handler_)
+                                error_handler_(*this, "Message length exceeds maximum payload.");
+                            check_destroy();
+                        }
+                        else if (has_mask_)
+                        {
+                            asio::async_read(
+                              adaptor_.socket(), asio::buffer((char*)&mask_, 4),
+                              [this](const error_code& ec, std::size_t
+#ifdef CROW_ENABLE_DEBUG
+                                                                   bytes_transferred
+#endif
+                              ) {
+                                  is_reading = false;
+#ifdef CROW_ENABLE_DEBUG
+                                  if (!ec && bytes_transferred != 4)
+                                  {
+                                      throw std::runtime_error("WebSocket:Mask:async_read fail:asio bug?");
+                                  }
+#endif
+
+                                  if (!ec)
+                                  {
+                                      state_ = WebSocketReadState::Payload;
+                                      do_read();
+                                  }
+                                  else
+                                  {
+                                      close_connection_ = true;
+                                      if (error_handler_)
+                                          error_handler_(*this, ec.message());
+                                      adaptor_.shutdown_readwrite();
+                                      adaptor_.close();
+                                      check_destroy();
+                                  }
+                              });
+                        }
+                        else
+                        {
+                            state_ = WebSocketReadState::Payload;
+                            do_read();
+                        }
+                        break;
+                    case WebSocketReadState::Payload:
+                    {
+                        auto to_read = static_cast<std::uint64_t>(buffer_.size());
+                        if (remaining_length_ < to_read)
+                            to_read = remaining_length_;
+                        adaptor_.socket().async_read_some(
+                          asio::buffer(buffer_, static_cast<std::size_t>(to_read)),
+                          [this](const error_code& ec, std::size_t bytes_transferred) {
+                              is_reading = false;
+
+                              if (!ec)
+                              {
+                                  fragment_.insert(fragment_.end(), buffer_.begin(), buffer_.begin() + bytes_transferred);
+                                  remaining_length_ -= bytes_transferred;
+                                  if (remaining_length_ == 0)
+                                  {
+                                      if (handle_fragment())
+                                      {
+                                          state_ = WebSocketReadState::MiniHeader;
+                                          do_read();
+                                      }
+                                  }
+                                  else
+                                      do_read();
+                              }
+                              else
+                              {
+                                  close_connection_ = true;
+                                  if (error_handler_)
+                                      error_handler_(*this, ec.message());
+                                  adaptor_.shutdown_readwrite();
+                                  adaptor_.close();
+                                  check_destroy();
+                              }
+                          });
+                    }
+                    break;
+                }
+            }
+
+            /// Check if the FIN bit is set.
+            bool is_FIN()
+            {
+                return mini_header_ & 0x8000;
+            }
+
+            /// Extract the opcode from the header.
+            int opcode()
+            {
+                return (mini_header_ & 0x0f00) >> 8;
+            }
+
+            /// Process the payload fragment.
+
+            ///
+            /// Unmasks the fragment, checks the opcode, merges fragments into 1 message body, and calls the appropriate handler.
+            bool handle_fragment()
+            {
+                if (has_mask_)
+                {
+                    for (decltype(fragment_.length()) i = 0; i < fragment_.length(); i++)
+                    {
+                        fragment_[i] ^= ((char*)&mask_)[i % 4];
+                    }
+                }
+                switch (opcode())
+                {
+                    case 0: // Continuation
+                    {
+                        message_ += fragment_;
+                        if (is_FIN())
+                        {
+                            if (message_handler_)
+                                message_handler_(*this, message_, is_binary_);
+                            message_.clear();
+                        }
+                    }
+                    break;
+                    case 1: // Text
+                    {
+                        is_binary_ = false;
+                        message_ += fragment_;
+                        if (is_FIN())
+                        {
+                            if (message_handler_)
+                                message_handler_(*this, message_, is_binary_);
+                            message_.clear();
+                        }
+                    }
+                    break;
+                    case 2: // Binary
+                    {
+                        is_binary_ = true;
+                        message_ += fragment_;
+                        if (is_FIN())
+                        {
+                            if (message_handler_)
+                                message_handler_(*this, message_, is_binary_);
+                            message_.clear();
+                        }
+                    }
+                    break;
+                    case 0x8: // Close
+                    {
+                        has_recv_close_ = true;
+                        if (!has_sent_close_)
+                        {
+                            close(fragment_);
+                        }
+                        else
+                        {
+                            adaptor_.shutdown_readwrite();
+                            adaptor_.close();
+                            close_connection_ = true;
+                            if (!is_close_handler_called_)
+                            {
+                                if (close_handler_)
+                                    close_handler_(*this, fragment_);
+                                is_close_handler_called_ = true;
+                            }
+                            check_destroy();
+                            return false;
+                        }
+                    }
+                    break;
+                    case 0x9: // Ping
+                    {
+                        send_pong(fragment_);
+                    }
+                    break;
+                    case 0xA: // Pong
+                    {
+                        pong_received_ = true;
+                    }
+                    break;
+                }
+
+                fragment_.clear();
+                return true;
+            }
+
+            /// Send the buffers' data through the socket.
+
+            ///
+            /// Also destroys the object if the Close flag is set.
+            void do_write()
+            {
+                if (sending_buffers_.empty())
+                {
+                    sending_buffers_.swap(write_buffers_);
+                    std::vector<asio::const_buffer> buffers;
+                    buffers.reserve(sending_buffers_.size());
+                    for (auto& s : sending_buffers_)
+                    {
+                        buffers.emplace_back(asio::buffer(s));
+                    }
+                    auto watch = std::weak_ptr<void>{anchor_};
+                    asio::async_write(
+                      adaptor_.socket(), buffers,
+                      [&, watch](const error_code& ec, std::size_t /*bytes_transferred*/) {
+                          if (!ec && !close_connection_)
+                          {
+                              sending_buffers_.clear();
+                              if (!write_buffers_.empty())
+                                  do_write();
+                              if (has_sent_close_)
+                                  close_connection_ = true;
+                          }
+                          else
+                          {
+                              auto anchor = watch.lock();
+                              if (anchor == nullptr) { return; }
+
+                              sending_buffers_.clear();
+                              close_connection_ = true;
+                              check_destroy();
+                          }
+                      });
+                }
+            }
+
+            /// Destroy the Connection.
+            void check_destroy()
+            {
+                //if (has_sent_close_ && has_recv_close_)
+                if (!is_close_handler_called_)
+                    if (close_handler_)
+                        close_handler_(*this, "uncleanly");
+                handler_->remove_websocket(this);
+                if (sending_buffers_.empty() && !is_reading)
+                    delete this;
+            }
+
+
+            struct SendMessageType
+            {
+                std::string payload;
+                Connection* self;
+                int opcode;
+
+                void operator()()
+                {
+                    self->send_data_impl(this);
+                }
+            };
+
+            void send_data_impl(SendMessageType* s)
+            {
+                auto header = build_header(s->opcode, s->payload.size());
+                write_buffers_.emplace_back(std::move(header));
+                write_buffers_.emplace_back(std::move(s->payload));
+                do_write();
+            }
+
+            void send_data(int opcode, std::string&& msg)
+            {
+                SendMessageType event_arg{
+                  std::move(msg),
+                  this,
+                  opcode};
+
+                post(std::move(event_arg));
+            }
+
+        private:
+            Adaptor adaptor_;
+            Handler* handler_;
+
+            std::vector<std::string> sending_buffers_;
+            std::vector<std::string> write_buffers_;
+
+            std::array<char, 4096> buffer_;
+            bool is_binary_;
+            std::string message_;
+            std::string fragment_;
+            WebSocketReadState state_{WebSocketReadState::MiniHeader};
+            uint16_t remaining_length16_{0};
+            uint64_t remaining_length_{0};
+            uint64_t max_payload_bytes_{UINT64_MAX};
+            bool close_connection_{false};
+            bool is_reading{false};
+            bool has_mask_{false};
+            uint32_t mask_;
+            uint16_t mini_header_;
+            bool has_sent_close_{false};
+            bool has_recv_close_{false};
+            bool error_occurred_{false};
+            bool pong_received_{false};
+            bool is_close_handler_called_{false};
+
+            std::shared_ptr<void> anchor_ = std::make_shared<int>(); // Value is just for placeholding
+
+            std::function<void(crow::websocket::connection&)> open_handler_;
+            std::function<void(crow::websocket::connection&, const std::string&, bool)> message_handler_;
+            std::function<void(crow::websocket::connection&, const std::string&)> close_handler_;
+            std::function<void(crow::websocket::connection&, const std::string&)> error_handler_;
+            std::function<bool(const crow::request&, void**)> accept_handler_;
+        };
+    } // namespace websocket
+} // namespace crow
+
+
+namespace crow
+{
+    constexpr const char VERSION[] = "master";
+}
+
+
+#ifdef CROW_USE_BOOST
+#include <boost/asio.hpp>
+#ifdef CROW_ENABLE_SSL
+#include <boost/asio/ssl.hpp>
+#endif
+#else
+#ifndef ASIO_STANDALONE
+#define ASIO_STANDALONE
+#endif
+#include <asio.hpp>
+#ifdef CROW_ENABLE_SSL
+#include <asio/ssl.hpp>
+#endif
+#endif
+
+#include <atomic>
+#include <chrono>
+#include <cstdint>
+#include <future>
+#include <memory>
+#include <vector>
+
+
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+#ifdef CROW_USE_BOOST
+    namespace asio = boost::asio;
+    using error_code = boost::system::error_code;
+#else
+    using error_code = asio::error_code;
+#endif
+    using tcp = asio::ip::tcp;
+
+    template<typename Handler, typename Adaptor = SocketAdaptor, typename... Middlewares>
+    class Server
+    {
+    public:
+        Server(Handler* handler, std::string bindaddr, uint16_t port, std::string server_name = std::string("Crow/") + VERSION, std::tuple<Middlewares...>* middlewares = nullptr, uint16_t concurrency = 1, uint8_t timeout = 5, typename Adaptor::context* adaptor_ctx = nullptr):
+          acceptor_(io_service_, tcp::endpoint(asio::ip::address::from_string(bindaddr), port)),
+          signals_(io_service_),
+          tick_timer_(io_service_),
+          handler_(handler),
+          concurrency_(concurrency),
+          timeout_(timeout),
+          server_name_(server_name),
+          port_(port),
+          bindaddr_(bindaddr),
+          task_queue_length_pool_(concurrency_ - 1),
+          middlewares_(middlewares),
+          adaptor_ctx_(adaptor_ctx)
+        {}
+
+        void set_tick_function(std::chrono::milliseconds d, std::function<void()> f)
+        {
+            tick_interval_ = d;
+            tick_function_ = f;
+        }
+
+        void on_tick()
+        {
+            tick_function_();
+            tick_timer_.expires_after(std::chrono::milliseconds(tick_interval_.count()));
+            tick_timer_.async_wait([this](const error_code& ec) {
+                if (ec)
+                    return;
+                on_tick();
+            });
+        }
+
+        void run()
+        {
+            uint16_t worker_thread_count = concurrency_ - 1;
+            for (int i = 0; i < worker_thread_count; i++)
+                io_service_pool_.emplace_back(new asio::io_service());
+            get_cached_date_str_pool_.resize(worker_thread_count);
+            task_timer_pool_.resize(worker_thread_count);
+
+            std::vector<std::future<void>> v;
+            std::atomic<int> init_count(0);
+            for (uint16_t i = 0; i < worker_thread_count; i++)
+                v.push_back(
+                  std::async(
+                    std::launch::async, [this, i, &init_count] {
+                        // thread local date string get function
+                        auto last = std::chrono::steady_clock::now();
+
+                        std::string date_str;
+                        auto update_date_str = [&] {
+                            auto last_time_t = time(0);
+                            tm my_tm;
+
+#if defined(_MSC_VER) || defined(__MINGW32__)
+                            gmtime_s(&my_tm, &last_time_t);
+#else
+                            gmtime_r(&last_time_t, &my_tm);
+#endif
+                            date_str.resize(100);
+                            size_t date_str_sz = strftime(&date_str[0], 99, "%a, %d %b %Y %H:%M:%S GMT", &my_tm);
+                            date_str.resize(date_str_sz);
+                        };
+                        update_date_str();
+                        get_cached_date_str_pool_[i] = [&]() -> std::string {
+                            if (std::chrono::steady_clock::now() - last >= std::chrono::seconds(1))
+                            {
+                                last = std::chrono::steady_clock::now();
+                                update_date_str();
+                            }
+                            return date_str;
+                        };
+
+                        // initializing task timers
+                        detail::task_timer task_timer(*io_service_pool_[i]);
+                        task_timer.set_default_timeout(timeout_);
+                        task_timer_pool_[i] = &task_timer;
+                        task_queue_length_pool_[i] = 0;
+
+                        init_count++;
+                        while (1)
+                        {
+                            try
+                            {
+                                if (io_service_pool_[i]->run() == 0)
+                                {
+                                    // when io_service.run returns 0, there are no more works to do.
+                                    break;
+                                }
+                            }
+                            catch (std::exception& e)
+                            {
+                                CROW_LOG_ERROR << "Worker Crash: An uncaught exception occurred: " << e.what();
+                            }
+                        }
+                    }));
+
+            if (tick_function_ && tick_interval_.count() > 0)
+            {
+                tick_timer_.expires_after(std::chrono::milliseconds(tick_interval_.count()));
+                tick_timer_.async_wait(
+                  [this](const error_code& ec) {
+                      if (ec)
+                          return;
+                      on_tick();
+                  });
+            }
+
+            port_ = acceptor_.local_endpoint().port();
+            handler_->port(port_);
+
+
+            CROW_LOG_INFO << server_name_ << " server is running at " << (handler_->ssl_used() ? "https://" : "http://") << bindaddr_ << ":" << acceptor_.local_endpoint().port() << " using " << concurrency_ << " threads";
+            CROW_LOG_INFO << "Call `app.loglevel(crow::LogLevel::Warning)` to hide Info level logs.";
+
+            signals_.async_wait(
+              [&](const error_code& /*error*/, int /*signal_number*/) {
+                  stop();
+              });
+
+            while (worker_thread_count != init_count)
+                std::this_thread::yield();
+
+            do_accept();
+
+            std::thread(
+              [this] {
+                  notify_start();
+                  io_service_.run();
+                  CROW_LOG_INFO << "Exiting.";
+              })
+              .join();
+        }
+
+        void stop()
+        {
+            shutting_down_ = true; // Prevent the acceptor from taking new connections
+            for (auto& io_service : io_service_pool_)
+            {
+                if (io_service != nullptr)
+                {
+                    CROW_LOG_INFO << "Closing IO service " << &io_service;
+                    io_service->stop(); // Close all io_services (and HTTP connections)
+                }
+            }
+
+            CROW_LOG_INFO << "Closing main IO service (" << &io_service_ << ')';
+            io_service_.stop(); // Close main io_service
+        }
+
+        /// Wait until the server has properly started
+        void wait_for_start()
+        {
+            std::unique_lock<std::mutex> lock(start_mutex_);
+            if (!server_started_)
+                cv_started_.wait(lock);
+        }
+
+        void signal_clear()
+        {
+            signals_.clear();
+        }
+
+        void signal_add(int signal_number)
+        {
+            signals_.add(signal_number);
+        }
+
+    private:
+        uint16_t pick_io_service_idx()
+        {
+            uint16_t min_queue_idx = 0;
+
+            // TODO improve load balancing
+            // size_t is used here to avoid the security issue https://codeql.github.com/codeql-query-help/cpp/cpp-comparison-with-wider-type/
+            // even though the max value of this can be only uint16_t as concurrency is uint16_t.
+            for (size_t i = 1; i < task_queue_length_pool_.size() && task_queue_length_pool_[min_queue_idx] > 0; i++)
+            // No need to check other io_services if the current one has no tasks
+            {
+                if (task_queue_length_pool_[i] < task_queue_length_pool_[min_queue_idx])
+                    min_queue_idx = i;
+            }
+            return min_queue_idx;
+        }
+
+        void do_accept()
+        {
+            if (!shutting_down_)
+            {
+                uint16_t service_idx = pick_io_service_idx();
+                asio::io_service& is = *io_service_pool_[service_idx];
+                task_queue_length_pool_[service_idx]++;
+                CROW_LOG_DEBUG << &is << " {" << service_idx << "} queue length: " << task_queue_length_pool_[service_idx];
+
+                auto p = std::make_shared<Connection<Adaptor, Handler, Middlewares...>>(
+                  is, handler_, server_name_, middlewares_,
+                  get_cached_date_str_pool_[service_idx], *task_timer_pool_[service_idx], adaptor_ctx_, task_queue_length_pool_[service_idx]);
+
+                acceptor_.async_accept(
+                  p->socket(),
+                  [this, p, &is, service_idx](error_code ec) {
+                      if (!ec)
+                      {
+                          is.post(
+                            [p] {
+                                p->start();
+                            });
+                      }
+                      else
+                      {
+                          task_queue_length_pool_[service_idx]--;
+                          CROW_LOG_DEBUG << &is << " {" << service_idx << "} queue length: " << task_queue_length_pool_[service_idx];
+                      }
+                      do_accept();
+                  });
+            }
+        }
+
+        /// Notify anything using `wait_for_start()` to proceed
+        void notify_start()
+        {
+            std::unique_lock<std::mutex> lock(start_mutex_);
+            server_started_ = true;
+            cv_started_.notify_all();
+        }
+
+    private:
+        std::vector<std::unique_ptr<asio::io_service>> io_service_pool_;
+        asio::io_service io_service_;
+        std::vector<detail::task_timer*> task_timer_pool_;
+        std::vector<std::function<std::string()>> get_cached_date_str_pool_;
+        tcp::acceptor acceptor_;
+        bool shutting_down_ = false;
+        bool server_started_{false};
+        std::condition_variable cv_started_;
+        std::mutex start_mutex_;
+        asio::signal_set signals_;
+
+        asio::basic_waitable_timer<std::chrono::high_resolution_clock> tick_timer_;
+
+        Handler* handler_;
+        uint16_t concurrency_{2};
+        std::uint8_t timeout_;
+        std::string server_name_;
+        uint16_t port_;
+        std::string bindaddr_;
+        std::vector<std::atomic<unsigned int>> task_queue_length_pool_;
+
+        std::chrono::milliseconds tick_interval_;
+        std::function<void()> tick_function_;
+
+        std::tuple<Middlewares...>* middlewares_;
+
+        typename Adaptor::context* adaptor_ctx_;
+    };
+} // namespace crow
+
+#include <string>
+#include <vector>
+#include <fstream>
+#include <iterator>
+#include <functional>
+
+namespace crow
+{
+    namespace mustache
+    {
+        using context = json::wvalue;
+
+        template_t load(const std::string& filename);
+
+        class invalid_template_exception : public std::exception
+        {
+        public:
+            invalid_template_exception(const std::string& msg):
+              msg("crow::mustache error: " + msg)
+            {}
+            virtual const char* what() const throw()
+            {
+                return msg.c_str();
+            }
+            std::string msg;
+        };
+
+        struct rendered_template : returnable
+        {
+            rendered_template():
+              returnable("text/html") {}
+
+            rendered_template(std::string& body):
+              returnable("text/html"), body_(std::move(body)) {}
+
+            std::string body_;
+
+            std::string dump() const override
+            {
+                return body_;
+            }
+        };
+
+        enum class ActionType
+        {
+            Ignore,
+            Tag,
+            UnescapeTag,
+            OpenBlock,
+            CloseBlock,
+            ElseBlock,
+            Partial,
+        };
+
+        struct Action
+        {
+            int start;
+            int end;
+            int pos;
+            ActionType t;
+            Action(ActionType t, size_t start, size_t end, size_t pos = 0):
+              start(static_cast<int>(start)), end(static_cast<int>(end)), pos(static_cast<int>(pos)), t(t)
+            {
+            }
+        };
+
+        /// A mustache template object.
+        class template_t
+        {
+        public:
+            template_t(std::string body):
+              body_(std::move(body))
+            {
+                // {{ {{# {{/ {{^ {{! {{> {{=
+                parse();
+            }
+
+        private:
+            std::string tag_name(const Action& action) const
+            {
+                return body_.substr(action.start, action.end - action.start);
+            }
+            auto find_context(const std::string& name, const std::vector<const context*>& stack, bool shouldUseOnlyFirstStackValue = false) const -> std::pair<bool, const context&>
+            {
+                if (name == ".")
+                {
+                    return {true, *stack.back()};
+                }
+                static json::wvalue empty_str;
+                empty_str = "";
+
+                int dotPosition = name.find(".");
+                if (dotPosition == static_cast<int>(name.npos))
+                {
+                    for (auto it = stack.rbegin(); it != stack.rend(); ++it)
+                    {
+                        if ((*it)->t() == json::type::Object)
+                        {
+                            if ((*it)->count(name))
+                                return {true, (**it)[name]};
+                        }
+                    }
+                }
+                else
+                {
+                    std::vector<int> dotPositions;
+                    dotPositions.push_back(-1);
+                    while (dotPosition != static_cast<int>(name.npos))
+                    {
+                        dotPositions.push_back(dotPosition);
+                        dotPosition = name.find(".", dotPosition + 1);
+                    }
+                    dotPositions.push_back(name.size());
+                    std::vector<std::string> names;
+                    names.reserve(dotPositions.size() - 1);
+                    for (int i = 1; i < static_cast<int>(dotPositions.size()); i++)
+                        names.emplace_back(name.substr(dotPositions[i - 1] + 1, dotPositions[i] - dotPositions[i - 1] - 1));
+
+                    for (auto it = stack.rbegin(); it != stack.rend(); ++it)
+                    {
+                        const context* view = *it;
+                        bool found = true;
+                        for (auto jt = names.begin(); jt != names.end(); ++jt)
+                        {
+                            if (view->t() == json::type::Object &&
+                                view->count(*jt))
+                            {
+                                view = &(*view)[*jt];
+                            }
+                            else
+                            {
+                                if (shouldUseOnlyFirstStackValue)
+                                {
+                                    return {false, empty_str};
+                                }
+                                found = false;
+                                break;
+                            }
+                        }
+                        if (found)
+                            return {true, *view};
+                    }
+                }
+
+                return {false, empty_str};
+            }
+
+            void escape(const std::string& in, std::string& out) const
+            {
+                out.reserve(out.size() + in.size());
+                for (auto it = in.begin(); it != in.end(); ++it)
+                {
+                    switch (*it)
+                    {
+                        case '&': out += "&amp;"; break;
+                        case '<': out += "&lt;"; break;
+                        case '>': out += "&gt;"; break;
+                        case '"': out += "&quot;"; break;
+                        case '\'': out += "&#39;"; break;
+                        case '/': out += "&#x2F;"; break;
+                        case '`': out += "&#x60;"; break;
+                        case '=': out += "&#x3D;"; break;
+                        default: out += *it; break;
+                    }
+                }
+            }
+
+            bool isTagInsideObjectBlock(const int& current, const std::vector<const context*>& stack) const
+            {
+                int openedBlock = 0;
+                for (int i = current; i > 0; --i)
+                {
+                    auto& action = actions_[i - 1];
+
+                    if (action.t == ActionType::OpenBlock)
+                    {
+                        if (openedBlock == 0 && (*stack.rbegin())->t() == json::type::Object)
+                        {
+                            return true;
+                        }
+                        --openedBlock;
+                    }
+                    else if (action.t == ActionType::CloseBlock)
+                    {
+                        ++openedBlock;
+                    }
+                }
+
+                return false;
+            }
+
+            void render_internal(int actionBegin, int actionEnd, std::vector<const context*>& stack, std::string& out, int indent) const
+            {
+                int current = actionBegin;
+
+                if (indent)
+                    out.insert(out.size(), indent, ' ');
+
+                while (current < actionEnd)
+                {
+                    auto& fragment = fragments_[current];
+                    auto& action = actions_[current];
+                    render_fragment(fragment, indent, out);
+                    switch (action.t)
+                    {
+                        case ActionType::Ignore:
+                            // do nothing
+                            break;
+                        case ActionType::Partial:
+                        {
+                            std::string partial_name = tag_name(action);
+                            auto partial_templ = load(partial_name);
+                            int partial_indent = action.pos;
+                            partial_templ.render_internal(0, partial_templ.fragments_.size() - 1, stack, out, partial_indent ? indent + partial_indent : 0);
+                        }
+                        break;
+                        case ActionType::UnescapeTag:
+                        case ActionType::Tag:
+                        {
+                            bool shouldUseOnlyFirstStackValue = false;
+                            if (isTagInsideObjectBlock(current, stack))
+                            {
+                                shouldUseOnlyFirstStackValue = true;
+                            }
+                            auto optional_ctx = find_context(tag_name(action), stack, shouldUseOnlyFirstStackValue);
+                            auto& ctx = optional_ctx.second;
+                            switch (ctx.t())
+                            {
+                                case json::type::False:
+                                case json::type::True:
+                                case json::type::Number:
+                                    out += ctx.dump();
+                                    break;
+                                case json::type::String:
+                                    if (action.t == ActionType::Tag)
+                                        escape(ctx.s, out);
+                                    else
+                                        out += ctx.s;
+                                    break;
+                                case json::type::Function:
+                                {
+                                    std::string execute_result = ctx.execute();
+                                    while (execute_result.find("{{") != std::string::npos)
+                                    {
+                                        template_t result_plug(execute_result);
+                                        execute_result = result_plug.render_string(*(stack[0]));
+                                    }
+
+                                    if (action.t == ActionType::Tag)
+                                        escape(execute_result, out);
+                                    else
+                                        out += execute_result;
+                                }
+                                break;
+                                default:
+                                    throw std::runtime_error("not implemented tag type" + utility::lexical_cast<std::string>(static_cast<int>(ctx.t())));
+                            }
+                        }
+                        break;
+                        case ActionType::ElseBlock:
+                        {
+                            static context nullContext;
+                            auto optional_ctx = find_context(tag_name(action), stack);
+                            if (!optional_ctx.first)
+                            {
+                                stack.emplace_back(&nullContext);
+                                break;
+                            }
+
+                            auto& ctx = optional_ctx.second;
+                            switch (ctx.t())
+                            {
+                                case json::type::List:
+                                    if (ctx.l && !ctx.l->empty())
+                                        current = action.pos;
+                                    else
+                                        stack.emplace_back(&nullContext);
+                                    break;
+                                case json::type::False:
+                                case json::type::Null:
+                                    stack.emplace_back(&nullContext);
+                                    break;
+                                default:
+                                    current = action.pos;
+                                    break;
+                            }
+                            break;
+                        }
+                        case ActionType::OpenBlock:
+                        {
+                            auto optional_ctx = find_context(tag_name(action), stack);
+                            if (!optional_ctx.first)
+                            {
+                                current = action.pos;
+                                break;
+                            }
+
+                            auto& ctx = optional_ctx.second;
+                            switch (ctx.t())
+                            {
+                                case json::type::List:
+                                    if (ctx.l)
+                                        for (auto it = ctx.l->begin(); it != ctx.l->end(); ++it)
+                                        {
+                                            stack.push_back(&*it);
+                                            render_internal(current + 1, action.pos, stack, out, indent);
+                                            stack.pop_back();
+                                        }
+                                    current = action.pos;
+                                    break;
+                                case json::type::Number:
+                                case json::type::String:
+                                case json::type::Object:
+                                case json::type::True:
+                                    stack.push_back(&ctx);
+                                    break;
+                                case json::type::False:
+                                case json::type::Null:
+                                    current = action.pos;
+                                    break;
+                                default:
+                                    throw std::runtime_error("{{#: not implemented context type: " + utility::lexical_cast<std::string>(static_cast<int>(ctx.t())));
+                                    break;
+                            }
+                            break;
+                        }
+                        case ActionType::CloseBlock:
+                            stack.pop_back();
+                            break;
+                        default:
+                            throw std::runtime_error("not implemented " + utility::lexical_cast<std::string>(static_cast<int>(action.t)));
+                    }
+                    current++;
+                }
+                auto& fragment = fragments_[actionEnd];
+                render_fragment(fragment, indent, out);
+            }
+            void render_fragment(const std::pair<int, int> fragment, int indent, std::string& out) const
+            {
+                if (indent)
+                {
+                    for (int i = fragment.first; i < fragment.second; i++)
+                    {
+                        out += body_[i];
+                        if (body_[i] == '\n' && i + 1 != static_cast<int>(body_.size()))
+                            out.insert(out.size(), indent, ' ');
+                    }
+                }
+                else
+                    out.insert(out.size(), body_, fragment.first, fragment.second - fragment.first);
+            }
+
+        public:
+            /// Output a returnable template from this mustache template
+            rendered_template render() const
+            {
+                context empty_ctx;
+                std::vector<const context*> stack;
+                stack.emplace_back(&empty_ctx);
+
+                std::string ret;
+                render_internal(0, fragments_.size() - 1, stack, ret, 0);
+                return rendered_template(ret);
+            }
+
+            /// Apply the values from the context provided and output a returnable template from this mustache template
+            rendered_template render(const context& ctx) const
+            {
+                std::vector<const context*> stack;
+                stack.emplace_back(&ctx);
+
+                std::string ret;
+                render_internal(0, fragments_.size() - 1, stack, ret, 0);
+                return rendered_template(ret);
+            }
+
+            /// Apply the values from the context provided and output a returnable template from this mustache template
+            rendered_template render(const context&& ctx) const
+            {
+                return render(ctx);
+            }
+
+            /// Output a returnable template from this mustache template
+            std::string render_string() const
+            {
+                context empty_ctx;
+                std::vector<const context*> stack;
+                stack.emplace_back(&empty_ctx);
+
+                std::string ret;
+                render_internal(0, fragments_.size() - 1, stack, ret, 0);
+                return ret;
+            }
+
+            /// Apply the values from the context provided and output a returnable template from this mustache template
+            std::string render_string(const context& ctx) const
+            {
+                std::vector<const context*> stack;
+                stack.emplace_back(&ctx);
+
+                std::string ret;
+                render_internal(0, fragments_.size() - 1, stack, ret, 0);
+                return ret;
+            }
+
+        private:
+            void parse()
+            {
+                std::string tag_open = "{{";
+                std::string tag_close = "}}";
+
+                std::vector<int> blockPositions;
+
+                size_t current = 0;
+                while (1)
+                {
+                    size_t idx = body_.find(tag_open, current);
+                    if (idx == body_.npos)
+                    {
+                        fragments_.emplace_back(static_cast<int>(current), static_cast<int>(body_.size()));
+                        actions_.emplace_back(ActionType::Ignore, 0, 0);
+                        break;
+                    }
+                    fragments_.emplace_back(static_cast<int>(current), static_cast<int>(idx));
+
+                    idx += tag_open.size();
+                    size_t endIdx = body_.find(tag_close, idx);
+                    if (endIdx == idx)
+                    {
+                        throw invalid_template_exception("empty tag is not allowed");
+                    }
+                    if (endIdx == body_.npos)
+                    {
+                        // error, no matching tag
+                        throw invalid_template_exception("not matched opening tag");
+                    }
+                    current = endIdx + tag_close.size();
+                    switch (body_[idx])
+                    {
+                        case '#':
+                            idx++;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            blockPositions.emplace_back(static_cast<int>(actions_.size()));
+                            actions_.emplace_back(ActionType::OpenBlock, idx, endIdx);
+                            break;
+                        case '/':
+                            idx++;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            {
+                                auto& matched = actions_[blockPositions.back()];
+                                if (body_.compare(idx, endIdx - idx,
+                                                  body_, matched.start, matched.end - matched.start) != 0)
+                                {
+                                    throw invalid_template_exception("not matched {{# {{/ pair: " +
+                                                                     body_.substr(matched.start, matched.end - matched.start) + ", " +
+                                                                     body_.substr(idx, endIdx - idx));
+                                }
+                                matched.pos = actions_.size();
+                            }
+                            actions_.emplace_back(ActionType::CloseBlock, idx, endIdx, blockPositions.back());
+                            blockPositions.pop_back();
+                            break;
+                        case '^':
+                            idx++;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            blockPositions.emplace_back(static_cast<int>(actions_.size()));
+                            actions_.emplace_back(ActionType::ElseBlock, idx, endIdx);
+                            break;
+                        case '!':
+                            // do nothing action
+                            actions_.emplace_back(ActionType::Ignore, idx + 1, endIdx);
+                            break;
+                        case '>': // partial
+                            idx++;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            actions_.emplace_back(ActionType::Partial, idx, endIdx);
+                            break;
+                        case '{':
+                            if (tag_open != "{{" || tag_close != "}}")
+                                throw invalid_template_exception("cannot use triple mustache when delimiter changed");
+
+                            idx++;
+                            if (body_[endIdx + 2] != '}')
+                            {
+                                throw invalid_template_exception("{{{: }}} not matched");
+                            }
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            actions_.emplace_back(ActionType::UnescapeTag, idx, endIdx);
+                            current++;
+                            break;
+                        case '&':
+                            idx++;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            actions_.emplace_back(ActionType::UnescapeTag, idx, endIdx);
+                            break;
+                        case '=':
+                            // tag itself is no-op
+                            idx++;
+                            actions_.emplace_back(ActionType::Ignore, idx, endIdx);
+                            endIdx--;
+                            if (body_[endIdx] != '=')
+                                throw invalid_template_exception("{{=: not matching = tag: " + body_.substr(idx, endIdx - idx));
+                            endIdx--;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx] == ' ')
+                                endIdx--;
+                            endIdx++;
+                            {
+                                bool succeeded = false;
+                                for (size_t i = idx; i < endIdx; i++)
+                                {
+                                    if (body_[i] == ' ')
+                                    {
+                                        tag_open = body_.substr(idx, i - idx);
+                                        while (body_[i] == ' ')
+                                            i++;
+                                        tag_close = body_.substr(i, endIdx - i);
+                                        if (tag_open.empty())
+                                            throw invalid_template_exception("{{=: empty open tag");
+                                        if (tag_close.empty())
+                                            throw invalid_template_exception("{{=: empty close tag");
+
+                                        if (tag_close.find(" ") != tag_close.npos)
+                                            throw invalid_template_exception("{{=: invalid open/close tag: " + tag_open + " " + tag_close);
+                                        succeeded = true;
+                                        break;
+                                    }
+                                }
+                                if (!succeeded)
+                                    throw invalid_template_exception("{{=: cannot find space between new open/close tags");
+                            }
+                            break;
+                        default:
+                            // normal tag case;
+                            while (body_[idx] == ' ')
+                                idx++;
+                            while (body_[endIdx - 1] == ' ')
+                                endIdx--;
+                            actions_.emplace_back(ActionType::Tag, idx, endIdx);
+                            break;
+                    }
+                }
+
+                // removing standalones
+                for (int i = actions_.size() - 2; i >= 0; i--)
+                {
+                    if (actions_[i].t == ActionType::Tag || actions_[i].t == ActionType::UnescapeTag)
+                        continue;
+                    auto& fragment_before = fragments_[i];
+                    auto& fragment_after = fragments_[i + 1];
+                    bool is_last_action = i == static_cast<int>(actions_.size()) - 2;
+                    bool all_space_before = true;
+                    int j, k;
+                    for (j = fragment_before.second - 1; j >= fragment_before.first; j--)
+                    {
+                        if (body_[j] != ' ')
+                        {
+                            all_space_before = false;
+                            break;
+                        }
+                    }
+                    if (all_space_before && i > 0)
+                        continue;
+                    if (!all_space_before && body_[j] != '\n')
+                        continue;
+                    bool all_space_after = true;
+                    for (k = fragment_after.first; k < static_cast<int>(body_.size()) && k < fragment_after.second; k++)
+                    {
+                        if (body_[k] != ' ')
+                        {
+                            all_space_after = false;
+                            break;
+                        }
+                    }
+                    if (all_space_after && !is_last_action)
+                        continue;
+                    if (!all_space_after &&
+                        !(
+                          body_[k] == '\n' ||
+                          (body_[k] == '\r' &&
+                           k + 1 < static_cast<int>(body_.size()) &&
+                           body_[k + 1] == '\n')))
+                        continue;
+                    if (actions_[i].t == ActionType::Partial)
+                    {
+                        actions_[i].pos = fragment_before.second - j - 1;
+                    }
+                    fragment_before.second = j + 1;
+                    if (!all_space_after)
+                    {
+                        if (body_[k] == '\n')
+                            k++;
+                        else
+                            k += 2;
+                        fragment_after.first = k;
+                    }
+                }
+            }
+
+            std::vector<std::pair<int, int>> fragments_;
+            std::vector<Action> actions_;
+            std::string body_;
+        };
+
+        inline template_t compile(const std::string& body)
+        {
+            return template_t(body);
+        }
+        namespace detail
+        {
+            inline std::string& get_template_base_directory_ref()
+            {
+                static std::string template_base_directory = "templates";
+                return template_base_directory;
+            }
+
+            /// A base directory not related to any blueprint
+            inline std::string& get_global_template_base_directory_ref()
+            {
+                static std::string template_base_directory = "templates";
+                return template_base_directory;
+            }
+        } // namespace detail
+
+        inline std::string default_loader(const std::string& filename)
+        {
+            std::string path = detail::get_template_base_directory_ref();
+            std::ifstream inf(utility::join_path(path, filename));
+            if (!inf)
+            {
+                CROW_LOG_WARNING << "Template \"" << filename << "\" not found.";
+                return {};
+            }
+            return {std::istreambuf_iterator<char>(inf), std::istreambuf_iterator<char>()};
+        }
+
+        namespace detail
+        {
+            inline std::function<std::string(std::string)>& get_loader_ref()
+            {
+                static std::function<std::string(std::string)> loader = default_loader;
+                return loader;
+            }
+        } // namespace detail
+
+        inline void set_base(const std::string& path)
+        {
+            auto& base = detail::get_template_base_directory_ref();
+            base = path;
+            if (base.back() != '\\' &&
+                base.back() != '/')
+            {
+                base += '/';
+            }
+        }
+
+        inline void set_global_base(const std::string& path)
+        {
+            auto& base = detail::get_global_template_base_directory_ref();
+            base = path;
+            if (base.back() != '\\' &&
+                base.back() != '/')
+            {
+                base += '/';
+            }
+        }
+
+        inline void set_loader(std::function<std::string(std::string)> loader)
+        {
+            detail::get_loader_ref() = std::move(loader);
+        }
+
+        inline std::string load_text(const std::string& filename)
+        {
+            std::string filename_sanitized(filename);
+            utility::sanitize_filename(filename_sanitized);
+            return detail::get_loader_ref()(filename_sanitized);
+        }
+
+        inline std::string load_text_unsafe(const std::string& filename)
+        {
+            return detail::get_loader_ref()(filename);
+        }
+
+        inline template_t load(const std::string& filename)
+        {
+            std::string filename_sanitized(filename);
+            utility::sanitize_filename(filename_sanitized);
+            return compile(detail::get_loader_ref()(filename_sanitized));
+        }
+
+        inline template_t load_unsafe(const std::string& filename)
+        {
+            return compile(detail::get_loader_ref()(filename));
+        }
+    } // namespace mustache
+} // namespace crow
+
+
+#include <cstdint>
+#include <utility>
+#include <tuple>
+#include <unordered_map>
+#include <memory>
+#include <vector>
+#include <algorithm>
+#include <type_traits>
+
+
+namespace crow // NOTE: Already documented in "crow/app.h"
+{
+
+    constexpr const uint16_t INVALID_BP_ID{((uint16_t)-1)};
+
+    namespace detail
+    {
+        /// Typesafe wrapper for storing lists of middleware as their indices in the App
+        struct middleware_indices
+        {
+            template<typename App>
+            void push()
+            {}
+
+            template<typename App, typename MW, typename... Middlewares>
+            void push()
+            {
+                using MwContainer = typename App::mw_container_t;
+                static_assert(black_magic::has_type<MW, MwContainer>::value, "Middleware must be present in app");
+                static_assert(std::is_base_of<crow::ILocalMiddleware, MW>::value, "Middleware must extend ILocalMiddleware");
+                int idx = black_magic::tuple_index<MW, MwContainer>::value;
+                indices_.push_back(idx);
+                push<App, Middlewares...>();
+            }
+
+            void merge_front(const detail::middleware_indices& other)
+            {
+                indices_.insert(indices_.begin(), other.indices_.cbegin(), other.indices_.cend());
+            }
+
+            void merge_back(const detail::middleware_indices& other)
+            {
+                indices_.insert(indices_.end(), other.indices_.cbegin(), other.indices_.cend());
+            }
+
+            void pop_back(const detail::middleware_indices& other)
+            {
+                indices_.resize(indices_.size() - other.indices_.size());
+            }
+
+            bool empty() const
+            {
+                return indices_.empty();
+            }
+
+            // Sorts indices and filters out duplicates to allow fast lookups with traversal
+            void pack()
+            {
+                std::sort(indices_.begin(), indices_.end());
+                indices_.erase(std::unique(indices_.begin(), indices_.end()), indices_.end());
+            }
+
+            const std::vector<int>& indices()
+            {
+                return indices_;
+            }
+
+        private:
+            std::vector<int> indices_;
+        };
+    } // namespace detail
+
+    /// A base class for all rules.
+
+    ///
+    /// Used to provide a common interface for code dealing with different types of rules.<br>
+    /// A Rule provides a URL, allowed HTTP methods, and handlers.
+    class BaseRule
+    {
+    public:
+        BaseRule(std::string rule):
+          rule_(std::move(rule))
+        {}
+
+        virtual ~BaseRule()
+        {}
+
+        virtual void validate() = 0;
+        
+        void set_added() {
+            added_ = true;
+        }
+
+        bool is_added() {
+            return added_;
+        }
+
+        std::unique_ptr<BaseRule> upgrade()
+        {
+            if (rule_to_upgrade_)
+                return std::move(rule_to_upgrade_);
+            return {};
+        }
+
+        virtual void handle(request&, response&, const routing_params&) = 0;
+        virtual void handle_upgrade(const request&, response& res, SocketAdaptor&&)
+        {
+            res = response(404);
+            res.end();
+        }
+#ifdef CROW_ENABLE_SSL
+        virtual void handle_upgrade(const request&, response& res, SSLAdaptor&&)
+        {
+            res = response(404);
+            res.end();
+        }
+#endif
+
+        uint32_t get_methods()
+        {
+            return methods_;
+        }
+
+        template<typename F>
+        void foreach_method(F f)
+        {
+            for (uint32_t method = 0, method_bit = 1; method < static_cast<uint32_t>(HTTPMethod::InternalMethodCount); method++, method_bit <<= 1)
+            {
+                if (methods_ & method_bit)
+                    f(method);
+            }
+        }
+
+        std::string custom_templates_base;
+
+        const std::string& rule() { return rule_; }
+
+    protected:
+        uint32_t methods_{1 << static_cast<int>(HTTPMethod::Get)};
+
+        std::string rule_;
+        std::string name_;
+        bool added_{false};
+
+        std::unique_ptr<BaseRule> rule_to_upgrade_;
+
+        detail::middleware_indices mw_indices_;
+
+        friend class Router;
+        friend class Blueprint;
+        template<typename T>
+        friend struct RuleParameterTraits;
+    };
+
+
+    namespace detail
+    {
+        namespace routing_handler_call_helper
+        {
+            template<typename T, int Pos>
+            struct call_pair
+            {
+                using type = T;
+                static const int pos = Pos;
+            };
+
+            template<typename H1>
+            struct call_params
+            {
+                H1& handler;
+                const routing_params& params;
+                request& req;
+                response& res;
+            };
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename S1, typename S2>
+            struct call
+            {};
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename... Args1, typename... Args2>
+            struct call<F, NInt, NUint, NDouble, NString, black_magic::S<int64_t, Args1...>, black_magic::S<Args2...>>
+            {
+                void operator()(F cparams)
+                {
+                    using pushed = typename black_magic::S<Args2...>::template push_back<call_pair<int64_t, NInt>>;
+                    call<F, NInt + 1, NUint, NDouble, NString, black_magic::S<Args1...>, pushed>()(cparams);
+                }
+            };
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename... Args1, typename... Args2>
+            struct call<F, NInt, NUint, NDouble, NString, black_magic::S<uint64_t, Args1...>, black_magic::S<Args2...>>
+            {
+                void operator()(F cparams)
+                {
+                    using pushed = typename black_magic::S<Args2...>::template push_back<call_pair<uint64_t, NUint>>;
+                    call<F, NInt, NUint + 1, NDouble, NString, black_magic::S<Args1...>, pushed>()(cparams);
+                }
+            };
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename... Args1, typename... Args2>
+            struct call<F, NInt, NUint, NDouble, NString, black_magic::S<double, Args1...>, black_magic::S<Args2...>>
+            {
+                void operator()(F cparams)
+                {
+                    using pushed = typename black_magic::S<Args2...>::template push_back<call_pair<double, NDouble>>;
+                    call<F, NInt, NUint, NDouble + 1, NString, black_magic::S<Args1...>, pushed>()(cparams);
+                }
+            };
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename... Args1, typename... Args2>
+            struct call<F, NInt, NUint, NDouble, NString, black_magic::S<std::string, Args1...>, black_magic::S<Args2...>>
+            {
+                void operator()(F cparams)
+                {
+                    using pushed = typename black_magic::S<Args2...>::template push_back<call_pair<std::string, NString>>;
+                    call<F, NInt, NUint, NDouble, NString + 1, black_magic::S<Args1...>, pushed>()(cparams);
+                }
+            };
+
+            template<typename F, int NInt, int NUint, int NDouble, int NString, typename... Args1>
+            struct call<F, NInt, NUint, NDouble, NString, black_magic::S<>, black_magic::S<Args1...>>
+            {
+                void operator()(F cparams)
+                {
+                    cparams.handler(
+                      cparams.req,
+                      cparams.res,
+                      cparams.params.template get<typename Args1::type>(Args1::pos)...);
+                }
+            };
+
+            template<typename Func, typename... ArgsWrapped>
+            struct Wrapped
+            {
+                template<typename... Args>
+                void set_(Func f, typename std::enable_if<!std::is_same<typename std::tuple_element<0, std::tuple<Args..., void>>::type, const request&>::value, int>::type = 0)
+                {
+                    handler_ = (
+#ifdef CROW_CAN_USE_CPP14
+                      [f = std::move(f)]
+#else
+                      [f]
+#endif
+                      (const request&, response& res, Args... args) {
+                          res = response(f(args...));
+                          res.end();
+                      });
+                }
+
+                template<typename Req, typename... Args>
+                struct req_handler_wrapper
+                {
+                    req_handler_wrapper(Func f):
+                      f(std::move(f))
+                    {
+                    }
+
+                    void operator()(const request& req, response& res, Args... args)
+                    {
+                        res = response(f(req, args...));
+                        res.end();
+                    }
+
+                    Func f;
+                };
+
+                template<typename... Args>
+                void set_(Func f, typename std::enable_if<
+                                    std::is_same<typename std::tuple_element<0, std::tuple<Args..., void>>::type, const request&>::value &&
+                                      !std::is_same<typename std::tuple_element<1, std::tuple<Args..., void, void>>::type, response&>::value,
+                                    int>::type = 0)
+                {
+                    handler_ = req_handler_wrapper<Args...>(std::move(f));
+                    /*handler_ = (
+                        [f = std::move(f)]
+                        (const request& req, response& res, Args... args){
+                             res = response(f(req, args...));
+                             res.end();
+                        });*/
+                }
+
+                template<typename... Args>
+                void set_(Func f, typename std::enable_if<
+                                    std::is_same<typename std::tuple_element<0, std::tuple<Args..., void>>::type, const request&>::value &&
+                                      std::is_same<typename std::tuple_element<1, std::tuple<Args..., void, void>>::type, response&>::value,
+                                    int>::type = 0)
+                {
+                    handler_ = std::move(f);
+                }
+
+                template<typename... Args>
+                struct handler_type_helper
+                {
+                    using type = std::function<void(const crow::request&, crow::response&, Args...)>;
+                    using args_type = black_magic::S<typename black_magic::promote_t<Args>...>;
+                };
+
+                template<typename... Args>
+                struct handler_type_helper<const request&, Args...>
+                {
+                    using type = std::function<void(const crow::request&, crow::response&, Args...)>;
+                    using args_type = black_magic::S<typename black_magic::promote_t<Args>...>;
+                };
+
+                template<typename... Args>
+                struct handler_type_helper<const request&, response&, Args...>
+                {
+                    using type = std::function<void(const crow::request&, crow::response&, Args...)>;
+                    using args_type = black_magic::S<typename black_magic::promote_t<Args>...>;
+                };
+
+                typename handler_type_helper<ArgsWrapped...>::type handler_;
+
+                void operator()(request& req, response& res, const routing_params& params)
+                {
+                    detail::routing_handler_call_helper::call<
+                      detail::routing_handler_call_helper::call_params<
+                        decltype(handler_)>,
+                      0, 0, 0, 0,
+                      typename handler_type_helper<ArgsWrapped...>::args_type,
+                      black_magic::S<>>()(
+                      detail::routing_handler_call_helper::call_params<
+                        decltype(handler_)>{handler_, params, req, res});
+                }
+            };
+
+        } // namespace routing_handler_call_helper
+    }     // namespace detail
+
+
+    class CatchallRule
+    {
+    public:
+        /// @cond SKIP
+        CatchallRule() {}
+
+        template<typename Func>
+        typename std::enable_if<black_magic::CallHelper<Func, black_magic::S<>>::value, void>::type
+          operator()(Func&& f)
+        {
+            static_assert(!std::is_same<void, decltype(f())>::value,
+                          "Handler function cannot have void return type; valid return types: string, int, crow::response, crow::returnable");
+
+            handler_ = (
+#ifdef CROW_CAN_USE_CPP14
+              [f = std::move(f)]
+#else
+              [f]
+#endif
+              (const request&, response& res) {
+                  res = response(f());
+                  res.end();
+              });
+        }
+
+        template<typename Func>
+        typename std::enable_if<
+          !black_magic::CallHelper<Func, black_magic::S<>>::value &&
+            black_magic::CallHelper<Func, black_magic::S<crow::request>>::value,
+          void>::type
+          operator()(Func&& f)
+        {
+            static_assert(!std::is_same<void, decltype(f(std::declval<crow::request>()))>::value,
+                          "Handler function cannot have void return type; valid return types: string, int, crow::response, crow::returnable");
+
+            handler_ = (
+#ifdef CROW_CAN_USE_CPP14
+              [f = std::move(f)]
+#else
+              [f]
+#endif
+              (const crow::request& req, crow::response& res) {
+                  res = response(f(req));
+                  res.end();
+              });
+        }
+
+        template<typename Func>
+        typename std::enable_if<
+          !black_magic::CallHelper<Func, black_magic::S<>>::value &&
+            !black_magic::CallHelper<Func, black_magic::S<crow::request>>::value &&
+            black_magic::CallHelper<Func, black_magic::S<crow::response&>>::value,
+          void>::type
+          operator()(Func&& f)
+        {
+            static_assert(std::is_same<void, decltype(f(std::declval<crow::response&>()))>::value,
+                          "Handler function with response argument should have void return type");
+            handler_ = (
+#ifdef CROW_CAN_USE_CPP14
+              [f = std::move(f)]
+#else
+              [f]
+#endif
+              (const crow::request&, crow::response& res) {
+                  f(res);
+              });
+        }
+
+        template<typename Func>
+        typename std::enable_if<
+          !black_magic::CallHelper<Func, black_magic::S<>>::value &&
+            !black_magic::CallHelper<Func, black_magic::S<crow::request>>::value &&
+            !black_magic::CallHelper<Func, black_magic::S<crow::response&>>::value,
+          void>::type
+          operator()(Func&& f)
+        {
+            static_assert(std::is_same<void, decltype(f(std::declval<crow::request>(), std::declval<crow::response&>()))>::value,
+                          "Handler function with response argument should have void return type");
+
+            handler_ = std::move(f);
+        }
+        /// @endcond
+        bool has_handler()
+        {
+            return (handler_ != nullptr);
+        }
+
+    protected:
+        friend class Router;
+
+    private:
+        std::function<void(const crow::request&, crow::response&)> handler_;
+    };
+
+
+    /// A rule dealing with websockets.
+
+    ///
+    /// Provides the interface for the user to put in the necessary handlers for a websocket to work.
+    template<typename App>
+    class WebSocketRule : public BaseRule
+    {
+        using self_t = WebSocketRule;
+
+    public:
+        WebSocketRule(std::string rule, App* app):
+          BaseRule(std::move(rule)),
+          app_(app),
+          max_payload_(UINT64_MAX)
+        {}
+
+        void validate() override
+        {}
+
+        void handle(request&, response& res, const routing_params&) override
+        {
+            res = response(404);
+            res.end();
+        }
+
+        void handle_upgrade(const request& req, response&, SocketAdaptor&& adaptor) override
+        {
+            max_payload_ = max_payload_override_ ? max_payload_ : app_->websocket_max_payload();
+            new crow::websocket::Connection<SocketAdaptor, App>(req, std::move(adaptor), app_, max_payload_, open_handler_, message_handler_, close_handler_, error_handler_, accept_handler_);
+        }
+#ifdef CROW_ENABLE_SSL
+        void handle_upgrade(const request& req, response&, SSLAdaptor&& adaptor) override
+        {
+            new crow::websocket::Connection<SSLAdaptor, App>(req, std::move(adaptor), app_, max_payload_, open_handler_, message_handler_, close_handler_, error_handler_, accept_handler_);
+        }
+#endif
+
+        /// Override the global payload limit for this single WebSocket rule
+        self_t& max_payload(uint64_t max_payload)
+        {
+            max_payload_ = max_payload;
+            max_payload_override_ = true;
+            return *this;
+        }
+
+        template<typename Func>
+        self_t& onopen(Func f)
+        {
+            open_handler_ = f;
+            return *this;
+        }
+
+        template<typename Func>
+        self_t& onmessage(Func f)
+        {
+            message_handler_ = f;
+            return *this;
+        }
+
+        template<typename Func>
+        self_t& onclose(Func f)
+        {
+            close_handler_ = f;
+            return *this;
+        }
+
+        template<typename Func>
+        self_t& onerror(Func f)
+        {
+            error_handler_ = f;
+            return *this;
+        }
+
+        template<typename Func>
+        self_t& onaccept(Func f)
+        {
+            accept_handler_ = f;
+            return *this;
+        }
+
+    protected:
+        App* app_;
+        std::function<void(crow::websocket::connection&)> open_handler_;
+        std::function<void(crow::websocket::connection&, const std::string&, bool)> message_handler_;
+        std::function<void(crow::websocket::connection&, const std::string&)> close_handler_;
+        std::function<void(crow::websocket::connection&, const std::string&)> error_handler_;
+        std::function<bool(const crow::request&, void**)> accept_handler_;
+        uint64_t max_payload_;
+        bool max_payload_override_ = false;
+    };
+
+    /// Allows the user to assign parameters using functions.
+
+    ///
+    /// `rule.name("name").methods(HTTPMethod::POST)`
+    template<typename T>
+    struct RuleParameterTraits
+    {
+        using self_t = T;
+
+        template<typename App>
+        WebSocketRule<App>& websocket(App* app)
+        {
+            auto p = new WebSocketRule<App>(static_cast<self_t*>(this)->rule_, app);
+            static_cast<self_t*>(this)->rule_to_upgrade_.reset(p);
+            return *p;
+        }
+
+        self_t& name(std::string name) noexcept
+        {
+            static_cast<self_t*>(this)->name_ = std::move(name);
+            return static_cast<self_t&>(*this);
+        }
+
+        self_t& methods(HTTPMethod method)
+        {
+            static_cast<self_t*>(this)->methods_ = 1 << static_cast<int>(method);
+            return static_cast<self_t&>(*this);
+        }
+
+        template<typename... MethodArgs>
+        self_t& methods(HTTPMethod method, MethodArgs... args_method)
+        {
+            methods(args_method...);
+            static_cast<self_t*>(this)->methods_ |= 1 << static_cast<int>(method);
+            return static_cast<self_t&>(*this);
+        }
+
+        /// Enable local middleware for this handler
+        template<typename App, typename... Middlewares>
+        self_t& middlewares()
+        {
+            static_cast<self_t*>(this)->mw_indices_.template push<App, Middlewares...>();
+            return static_cast<self_t&>(*this);
+        }
+    };
+
+    /// A rule that can change its parameters during runtime.
+    class DynamicRule : public BaseRule, public RuleParameterTraits<DynamicRule>
+    {
+    public:
+        DynamicRule(std::string rule):
+          BaseRule(std::move(rule))
+        {}
+
+        void validate() override
+        {
+            if (!erased_handler_)
+            {
+                throw std::runtime_error(name_ + (!name_.empty() ? ": " : "") + "no handler for url " + rule_);
+            }
+        }
+
+        void handle(request& req, response& res, const routing_params& params) override
+        {
+            if (!custom_templates_base.empty())
+                mustache::set_base(custom_templates_base);
+            else if (mustache::detail::get_template_base_directory_ref() != "templates")
+                mustache::set_base("templates");
+            erased_handler_(req, res, params);
+        }
+
+        template<typename Func>
+        void operator()(Func f)
+        {
+#ifdef CROW_MSVC_WORKAROUND
+            using function_t = utility::function_traits<decltype(&Func::operator())>;
+#else
+            using function_t = utility::function_traits<Func>;
+#endif
+            erased_handler_ = wrap(std::move(f), black_magic::gen_seq<function_t::arity>());
+        }
+
+        // enable_if Arg1 == request && Arg2 == response
+        // enable_if Arg1 == request && Arg2 != resposne
+        // enable_if Arg1 != request
+#ifdef CROW_MSVC_WORKAROUND
+        template<typename Func, size_t... Indices>
+#else
+        template<typename Func, unsigned... Indices>
+#endif
+        std::function<void(request&, response&, const routing_params&)>
+          wrap(Func f, black_magic::seq<Indices...>)
+        {
+#ifdef CROW_MSVC_WORKAROUND
+            using function_t = utility::function_traits<decltype(&Func::operator())>;
+#else
+            using function_t = utility::function_traits<Func>;
+#endif
+            if (!black_magic::is_parameter_tag_compatible(
+                  black_magic::get_parameter_tag_runtime(rule_.c_str()),
+                  black_magic::compute_parameter_tag_from_args_list<
+                    typename function_t::template arg<Indices>...>::value))
+            {
+                throw std::runtime_error("route_dynamic: Handler type is mismatched with URL parameters: " + rule_);
+            }
+            auto ret = detail::routing_handler_call_helper::Wrapped<Func, typename function_t::template arg<Indices>...>();
+            ret.template set_<
+              typename function_t::template arg<Indices>...>(std::move(f));
+            return ret;
+        }
+
+        template<typename Func>
+        void operator()(std::string name, Func&& f)
+        {
+            name_ = std::move(name);
+            (*this).template operator()<Func>(std::forward(f));
+        }
+
+    private:
+        std::function<void(request&, response&, const routing_params&)> erased_handler_;
+    };
+
+    /// Default rule created when CROW_ROUTE is called.
+    template<typename... Args>
+    class TaggedRule : public BaseRule, public RuleParameterTraits<TaggedRule<Args...>>
+    {
+    public:
+        using self_t = TaggedRule<Args...>;
+
+        TaggedRule(std::string rule):
+          BaseRule(std::move(rule))
+        {}
+
+        void validate() override
+        {
+            if (rule_.at(0) != '/')
+                throw std::runtime_error("Internal error: Routes must start with a '/'");
+
+            if (!handler_)
+            {
+                throw std::runtime_error(name_ + (!name_.empty() ? ": " : "") + "no handler for url " + rule_);
+            }
+        }
+
+        template<typename Func>
+        void operator()(Func&& f)
+        {
+            handler_ = (
+#ifdef CROW_CAN_USE_CPP14
+              [f = std::move(f)]
+#else
+              [f]
+#endif
+              (crow::request& req, crow::response& res, Args... args) {
+                  detail::wrapped_handler_call(req, res, f, std::forward<Args>(args)...);
+              });
+        }
+
+        template<typename Func>
+        void operator()(std::string name, Func&& f)
+        {
+            name_ = std::move(name);
+            (*this).template operator()<Func>(std::forward(f));
+        }
+
+        void handle(request& req, response& res, const routing_params& params) override
+        {
+            if (!custom_templates_base.empty())
+                mustache::set_base(custom_templates_base);
+            else if (mustache::detail::get_template_base_directory_ref() != mustache::detail::get_global_template_base_directory_ref())
+                mustache::set_base(mustache::detail::get_global_template_base_directory_ref());
+
+            detail::routing_handler_call_helper::call<
+              detail::routing_handler_call_helper::call_params<decltype(handler_)>,
+              0, 0, 0, 0,
+              black_magic::S<Args...>,
+              black_magic::S<>>()(
+              detail::routing_handler_call_helper::call_params<decltype(handler_)>{handler_, params, req, res});
+        }
+
+    private:
+        std::function<void(crow::request&, crow::response&, Args...)> handler_;
+    };
+
+    const int RULE_SPECIAL_REDIRECT_SLASH = 1;
+
+
+    /// A search tree.
+    class Trie
+    {
+    public:
+        struct Node
+        {
+            uint16_t rule_index{};
+            // Assign the index to the maximum 32 unsigned integer value by default so that any other number (specifically 0) is a valid BP id.
+            uint16_t blueprint_index{INVALID_BP_ID};
+            std::string key;
+            ParamType param = ParamType::MAX; // MAX = No param.
+            std::vector<Node> children;
+
+            bool IsSimpleNode() const
+            {
+                return !rule_index &&
+                       blueprint_index == INVALID_BP_ID &&
+                       children.size() < 2 &&
+                       param == ParamType::MAX &&
+                       std::all_of(std::begin(children), std::end(children), [](const Node& x) {
+                           return x.param == ParamType::MAX;
+                       });
+            }
+
+            Node& add_child_node()
+            {
+                children.emplace_back();
+                return children.back();
+            }
+        };
+
+
+        Trie()
+        {}
+
+        /// Check whether or not the trie is empty.
+        bool is_empty()
+        {
+            return head_.children.empty();
+        }
+
+        void optimize()
+        {
+            for (auto& child : head_.children)
+            {
+                optimizeNode(child);
+            }
+        }
+
+
+    private:
+        void optimizeNode(Node& node)
+        {
+            if (node.children.empty())
+                return;
+            if (node.IsSimpleNode())
+            {
+                auto children_temp = std::move(node.children);
+                auto& child_temp = children_temp[0];
+                node.key += child_temp.key;
+                node.rule_index = child_temp.rule_index;
+                node.blueprint_index = child_temp.blueprint_index;
+                node.children = std::move(child_temp.children);
+                optimizeNode(node);
+            }
+            else
+            {
+                for (auto& child : node.children)
+                {
+                    optimizeNode(child);
+                }
+            }
+        }
+
+        void debug_node_print(const Node& node, int level)
+        {
+            if (node.param != ParamType::MAX)
+            {
+                switch (node.param)
+                {
+                    case ParamType::INT:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<int>";
+                        break;
+                    case ParamType::UINT:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<uint>";
+                        break;
+                    case ParamType::DOUBLE:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<double>";
+                        break;
+                    case ParamType::STRING:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<string>";
+                        break;
+                    case ParamType::PATH:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<path>";
+                        break;
+                    default:
+                        CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ "
+                                       << "<ERROR>";
+                        break;
+                }
+            }
+            else
+                CROW_LOG_DEBUG << std::string(3 * level, ' ') << "└➝ " << node.key;
+
+            for (const auto& child : node.children)
+            {
+                debug_node_print(child, level + 1);
+            }
+        }
+
+    public:
+        void debug_print()
+        {
+            CROW_LOG_DEBUG << "└➙ ROOT";
+            for (const auto& child : head_.children)
+                debug_node_print(child, 1);
+        }
+
+        void validate()
+        {
+            if (!head_.IsSimpleNode())
+                throw std::runtime_error("Internal error: Trie header should be simple!");
+            optimize();
+        }
+
+        //Rule_index, Blueprint_index, routing_params
+        routing_handle_result find(const std::string& req_url, const Node& node, unsigned pos = 0, routing_params* params = nullptr, std::vector<uint16_t>* blueprints = nullptr) const
+        {
+            //start params as an empty struct
+            routing_params empty;
+            if (params == nullptr)
+                params = &empty;
+            //same for blueprint vector
+            std::vector<uint16_t> MT;
+            if (blueprints == nullptr)
+                blueprints = &MT;
+
+            uint16_t found{};               //The rule index to be found
+            std::vector<uint16_t> found_BP; //The Blueprint indices to be found
+            routing_params match_params;    //supposedly the final matched parameters
+
+            auto update_found = [&found, &found_BP, &match_params](routing_handle_result& ret) {
+                found_BP = std::move(ret.blueprint_indices);
+                if (ret.rule_index && (!found || found > ret.rule_index))
+                {
+                    found = ret.rule_index;
+                    match_params = std::move(ret.r_params);
+                }
+            };
+
+            //if the function was called on a node at the end of the string (the last recursion), return the nodes rule index, and whatever params were passed to the function
+            if (pos == req_url.size())
+            {
+                found_BP = std::move(*blueprints);
+                return routing_handle_result{node.rule_index, *blueprints, *params};
+            }
+
+            bool found_fragment = false;
+
+            for (const auto& child : node.children)
+            {
+                if (child.param != ParamType::MAX)
+                {
+                    if (child.param == ParamType::INT)
+                    {
+                        char c = req_url[pos];
+                        if ((c >= '0' && c <= '9') || c == '+' || c == '-')
+                        {
+                            char* eptr;
+                            errno = 0;
+                            long long int value = strtoll(req_url.data() + pos, &eptr, 10);
+                            if (errno != ERANGE && eptr != req_url.data() + pos)
+                            {
+                                found_fragment = true;
+                                params->int_params.push_back(value);
+                                if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                                auto ret = find(req_url, child, eptr - req_url.data(), params, blueprints);
+                                update_found(ret);
+                                params->int_params.pop_back();
+                                if (!blueprints->empty()) blueprints->pop_back();
+                            }
+                        }
+                    }
+
+                    else if (child.param == ParamType::UINT)
+                    {
+                        char c = req_url[pos];
+                        if ((c >= '0' && c <= '9') || c == '+')
+                        {
+                            char* eptr;
+                            errno = 0;
+                            unsigned long long int value = strtoull(req_url.data() + pos, &eptr, 10);
+                            if (errno != ERANGE && eptr != req_url.data() + pos)
+                            {
+                                found_fragment = true;
+                                params->uint_params.push_back(value);
+                                if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                                auto ret = find(req_url, child, eptr - req_url.data(), params, blueprints);
+                                update_found(ret);
+                                params->uint_params.pop_back();
+                                if (!blueprints->empty()) blueprints->pop_back();
+                            }
+                        }
+                    }
+
+                    else if (child.param == ParamType::DOUBLE)
+                    {
+                        char c = req_url[pos];
+                        if ((c >= '0' && c <= '9') || c == '+' || c == '-' || c == '.')
+                        {
+                            char* eptr;
+                            errno = 0;
+                            double value = strtod(req_url.data() + pos, &eptr);
+                            if (errno != ERANGE && eptr != req_url.data() + pos)
+                            {
+                                found_fragment = true;
+                                params->double_params.push_back(value);
+                                if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                                auto ret = find(req_url, child, eptr - req_url.data(), params, blueprints);
+                                update_found(ret);
+                                params->double_params.pop_back();
+                                if (!blueprints->empty()) blueprints->pop_back();
+                            }
+                        }
+                    }
+
+                    else if (child.param == ParamType::STRING)
+                    {
+                        size_t epos = pos;
+                        for (; epos < req_url.size(); epos++)
+                        {
+                            if (req_url[epos] == '/')
+                                break;
+                        }
+
+                        if (epos != pos)
+                        {
+                            found_fragment = true;
+                            params->string_params.push_back(req_url.substr(pos, epos - pos));
+                            if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                            auto ret = find(req_url, child, epos, params, blueprints);
+                            update_found(ret);
+                            params->string_params.pop_back();
+                            if (!blueprints->empty()) blueprints->pop_back();
+                        }
+                    }
+
+                    else if (child.param == ParamType::PATH)
+                    {
+                        size_t epos = req_url.size();
+
+                        if (epos != pos)
+                        {
+                            found_fragment = true;
+                            params->string_params.push_back(req_url.substr(pos, epos - pos));
+                            if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                            auto ret = find(req_url, child, epos, params, blueprints);
+                            update_found(ret);
+                            params->string_params.pop_back();
+                            if (!blueprints->empty()) blueprints->pop_back();
+                        }
+                    }
+                }
+
+                else
+                {
+                    const std::string& fragment = child.key;
+                    if (req_url.compare(pos, fragment.size(), fragment) == 0)
+                    {
+                        found_fragment = true;
+                        if (child.blueprint_index != INVALID_BP_ID) blueprints->push_back(child.blueprint_index);
+                        auto ret = find(req_url, child, pos + fragment.size(), params, blueprints);
+                        update_found(ret);
+                        if (!blueprints->empty()) blueprints->pop_back();
+                    }
+                }
+            }
+
+            if (!found_fragment)
+                found_BP = std::move(*blueprints);
+
+            return routing_handle_result{found, found_BP, match_params}; //Called after all the recursions have been done
+        }
+
+        routing_handle_result find(const std::string& req_url) const
+        {
+            return find(req_url, head_);
+        }
+
+        //This functions assumes any blueprint info passed is valid
+        void add(const std::string& url, uint16_t rule_index, unsigned bp_prefix_length = 0, uint16_t blueprint_index = INVALID_BP_ID)
+        {
+            auto idx = &head_;
+
+            bool has_blueprint = bp_prefix_length != 0 && blueprint_index != INVALID_BP_ID;
+
+            for (unsigned i = 0; i < url.size(); i++)
+            {
+                char c = url[i];
+                if (c == '<')
+                {
+                    static struct ParamTraits
+                    {
+                        ParamType type;
+                        std::string name;
+                    } paramTraits[] =
+                      {
+                        {ParamType::INT, "<int>"},
+                        {ParamType::UINT, "<uint>"},
+                        {ParamType::DOUBLE, "<float>"},
+                        {ParamType::DOUBLE, "<double>"},
+                        {ParamType::STRING, "<str>"},
+                        {ParamType::STRING, "<string>"},
+                        {ParamType::PATH, "<path>"},
+                      };
+
+                    for (const auto& x : paramTraits)
+                    {
+                        if (url.compare(i, x.name.size(), x.name) == 0)
+                        {
+                            bool found = false;
+                            for (auto& child : idx->children)
+                            {
+                                if (child.param == x.type)
+                                {
+                                    idx = &child;
+                                    i += x.name.size();
+                                    found = true;
+                                    break;
+                                }
+                            }
+                            if (found)
+                                break;
+
+                            auto new_node_idx = &idx->add_child_node();
+                            new_node_idx->param = x.type;
+                            idx = new_node_idx;
+                            i += x.name.size();
+                            break;
+                        }
+                    }
+
+                    i--;
+                }
+                else
+                {
+                    //This part assumes the tree is unoptimized (every node has a max 1 character key)
+                    bool piece_found = false;
+                    for (auto& child : idx->children)
+                    {
+                        if (child.key[0] == c)
+                        {
+                            idx = &child;
+                            piece_found = true;
+                            break;
+                        }
+                    }
+                    if (!piece_found)
+                    {
+                        auto new_node_idx = &idx->add_child_node();
+                        new_node_idx->key = c;
+                        //The assumption here is that you'd only need to add a blueprint index if the tree didn't have the BP prefix.
+                        if (has_blueprint && i == bp_prefix_length)
+                            new_node_idx->blueprint_index = blueprint_index;
+                        idx = new_node_idx;
+                    }
+                }
+            }
+
+            //check if the last node already has a value (exact url already in Trie)
+            if (idx->rule_index)
+                throw std::runtime_error("handler already exists for " + url);
+            idx->rule_index = rule_index;
+        }
+
+    private:
+        Node head_;
+    };
+
+    /// A blueprint can be considered a smaller section of a Crow app, specifically where the router is conecerned.
+
+    ///
+    /// You can use blueprints to assign a common prefix to rules' prefix, set custom static and template folders, and set a custom catchall route.
+    /// You can also assign nest blueprints for maximum Compartmentalization.
+    class Blueprint
+    {
+    public:
+        Blueprint(const std::string& prefix):
+          prefix_(prefix){};
+
+        Blueprint(const std::string& prefix, const std::string& static_dir):
+          prefix_(prefix), static_dir_(static_dir){};
+
+        Blueprint(const std::string& prefix, const std::string& static_dir, const std::string& templates_dir):
+          prefix_(prefix), static_dir_(static_dir), templates_dir_(templates_dir){};
+
+        /*
+        Blueprint(Blueprint& other)
+        {
+            prefix_ = std::move(other.prefix_);
+            all_rules_ = std::move(other.all_rules_);
+        }
+
+        Blueprint(const Blueprint& other)
+        {
+            prefix_ = other.prefix_;
+            all_rules_ = other.all_rules_;
+        }
+*/
+        Blueprint(Blueprint&& value)
+        {
+            *this = std::move(value);
+        }
+
+        Blueprint& operator=(const Blueprint& value) = delete;
+
+        Blueprint& operator=(Blueprint&& value) noexcept
+        {
+            prefix_ = std::move(value.prefix_);
+            static_dir_ = std::move(value.static_dir_);
+            templates_dir_ = std::move(value.templates_dir_);
+            all_rules_ = std::move(value.all_rules_);
+            catchall_rule_ = std::move(value.catchall_rule_);
+            blueprints_ = std::move(value.blueprints_);
+            mw_indices_ = std::move(value.mw_indices_);
+            return *this;
+        }
+
+        bool operator==(const Blueprint& value)
+        {
+            return value.prefix() == prefix_;
+        }
+
+        bool operator!=(const Blueprint& value)
+        {
+            return value.prefix() != prefix_;
+        }
+
+        std::string prefix() const
+        {
+            return prefix_;
+        }
+
+        std::string static_dir() const
+        {
+            return static_dir_;
+        }
+
+        void set_added() {
+            added_ = true;
+        }
+
+        bool is_added() {
+            return added_;
+        }
+
+        DynamicRule& new_rule_dynamic(const std::string& rule)
+        {
+            std::string new_rule = '/' + prefix_ + rule;
+            auto ruleObject = new DynamicRule(std::move(new_rule));
+            ruleObject->custom_templates_base = templates_dir_;
+            all_rules_.emplace_back(ruleObject);
+
+            return *ruleObject;
+        }
+
+        template<uint64_t N>
+        typename black_magic::arguments<N>::type::template rebind<TaggedRule>& new_rule_tagged(const std::string& rule)
+        {
+            std::string new_rule = '/' + prefix_ + rule;
+            using RuleT = typename black_magic::arguments<N>::type::template rebind<TaggedRule>;
+
+            auto ruleObject = new RuleT(std::move(new_rule));
+            ruleObject->custom_templates_base = templates_dir_;
+            all_rules_.emplace_back(ruleObject);
+
+            return *ruleObject;
+        }
+
+        void register_blueprint(Blueprint& blueprint)
+        {
+            if (blueprints_.empty() || std::find(blueprints_.begin(), blueprints_.end(), &blueprint) == blueprints_.end())
+            {
+                apply_blueprint(blueprint);
+                blueprints_.emplace_back(&blueprint);
+            }
+            else
+                throw std::runtime_error("blueprint \"" + blueprint.prefix_ + "\" already exists in blueprint \"" + prefix_ + '\"');
+        }
+
+
+        CatchallRule& catchall_rule()
+        {
+            return catchall_rule_;
+        }
+
+        template<typename App, typename... Middlewares>
+        void middlewares()
+        {
+            mw_indices_.push<App, Middlewares...>();
+        }
+
+    private:
+        void apply_blueprint(Blueprint& blueprint)
+        {
+
+            blueprint.prefix_ = prefix_ + '/' + blueprint.prefix_;
+            blueprint.static_dir_ = static_dir_ + '/' + blueprint.static_dir_;
+            blueprint.templates_dir_ = templates_dir_ + '/' + blueprint.templates_dir_;
+            for (auto& rule : blueprint.all_rules_)
+            {
+                std::string new_rule = '/' + prefix_ + rule->rule_;
+                rule->rule_ = new_rule;
+            }
+            for (Blueprint* bp_child : blueprint.blueprints_)
+            {
+                Blueprint& bp_ref = *bp_child;
+                apply_blueprint(bp_ref);
+            }
+        }
+
+        std::string prefix_;
+        std::string static_dir_;
+        std::string templates_dir_;
+        std::vector<std::unique_ptr<BaseRule>> all_rules_;
+        CatchallRule catchall_rule_;
+        std::vector<Blueprint*> blueprints_;
+        detail::middleware_indices mw_indices_;
+        bool added_{false};
+
+        friend class Router;
+    };
+
+    /// Handles matching requests to existing rules and upgrade requests.
+    class Router
+    {
+    public:
+        Router()
+        {}
+
+        DynamicRule& new_rule_dynamic(const std::string& rule)
+        {
+            auto ruleObject = new DynamicRule(rule);
+            all_rules_.emplace_back(ruleObject);
+
+            return *ruleObject;
+        }
+
+        template<uint64_t N>
+        typename black_magic::arguments<N>::type::template rebind<TaggedRule>& new_rule_tagged(const std::string& rule)
+        {
+            using RuleT = typename black_magic::arguments<N>::type::template rebind<TaggedRule>;
+
+            auto ruleObject = new RuleT(rule);
+            all_rules_.emplace_back(ruleObject);
+
+            return *ruleObject;
+        }
+
+        CatchallRule& catchall_rule()
+        {
+            return catchall_rule_;
+        }
+
+        void internal_add_rule_object(const std::string& rule, BaseRule* ruleObject)
+        {
+            internal_add_rule_object(rule, ruleObject, INVALID_BP_ID, blueprints_);
+        }
+
+        void internal_add_rule_object(const std::string& rule, BaseRule* ruleObject, const uint16_t& BP_index, std::vector<Blueprint*>& blueprints)
+        {
+            bool has_trailing_slash = false;
+            std::string rule_without_trailing_slash;
+            if (rule.size() > 1 && rule.back() == '/')
+            {
+                has_trailing_slash = true;
+                rule_without_trailing_slash = rule;
+                rule_without_trailing_slash.pop_back();
+            }
+
+            ruleObject->mw_indices_.pack();
+
+            ruleObject->foreach_method([&](int method) {
+                per_methods_[method].rules.emplace_back(ruleObject);
+                per_methods_[method].trie.add(rule, per_methods_[method].rules.size() - 1, BP_index != INVALID_BP_ID ? blueprints[BP_index]->prefix().length() : 0, BP_index);
+
+                // directory case:
+                //   request to '/about' url matches '/about/' rule
+                if (has_trailing_slash)
+                {
+                    per_methods_[method].trie.add(rule_without_trailing_slash, RULE_SPECIAL_REDIRECT_SLASH, BP_index != INVALID_BP_ID ? blueprints[BP_index]->prefix().length() : 0, BP_index);
+                }
+            });
+
+            ruleObject->set_added();
+        }
+
+        void register_blueprint(Blueprint& blueprint)
+        {
+            if (std::find(blueprints_.begin(), blueprints_.end(), &blueprint) == blueprints_.end())
+            {
+                blueprints_.emplace_back(&blueprint);
+            }
+            else
+                throw std::runtime_error("blueprint \"" + blueprint.prefix_ + "\" already exists in router");
+        }
+
+        void get_recursive_child_methods(Blueprint* blueprint, std::vector<HTTPMethod>& methods)
+        {
+            //we only need to deal with children if the blueprint has absolutely no methods (meaning its index won't be added to the trie)
+            if (blueprint->static_dir_.empty() && blueprint->all_rules_.empty())
+            {
+                for (Blueprint* bp : blueprint->blueprints_)
+                {
+                    get_recursive_child_methods(bp, methods);
+                }
+            }
+            else if (!blueprint->static_dir_.empty())
+                methods.emplace_back(HTTPMethod::Get);
+            for (auto& rule : blueprint->all_rules_)
+            {
+                rule->foreach_method([&methods](unsigned method) {
+                    HTTPMethod method_final = static_cast<HTTPMethod>(method);
+                    if (std::find(methods.begin(), methods.end(), method_final) == methods.end())
+                        methods.emplace_back(method_final);
+                });
+            }
+        }
+
+        void validate_bp() {
+            //Take all the routes from the registered blueprints and add them to `all_rules_` to be processed.
+            detail::middleware_indices blueprint_mw;
+            validate_bp(blueprints_, blueprint_mw);
+        }
+
+        void validate_bp(std::vector<Blueprint*> blueprints, detail::middleware_indices& current_mw)
+        {
+            for (unsigned i = 0; i < blueprints.size(); i++)
+            {
+                Blueprint* blueprint = blueprints[i];
+
+                if (blueprint->is_added()) continue;
+
+                if (blueprint->static_dir_ == "" && blueprint->all_rules_.empty())
+                {
+                    std::vector<HTTPMethod> methods;
+                    get_recursive_child_methods(blueprint, methods);
+                    for (HTTPMethod x : methods)
+                    {
+                        int i = static_cast<int>(x);
+                        per_methods_[i].trie.add(blueprint->prefix(), 0, blueprint->prefix().length(), i);
+                    }
+                }
+
+                current_mw.merge_back(blueprint->mw_indices_);
+                for (auto& rule : blueprint->all_rules_)
+                {
+                    if (rule && !rule->is_added())
+                    {
+                        auto upgraded = rule->upgrade();
+                        if (upgraded)
+                            rule = std::move(upgraded);
+                        rule->validate();
+                        rule->mw_indices_.merge_front(current_mw);
+                        internal_add_rule_object(rule->rule(), rule.get(), i, blueprints);
+                    }
+                }
+                validate_bp(blueprint->blueprints_, current_mw);
+                current_mw.pop_back(blueprint->mw_indices_);
+                blueprint->set_added();
+            }
+        }
+
+        void validate()
+        {
+            for (auto& rule : all_rules_)
+            {
+                if (rule && !rule->is_added())
+                {
+                    auto upgraded = rule->upgrade();
+                    if (upgraded)
+                        rule = std::move(upgraded);
+                    rule->validate();
+                    internal_add_rule_object(rule->rule(), rule.get());
+                }
+            }
+            for (auto& per_method : per_methods_)
+            {
+                per_method.trie.validate();
+            }
+        }
+
+        // TODO maybe add actual_method
+        template<typename Adaptor>
+        void handle_upgrade(const request& req, response& res, Adaptor&& adaptor)
+        {
+            if (req.method >= HTTPMethod::InternalMethodCount)
+                return;
+
+            auto& per_method = per_methods_[static_cast<int>(req.method)];
+            auto& rules = per_method.rules;
+            unsigned rule_index = per_method.trie.find(req.url).rule_index;
+
+            if (!rule_index)
+            {
+                for (auto& per_method : per_methods_)
+                {
+                    if (per_method.trie.find(req.url).rule_index)
+                    {
+                        CROW_LOG_DEBUG << "Cannot match method " << req.url << " " << method_name(req.method);
+                        res = response(405);
+                        res.end();
+                        return;
+                    }
+                }
+
+                CROW_LOG_INFO << "Cannot match rules " << req.url;
+                res = response(404);
+                res.end();
+                return;
+            }
+
+            if (rule_index >= rules.size())
+                throw std::runtime_error("Trie internal structure corrupted!");
+
+            if (rule_index == RULE_SPECIAL_REDIRECT_SLASH)
+            {
+                CROW_LOG_INFO << "Redirecting to a url with trailing slash: " << req.url;
+                res = response(301);
+
+                // TODO(ipkn) absolute url building
+                if (req.get_header_value("Host").empty())
+                {
+                    res.add_header("Location", req.url + "/");
+                }
+                else
+                {
+                    res.add_header("Location", "http://" + req.get_header_value("Host") + req.url + "/");
+                }
+                res.end();
+                return;
+            }
+
+            CROW_LOG_DEBUG << "Matched rule (upgrade) '" << rules[rule_index]->rule_ << "' " << static_cast<uint32_t>(req.method) << " / " << rules[rule_index]->get_methods();
+
+            try
+            {
+                rules[rule_index]->handle_upgrade(req, res, std::move(adaptor));
+            }
+            catch (...)
+            {
+                exception_handler_(res);
+                res.end();
+                return;
+            }
+        }
+
+        void get_found_bp(std::vector<uint16_t>& bp_i, std::vector<Blueprint*>& blueprints, std::vector<Blueprint*>& found_bps, uint16_t index = 0)
+        {
+            // This statement makes 3 assertions:
+            // 1. The index is above 0.
+            // 2. The index does not lie outside the given blueprint list.
+            // 3. The next blueprint we're adding has a prefix that starts the same as the already added blueprint + a slash (the rest is irrelevant).
+            //
+            // This is done to prevent a blueprint that has a prefix of "bp_prefix2" to be assumed as a child of one that has "bp_prefix".
+            //
+            // If any of the assertions is untrue, we delete the last item added, and continue using the blueprint list of the blueprint found before, the topmost being the router's list
+            auto verify_prefix = [&bp_i, &index, &blueprints, &found_bps]() {
+                return index > 0 &&
+                       bp_i[index] < blueprints.size() &&
+                       blueprints[bp_i[index]]->prefix().substr(0, found_bps[index - 1]->prefix().length() + 1).compare(std::string(found_bps[index - 1]->prefix() + '/')) == 0;
+            };
+            if (index < bp_i.size())
+            {
+
+                if (verify_prefix())
+                {
+                    found_bps.push_back(blueprints[bp_i[index]]);
+                    get_found_bp(bp_i, found_bps.back()->blueprints_, found_bps, ++index);
+                }
+                else
+                {
+                    if (found_bps.size() < 2)
+                    {
+                        found_bps.clear();
+                        found_bps.push_back(blueprints_[bp_i[index]]);
+                    }
+                    else
+                    {
+                        found_bps.pop_back();
+                        Blueprint* last_element = found_bps.back();
+                        found_bps.push_back(last_element->blueprints_[bp_i[index]]);
+                    }
+                    get_found_bp(bp_i, found_bps.back()->blueprints_, found_bps, ++index);
+                }
+            }
+        }
+
+        /// Is used to handle errors, you insert the error code, found route, request, and response. and it'll either call the appropriate catchall route (considering the blueprint system) and send you a status string (which is mainly used for debug messages), or just set the response code to the proper error code.
+        std::string get_error(unsigned short code, routing_handle_result& found, const request& req, response& res)
+        {
+            res.code = code;
+            std::vector<Blueprint*> bps_found;
+            get_found_bp(found.blueprint_indices, blueprints_, bps_found);
+            for (int i = bps_found.size() - 1; i > 0; i--)
+            {
+                std::vector<uint16_t> bpi = found.blueprint_indices;
+                if (bps_found[i]->catchall_rule().has_handler())
+                {
+                    try
+                    {
+                        bps_found[i]->catchall_rule().handler_(req, res);
+                    }
+                    catch (...)
+                    {
+                        exception_handler_(res);
+                    }
+#ifdef CROW_ENABLE_DEBUG
+                    return std::string("Redirected to Blueprint \"" + bps_found[i]->prefix() + "\" Catchall rule");
+#else
+                    return std::string();
+#endif
+                }
+            }
+            if (catchall_rule_.has_handler())
+            {
+                try
+                {
+                    catchall_rule_.handler_(req, res);
+                }
+                catch (...)
+                {
+                    exception_handler_(res);
+                }
+#ifdef CROW_ENABLE_DEBUG
+                return std::string("Redirected to global Catchall rule");
+#else
+                return std::string();
+#endif
+            }
+            return std::string();
+        }
+
+        std::unique_ptr<routing_handle_result> handle_initial(request& req, response& res)
+        {
+            HTTPMethod method_actual = req.method;
+
+            std::unique_ptr<routing_handle_result> found{
+              new routing_handle_result(
+                0,
+                std::vector<uint16_t>(),
+                routing_params(),
+                HTTPMethod::InternalMethodCount)}; // This is always returned to avoid a null pointer dereference.
+
+            // NOTE(EDev): This most likely will never run since the parser should handle this situation and close the connection before it gets here.
+            if (CROW_UNLIKELY(req.method >= HTTPMethod::InternalMethodCount))
+                return found;
+            else if (req.method == HTTPMethod::Head)
+            {
+                *found = per_methods_[static_cast<int>(method_actual)].trie.find(req.url);
+                // support HEAD requests using GET if not defined as method for the requested URL
+                if (!found->rule_index)
+                {
+                    method_actual = HTTPMethod::Get;
+                    *found = per_methods_[static_cast<int>(method_actual)].trie.find(req.url);
+                    if (!found->rule_index) // If a route is still not found, return a 404 without executing the rest of the HEAD specific code.
+                    {
+                        CROW_LOG_DEBUG << "Cannot match rules " << req.url;
+                        res = response(404); //TODO(EDev): Should this redirect to catchall?
+                        res.end();
+                        return found;
+                    }
+                }
+
+                res.skip_body = true;
+                found->method = method_actual;
+                return found;
+            }
+            else if (req.method == HTTPMethod::Options)
+            {
+                std::string allow = "OPTIONS, HEAD, ";
+
+                if (req.url == "/*")
+                {
+                    for (int i = 0; i < static_cast<int>(HTTPMethod::InternalMethodCount); i++)
+                    {
+                        if (static_cast<int>(HTTPMethod::Head) == i)
+                            continue; // HEAD is always allowed
+
+                        if (!per_methods_[i].trie.is_empty())
+                        {
+                            allow += method_name(static_cast<HTTPMethod>(i)) + ", ";
+                        }
+                    }
+                    allow = allow.substr(0, allow.size() - 2);
+                    res = response(204);
+                    res.set_header("Allow", allow);
+                    res.end();
+                    found->method = method_actual;
+                    return found;
+                }
+                else
+                {
+                    bool rules_matched = false;
+                    for (int i = 0; i < static_cast<int>(HTTPMethod::InternalMethodCount); i++)
+                    {
+                        if (per_methods_[i].trie.find(req.url).rule_index)
+                        {
+                            rules_matched = true;
+
+                            if (static_cast<int>(HTTPMethod::Head) == i)
+                                continue; // HEAD is always allowed
+
+                            allow += method_name(static_cast<HTTPMethod>(i)) + ", ";
+                        }
+                    }
+                    if (rules_matched)
+                    {
+                        allow = allow.substr(0, allow.size() - 2);
+                        res = response(204);
+                        res.set_header("Allow", allow);
+                        res.end();
+                        found->method = method_actual;
+                        return found;
+                    }
+                    else
+                    {
+                        CROW_LOG_DEBUG << "Cannot match rules " << req.url;
+                        res = response(404); //TODO(EDev): Should this redirect to catchall?
+                        res.end();
+                        return found;
+                    }
+                }
+            }
+            else // Every request that isn't a HEAD or OPTIONS request
+            {
+                *found = per_methods_[static_cast<int>(method_actual)].trie.find(req.url);
+                // TODO(EDev): maybe ending the else here would allow the requests coming from above (after removing the return statement) to be checked on whether they actually point to a route
+                if (!found->rule_index)
+                {
+                    for (auto& per_method : per_methods_)
+                    {
+                        if (per_method.trie.find(req.url).rule_index) //Route found, but in another method
+                        {
+                            const std::string error_message(get_error(405, *found, req, res));
+                            CROW_LOG_DEBUG << "Cannot match method " << req.url << " " << method_name(method_actual) << ". " << error_message;
+                            res.end();
+                            return found;
+                        }
+                    }
+                    //Route does not exist anywhere
+
+                    const std::string error_message(get_error(404, *found, req, res));
+                    CROW_LOG_DEBUG << "Cannot match rules " << req.url << ". " << error_message;
+                    res.end();
+                    return found;
+                }
+
+                found->method = method_actual;
+                return found;
+            }
+        }
+
+        template<typename App>
+        void handle(request& req, response& res, routing_handle_result found)
+        {
+            HTTPMethod method_actual = found.method;
+            auto& rules = per_methods_[static_cast<int>(method_actual)].rules;
+            unsigned rule_index = found.rule_index;
+
+            if (rule_index >= rules.size())
+                throw std::runtime_error("Trie internal structure corrupted!");
+
+            if (rule_index == RULE_SPECIAL_REDIRECT_SLASH)
+            {
+                CROW_LOG_INFO << "Redirecting to a url with trailing slash: " << req.url;
+                res = response(301);
+
+                // TODO(ipkn) absolute url building
+                if (req.get_header_value("Host").empty())
+                {
+                    res.add_header("Location", req.url + "/");
+                }
+                else
+                {
+                    res.add_header("Location", "http://" + req.get_header_value("Host") + req.url + "/");
+                }
+                res.end();
+                return;
+            }
+
+            CROW_LOG_DEBUG << "Matched rule '" << rules[rule_index]->rule_ << "' " << static_cast<uint32_t>(req.method) << " / " << rules[rule_index]->get_methods();
+
+            try
+            {
+                auto& rule = rules[rule_index];
+                handle_rule<App>(rule, req, res, found.r_params);
+            }
+            catch (...)
+            {
+                exception_handler_(res);
+                res.end();
+                return;
+            }
+        }
+
+        template<typename App>
+        typename std::enable_if<std::tuple_size<typename App::mw_container_t>::value != 0, void>::type
+          handle_rule(BaseRule* rule, crow::request& req, crow::response& res, const crow::routing_params& rp)
+        {
+            if (!rule->mw_indices_.empty())
+            {
+                auto& ctx = *reinterpret_cast<typename App::context_t*>(req.middleware_context);
+                auto& container = *reinterpret_cast<typename App::mw_container_t*>(req.middleware_container);
+                detail::middleware_call_criteria_dynamic<false> crit_fwd(rule->mw_indices_.indices());
+
+                auto glob_completion_handler = std::move(res.complete_request_handler_);
+                res.complete_request_handler_ = [] {};
+
+                detail::middleware_call_helper<decltype(crit_fwd),
+                                               0, typename App::context_t, typename App::mw_container_t>(crit_fwd, container, req, res, ctx);
+
+                if (res.completed_)
+                {
+                    glob_completion_handler();
+                    return;
+                }
+
+                res.complete_request_handler_ = [&rule, &ctx, &container, &req, &res, glob_completion_handler] {
+                    detail::middleware_call_criteria_dynamic<true> crit_bwd(rule->mw_indices_.indices());
+
+                    detail::after_handlers_call_helper<
+                      decltype(crit_bwd),
+                      std::tuple_size<typename App::mw_container_t>::value - 1,
+                      typename App::context_t,
+                      typename App::mw_container_t>(crit_bwd, container, ctx, req, res);
+                    glob_completion_handler();
+                };
+            }
+            rule->handle(req, res, rp);
+        }
+
+        template<typename App>
+        typename std::enable_if<std::tuple_size<typename App::mw_container_t>::value == 0, void>::type
+          handle_rule(BaseRule* rule, crow::request& req, crow::response& res, const crow::routing_params& rp)
+        {
+            rule->handle(req, res, rp);
+        }
+
+        void debug_print()
+        {
+            for (int i = 0; i < static_cast<int>(HTTPMethod::InternalMethodCount); i++)
+            {
+                Trie& trie_ = per_methods_[i].trie;
+                if (!trie_.is_empty())
+                {
+                    CROW_LOG_DEBUG << method_name(static_cast<HTTPMethod>(i));
+                    trie_.debug_print();
+                }
+            }
+        }
+
+        std::vector<Blueprint*>& blueprints()
+        {
+            return blueprints_;
+        }
+
+        std::function<void(crow::response&)>& exception_handler()
+        {
+            return exception_handler_;
+        }
+
+        static void default_exception_handler(response& res)
+        {
+            // any uncaught exceptions become 500s
+            res = response(500);
+
+            try
+            {
+                throw;
+            }
+            catch (const std::exception& e)
+            {
+                CROW_LOG_ERROR << "An uncaught exception occurred: " << e.what();
+            }
+            catch (...)
+            {
+                CROW_LOG_ERROR << "An uncaught exception occurred. The type was unknown so no information was available.";
+            }
+        }
+
+    private:
+        CatchallRule catchall_rule_;
+
+        struct PerMethod
+        {
+            std::vector<BaseRule*> rules;
+            Trie trie;
+
+            // rule index 0, 1 has special meaning; preallocate it to avoid duplication.
+            PerMethod():
+              rules(2) {}
+        };
+        std::array<PerMethod, static_cast<int>(HTTPMethod::InternalMethodCount)> per_methods_;
+        std::vector<std::unique_ptr<BaseRule>> all_rules_;
+        std::vector<Blueprint*> blueprints_;
+        std::function<void(crow::response&)> exception_handler_ = &default_exception_handler;
+    };
+} // namespace crow
+
+
+namespace crow
+{
+    struct CORSHandler;
+
+    /// Used for tuning CORS policies
+    struct CORSRules
+    {
+        friend struct crow::CORSHandler;
+
+        /// Set Access-Control-Allow-Origin. Default is "*"
+        CORSRules& origin(const std::string& origin)
+        {
+            origin_ = origin;
+            return *this;
+        }
+
+        /// Set Access-Control-Allow-Methods. Default is "*"
+        CORSRules& methods(crow::HTTPMethod method)
+        {
+            add_list_item(methods_, crow::method_name(method));
+            return *this;
+        }
+
+        /// Set Access-Control-Allow-Methods. Default is "*"
+        template<typename... Methods>
+        CORSRules& methods(crow::HTTPMethod method, Methods... method_list)
+        {
+            add_list_item(methods_, crow::method_name(method));
+            methods(method_list...);
+            return *this;
+        }
+
+        /// Set Access-Control-Allow-Headers. Default is "*"
+        CORSRules& headers(const std::string& header)
+        {
+            add_list_item(headers_, header);
+            return *this;
+        }
+
+        /// Set Access-Control-Allow-Headers. Default is "*"
+        template<typename... Headers>
+        CORSRules& headers(const std::string& header, Headers... header_list)
+        {
+            add_list_item(headers_, header);
+            headers(header_list...);
+            return *this;
+        }
+
+        /// Set Access-Control-Max-Age. Default is none
+        CORSRules& max_age(int max_age)
+        {
+            max_age_ = std::to_string(max_age);
+            return *this;
+        }
+
+        /// Enable Access-Control-Allow-Credentials
+        CORSRules& allow_credentials()
+        {
+            allow_credentials_ = true;
+            return *this;
+        }
+
+        /// Ignore CORS and don't send any headers
+        void ignore()
+        {
+            ignore_ = true;
+        }
+
+        /// Handle CORS on specific prefix path
+        CORSRules& prefix(const std::string& prefix);
+
+        /// Handle CORS for specific blueprint
+        CORSRules& blueprint(const Blueprint& bp);
+
+        /// Global CORS policy
+        CORSRules& global();
+
+    private:
+        CORSRules() = delete;
+        CORSRules(CORSHandler* handler):
+          handler_(handler) {}
+
+        /// build comma separated list
+        void add_list_item(std::string& list, const std::string& val)
+        {
+            if (list == "*") list = "";
+            if (list.size() > 0) list += ", ";
+            list += val;
+        }
+
+        /// Set header `key` to `value` if it is not set
+        void set_header_no_override(const std::string& key, const std::string& value, crow::response& res)
+        {
+            if (value.size() == 0) return;
+            if (!get_header_value(res.headers, key).empty()) return;
+            res.add_header(key, value);
+        }
+
+        /// Set response headers
+        void apply(crow::response& res)
+        {
+            if (ignore_) return;
+            set_header_no_override("Access-Control-Allow-Origin", origin_, res);
+            set_header_no_override("Access-Control-Allow-Methods", methods_, res);
+            set_header_no_override("Access-Control-Allow-Headers", headers_, res);
+            set_header_no_override("Access-Control-Max-Age", max_age_, res);
+            if (allow_credentials_) set_header_no_override("Access-Control-Allow-Credentials", "true", res);
+        }
+
+        bool ignore_ = false;
+        // TODO: support multiple origins that are dynamically selected
+        std::string origin_ = "*";
+        std::string methods_ = "*";
+        std::string headers_ = "*";
+        std::string max_age_;
+        bool allow_credentials_ = false;
+
+        CORSHandler* handler_;
+    };
+
+    /// CORSHandler is a global middleware for setting CORS headers.
+
+    ///
+    /// By default, it sets Access-Control-Allow-Origin/Methods/Headers to "*".
+    /// The default behaviour can be changed with the `global()` cors rule.
+    /// Additional rules for prexies can be added with `prefix()`.
+    struct CORSHandler
+    {
+        struct context
+        {};
+
+        void before_handle(crow::request& /*req*/, crow::response& /*res*/, context& /*ctx*/)
+        {}
+
+        void after_handle(crow::request& req, crow::response& res, context& /*ctx*/)
+        {
+            auto& rule = find_rule(req.url);
+            rule.apply(res);
+        }
+
+        /// Handle CORS on a specific prefix path
+        CORSRules& prefix(const std::string& prefix)
+        {
+            rules.emplace_back(prefix, CORSRules(this));
+            return rules.back().second;
+        }
+
+        /// Handle CORS for a specific blueprint
+        CORSRules& blueprint(const Blueprint& bp)
+        {
+            rules.emplace_back(bp.prefix(), CORSRules(this));
+            return rules.back().second;
+        }
+
+        /// Get the global CORS policy
+        CORSRules& global()
+        {
+            return default_;
+        }
+
+    private:
+        CORSRules& find_rule(const std::string& path)
+        {
+            // TODO: use a trie in case of many rules
+            for (auto& rule : rules)
+            {
+                // Check if path starts with a rules prefix
+                if (path.rfind(rule.first, 0) == 0)
+                {
+                    return rule.second;
+                }
+            }
+            return default_;
+        }
+
+        std::vector<std::pair<std::string, CORSRules>> rules;
+        CORSRules default_ = CORSRules(this);
+    };
+
+    inline CORSRules& CORSRules::prefix(const std::string& prefix)
+    {
+        return handler_->prefix(prefix);
+    }
+
+    inline CORSRules& CORSRules::blueprint(const Blueprint& bp)
+    {
+        return handler_->blueprint(bp);
+    }
+
+    inline CORSRules& CORSRules::global()
+    {
+        return handler_->global();
+    }
+
+} // namespace crow
+
+/**
+ * \file crow/app.h
+ * \brief This file includes the definition of the crow::Crow class,
+ * the crow::App and crow::SimpleApp aliases, and some macros.
+ *
+ * In this file are defined:
+ * - crow::Crow
+ * - crow::App
+ * - crow::SimpleApp
+ * - \ref CROW_ROUTE
+ * - \ref CROW_BP_ROUTE
+ * - \ref CROW_WEBSOCKET_ROUTE
+ * - \ref CROW_MIDDLEWARES
+ * - \ref CROW_CATCHALL_ROUTE
+ * - \ref CROW_BP_CATCHALL_ROUTE
+ */
+
+
+#include <chrono>
+#include <string>
+#include <functional>
+#include <memory>
+#include <future>
+#include <cstdint>
+#include <type_traits>
+#include <thread>
+#include <condition_variable>
+
+#ifdef CROW_ENABLE_COMPRESSION
+#endif // #ifdef CROW_ENABLE_COMPRESSION
+
+
+#ifdef CROW_MSVC_WORKAROUND
+
+#define CROW_ROUTE(app, url) app.route_dynamic(url) // See the documentation in the comment below.
+#define CROW_BP_ROUTE(blueprint, url) blueprint.new_rule_dynamic(url) // See the documentation in the comment below.
+
+#else // #ifdef CROW_MSVC_WORKAROUND
+
+/**
+ * \def CROW_ROUTE(app, url)
+ * \brief Creates a route for app using a rule.
+ *
+ * It use crow::Crow::route_dynamic or crow::Crow::route to define
+ * a rule for your application. It's usage is like this:
+ *
+ * ```cpp
+ * auto app = crow::SimpleApp(); // or crow::App()
+ * CROW_ROUTE(app, "/")
+ * ([](){
+ *     return "<h1>Hello, world!</h1>";
+ * });
+ * ```
+ *
+ * This is the recommended way to define routes in a crow application.
+ * \see [Page of guide "Routes"](https://crowcpp.org/master/guides/routes/).
+ */
+#define CROW_ROUTE(app, url) app.template route<crow::black_magic::get_parameter_tag(url)>(url)
+
+/**
+ * \def CROW_BP_ROUTE(blueprint, url)
+ * \brief Creates a route for a blueprint using a rule.
+ *
+ * It may use crow::Blueprint::new_rule_dynamic or
+ * crow::Blueprint::new_rule_tagged to define a new rule for
+ * an given blueprint. It's usage is similar
+ * to CROW_ROUTE macro:
+ *
+ * ```cpp
+ * crow::Blueprint my_bp();
+ * CROW_BP_ROUTE(my_bp, "/")
+ * ([](){
+ *     return "<h1>Hello, world!</h1>";
+ * });
+ * ```
+ *
+ * This is the recommended way to define routes in a crow blueprint
+ * because of its compile-time capabilities.
+ *
+ * \see [Page of the guide "Blueprints"](https://crowcpp.org/master/guides/blueprints/).
+ */
+#define CROW_BP_ROUTE(blueprint, url) blueprint.new_rule_tagged<crow::black_magic::get_parameter_tag(url)>(url)
+
+/**
+ * \def CROW_WEBSOCKET_ROUTE(app, url)
+ * \brief Defines WebSocket route for app.
+ *
+ * It binds a WebSocket route to app. Easy solution to implement
+ * WebSockets in your app. The usage syntax of this macro is
+ * like this:
+ *
+ * ```cpp
+ * auto app = crow::SimpleApp(); // or crow::App()
+ * CROW_WEBSOCKET_ROUTE(app, "/ws")
+ *     .onopen([&](crow::websocket::connection& conn){
+ *                do_something();
+ *            })
+ *     .onclose([&](crow::websocket::connection& conn, const std::string& reason){
+ *                 do_something();
+ *             })
+ *     .onmessage([&](crow::websocket::connection&, const std::string& data, bool is_binary){
+ *                   if (is_binary)
+ *                       do_something(data);
+ *                   else
+ *                       do_something_else(data);
+ *               });
+ * ```
+ *
+ * \see [Page of the guide "WebSockets"](https://crowcpp.org/master/guides/websockets/).
+ */
+#define CROW_WEBSOCKET_ROUTE(app, url) app.route<crow::black_magic::get_parameter_tag(url)>(url).websocket<std::remove_reference<decltype(app)>::type>(&app)
+
+/**
+ * \def CROW_MIDDLEWARES(app, ...)
+ * \brief Enable a Middleware for an specific route in app
+ * or blueprint.
+ *
+ * It defines the usage of a Middleware in one route. And it
+ * can be used in both crow::SimpleApp (and crow::App) instances and
+ * crow::Blueprint. Its usage syntax is like this:
+ *
+ * ```cpp
+ * auto app = crow::SimpleApp(); // or crow::App()
+ * CROW_ROUTE(app, "/with_middleware")
+ * .CROW_MIDDLEWARES(app, LocalMiddleware) // Can be used more than one
+ * ([]() {                                 // middleware.
+ *     return "Hello world!";
+ * });
+ * ```
+ *
+ * \see [Page of the guide "Middlewares"](https://crowcpp.org/master/guides/middleware/).
+ */
+#define CROW_MIDDLEWARES(app, ...) template middlewares<typename std::remove_reference<decltype(app)>::type, __VA_ARGS__>()
+
+#endif // #ifdef CROW_MSVC_WORKAROUND
+
+/**
+ * \def CROW_CATCHALL_ROUTE(app)
+ * \brief Defines a custom catchall route for app using a
+ * custom rule.
+ *
+ * It defines a handler when the client make a request for an
+ * undefined route. Instead of just reply with a `404` status
+ * code (default behavior), you can define a custom handler
+ * using this macro.
+ *
+ * \see [Page of the guide "Routes" (Catchall routes)](https://crowcpp.org/master/guides/routes/#catchall-routes).
+ */ 
+#define CROW_CATCHALL_ROUTE(app) app.catchall_route()
+
+/**
+ * \def CROW_BP_CATCHALL_ROUTE(blueprint)
+ * \brief Defines a custom catchall route for blueprint
+ * using a custom rule.
+ *
+ * It defines a handler when the client make a request for an
+ * undefined route in the blueprint.
+ *
+ * \see [Page of the guide "Blueprint" (Define a custom Catchall route)](https://crowcpp.org/master/guides/blueprints/#define-a-custom-catchall-route).
+ */ 
+#define CROW_BP_CATCHALL_ROUTE(blueprint) blueprint.catchall_rule()
+
+
+/**
+ * \namespace crow
+ * \brief The main namespace of the library. In this namespace
+ * is defined the most important classes and functions of the
+ * library.
+ * 
+ * Within this namespace, the Crow class, Router class, Connection
+ * class, and other are defined.
+ */
+namespace crow
+{
+#ifdef CROW_ENABLE_SSL
+    using ssl_context_t = asio::ssl::context;
+#endif
+    /**
+     * \class Crow
+     * \brief The main server application class.
+     *
+     * Use crow::SimpleApp or crow::App<Middleware1, Middleware2, etc...> instead of
+     * directly instantiate this class.
+     */
+    template<typename... Middlewares>
+    class Crow
+    {
+    public:
+        /// \brief This is the crow application
+        using self_t = Crow;
+
+        /// \brief The HTTP server
+        using server_t = Server<Crow, SocketAdaptor, Middlewares...>;
+
+#ifdef CROW_ENABLE_SSL
+        /// \brief An HTTP server that runs on SSL with an SSLAdaptor
+        using ssl_server_t = Server<Crow, SSLAdaptor, Middlewares...>;
+#endif
+        Crow()
+        {}
+
+        /// \brief Construct Crow with a subset of middleware
+        template<typename... Ts>
+        Crow(Ts&&... ts):
+          middlewares_(make_middleware_tuple(std::forward<Ts>(ts)...))
+        {}
+
+        /// \brief Process an Upgrade request
+        ///
+        /// Currently used to upgrade an HTTP connection to a WebSocket connection
+        template<typename Adaptor>
+        void handle_upgrade(const request& req, response& res, Adaptor&& adaptor)
+        {
+            router_.handle_upgrade(req, res, adaptor);
+        }
+
+        /// \brief Process only the method and URL of a request and provide a route (or an error response)
+        std::unique_ptr<routing_handle_result> handle_initial(request& req, response& res)
+        {
+            return router_.handle_initial(req, res);
+        }
+
+        /// \brief Process the fully parsed request and generate a response for it
+        void handle(request& req, response& res, std::unique_ptr<routing_handle_result>& found)
+        {
+            router_.handle<self_t>(req, res, *found);
+        }
+
+        /// \brief Process a fully parsed request from start to finish (primarily used for debugging)
+        void handle_full(request& req, response& res)
+        {
+            auto found = handle_initial(req, res);
+            if (found->rule_index)
+                handle(req, res, found);
+        }
+
+        /// \brief Create a dynamic route using a rule (**Use CROW_ROUTE instead**)
+        DynamicRule& route_dynamic(const std::string& rule)
+        {
+            return router_.new_rule_dynamic(rule);
+        }
+
+        /// \brief Create a route using a rule (**Use CROW_ROUTE instead**)
+        template<uint64_t Tag>
+#ifdef CROW_GCC83_WORKAROUND
+        auto& route(const std::string& rule)
+#else
+        auto route(const std::string& rule)
+#endif
+#if defined CROW_CAN_USE_CPP17 && !defined CROW_GCC83_WORKAROUND
+          -> typename std::invoke_result<decltype(&Router::new_rule_tagged<Tag>), Router, const std::string&>::type
+#elif !defined CROW_GCC83_WORKAROUND
+          -> typename std::result_of<decltype (&Router::new_rule_tagged<Tag>)(Router, const std::string&)>::type
+#endif
+        {
+            return router_.new_rule_tagged<Tag>(rule);
+        }
+
+        /// \brief Create a route for any requests without a proper route (**Use CROW_CATCHALL_ROUTE instead**)
+        CatchallRule& catchall_route()
+        {
+            return router_.catchall_rule();
+        }
+
+        /// \brief Set the default max payload size for websockets
+        self_t& websocket_max_payload(uint64_t max_payload)
+        {
+            max_payload_ = max_payload;
+            return *this;
+        }
+
+        /// \brief Get the default max payload size for websockets
+        uint64_t websocket_max_payload()
+        {
+            return max_payload_;
+        }
+
+        self_t& signal_clear()
+        {
+            signals_.clear();
+            return *this;
+        }
+
+        self_t& signal_add(int signal_number)
+        {
+            signals_.push_back(signal_number);
+            return *this;
+        }
+
+        std::vector<int> signals()
+        {
+            return signals_;
+        }
+
+        /// \brief Set the port that Crow will handle requests on
+        self_t& port(std::uint16_t port)
+        {
+            port_ = port;
+            return *this;
+        }
+
+        /// \brief Get the port that Crow will handle requests on
+        std::uint16_t port()
+        {
+            return port_;
+        }
+
+        /// \brief Set the connection timeout in seconds (default is 5)
+        self_t& timeout(std::uint8_t timeout)
+        {
+            timeout_ = timeout;
+            return *this;
+        }
+
+        /// \brief Set the server name
+        self_t& server_name(std::string server_name)
+        {
+            server_name_ = server_name;
+            return *this;
+        }
+
+        /// \brief The IP address that Crow will handle requests on (default is 0.0.0.0)
+        self_t& bindaddr(std::string bindaddr)
+        {
+            bindaddr_ = bindaddr;
+            return *this;
+        }
+
+        /// \brief Get the address that Crow will handle requests on
+        std::string bindaddr()
+        {
+            return bindaddr_;
+        }
+
+        /// \brief Run the server on multiple threads using all available threads
+        self_t& multithreaded()
+        {
+            return concurrency(std::thread::hardware_concurrency());
+        }
+
+        /// \brief Run the server on multiple threads using a specific number
+        self_t& concurrency(std::uint16_t concurrency)
+        {
+            if (concurrency < 2) // Crow can have a minimum of 2 threads running
+                concurrency = 2;
+            concurrency_ = concurrency;
+            return *this;
+        }
+
+        /// \brief Get the number of threads that server is using
+        std::uint16_t concurrency()
+        {
+            return concurrency_;
+        }
+
+        /// \brief Set the server's log level
+        ///
+        /// Possible values are:
+        /// - crow::LogLevel::Debug       (0)
+        /// - crow::LogLevel::Info        (1)
+        /// - crow::LogLevel::Warning     (2)
+        /// - crow::LogLevel::Error       (3)
+        /// - crow::LogLevel::Critical    (4)
+        self_t& loglevel(LogLevel level)
+        {
+            crow::logger::setLogLevel(level);
+            return *this;
+        }
+
+        /// \brief Set the response body size (in bytes) beyond which Crow automatically streams responses (Default is 1MiB)
+        ///
+        /// Any streamed response is unaffected by Crow's timer, and therefore won't timeout before a response is fully sent.
+        self_t& stream_threshold(size_t threshold)
+        {
+            res_stream_threshold_ = threshold;
+            return *this;
+        }
+
+        /// \brief Get the response body size (in bytes) beyond which Crow automatically streams responses
+        size_t& stream_threshold()
+        {
+            return res_stream_threshold_;
+        }
+
+        
+        self_t& register_blueprint(Blueprint& blueprint)
+        {
+            router_.register_blueprint(blueprint);
+            return *this;
+        }
+
+        /// \brief Set the function to call to handle uncaught exceptions generated in routes (Default generates error 500).
+        ///
+        /// The function must have the following signature: void(crow::response&).
+        /// It must set the response passed in argument to the function, which will be sent back to the client.
+        /// See Router::default_exception_handler() for the default implementation.
+        template<typename Func>
+        self_t& exception_handler(Func&& f)
+        {
+            router_.exception_handler() = std::forward<Func>(f);
+            return *this;
+        }
+
+        std::function<void(crow::response&)>& exception_handler()
+        {
+            return router_.exception_handler();
+        }
+
+        /// \brief Set a custom duration and function to run on every tick
+        template<typename Duration, typename Func>
+        self_t& tick(Duration d, Func f)
+        {
+            tick_interval_ = std::chrono::duration_cast<std::chrono::milliseconds>(d);
+            tick_function_ = f;
+            return *this;
+        }
+
+#ifdef CROW_ENABLE_COMPRESSION
+        
+        self_t& use_compression(compression::algorithm algorithm)
+        {
+            comp_algorithm_ = algorithm;
+            compression_used_ = true;
+            return *this;
+        }
+
+        compression::algorithm compression_algorithm()
+        {
+            return comp_algorithm_;
+        }
+
+        bool compression_used() const
+        {
+            return compression_used_;
+        }
+#endif
+
+        /// \brief Apply blueprints
+        void add_blueprint()
+        {
+#if defined(__APPLE__) || defined(__MACH__)
+            if (router_.blueprints().empty()) return;
+#endif
+
+            for (Blueprint* bp : router_.blueprints())
+            {
+                if (bp->static_dir().empty()) continue;
+
+                auto static_dir_ = crow::utility::normalize_path(bp->static_dir());
+
+                bp->new_rule_tagged<crow::black_magic::get_parameter_tag(CROW_STATIC_ENDPOINT)>(CROW_STATIC_ENDPOINT)([static_dir_](crow::response& res, std::string file_path_partial) {
+                    utility::sanitize_filename(file_path_partial);
+                    res.set_static_file_info_unsafe(static_dir_ + file_path_partial);
+                    res.end();
+                });
+            }
+
+            router_.validate_bp();
+        }
+
+        /// \brief Go through the rules, upgrade them if possible, and add them to the list of rules
+        void add_static_dir()
+        {
+            if (are_static_routes_added()) return;
+            auto static_dir_ = crow::utility::normalize_path(CROW_STATIC_DIRECTORY);
+
+            route<crow::black_magic::get_parameter_tag(CROW_STATIC_ENDPOINT)>(CROW_STATIC_ENDPOINT)([static_dir_](crow::response& res, std::string file_path_partial) {
+                utility::sanitize_filename(file_path_partial);
+                res.set_static_file_info_unsafe(static_dir_ + file_path_partial);
+                res.end();
+            });
+            set_static_routes_added();
+        }
+
+        /// \brief A wrapper for `validate()` in the router
+        void validate()
+        {
+            router_.validate();
+        }
+
+        /// \brief Run the server
+        void run()
+        {
+#ifndef CROW_DISABLE_STATIC_DIR
+            add_blueprint();
+            add_static_dir();
+#endif
+            validate();
+
+#ifdef CROW_ENABLE_SSL
+            if (ssl_used_)
+            {
+                ssl_server_ = std::move(std::unique_ptr<ssl_server_t>(new ssl_server_t(this, bindaddr_, port_, server_name_, &middlewares_, concurrency_, timeout_, &ssl_context_)));
+                ssl_server_->set_tick_function(tick_interval_, tick_function_);
+                ssl_server_->signal_clear();
+                for (auto snum : signals_)
+                {
+                    ssl_server_->signal_add(snum);
+                }
+                notify_server_start();
+                ssl_server_->run();
+            }
+            else
+#endif
+            {
+                server_ = std::move(std::unique_ptr<server_t>(new server_t(this, bindaddr_, port_, server_name_, &middlewares_, concurrency_, timeout_, nullptr)));
+                server_->set_tick_function(tick_interval_, tick_function_);
+                for (auto snum : signals_)
+                {
+                    server_->signal_add(snum);
+                }
+                notify_server_start();
+                server_->run();
+            }
+        }
+
+        /// \brief Non-blocking version of \ref run()
+        ///
+        /// The output from this method needs to be saved into a variable!
+        /// Otherwise the call will be made on the same thread.
+        std::future<void> run_async()
+        {
+            return std::async(std::launch::async, [&] {
+                this->run();
+            });
+        }
+
+        /// \brief Stop the server
+        void stop()
+        {
+#ifdef CROW_ENABLE_SSL
+            if (ssl_used_)
+            {
+                if (ssl_server_) { ssl_server_->stop(); }
+            }
+            else
+#endif
+            {
+                // TODO(EDev): Move these 6 lines to a method in http_server.
+                std::vector<crow::websocket::connection*> websockets_to_close = websockets_;
+                for (auto websocket : websockets_to_close)
+                {
+                    CROW_LOG_INFO << "Quitting Websocket: " << websocket;
+                    websocket->close("Server Application Terminated");
+                }
+                if (server_) { server_->stop(); }
+            }
+        }
+
+        void add_websocket(crow::websocket::connection* conn)
+        {
+            websockets_.push_back(conn);
+        }
+
+        void remove_websocket(crow::websocket::connection* conn)
+        {
+            websockets_.erase(std::remove(websockets_.begin(), websockets_.end(), conn), websockets_.end());
+        }
+
+        /// \brief Print the routing paths defined for each HTTP method
+        void debug_print()
+        {
+            CROW_LOG_DEBUG << "Routing:";
+            router_.debug_print();
+        }
+
+
+#ifdef CROW_ENABLE_SSL
+
+        /// \brief Use certificate and key files for SSL
+        self_t& ssl_file(const std::string& crt_filename, const std::string& key_filename)
+        {
+            ssl_used_ = true;
+            ssl_context_.set_verify_mode(asio::ssl::verify_peer);
+            ssl_context_.set_verify_mode(asio::ssl::verify_client_once);
+            ssl_context_.use_certificate_file(crt_filename, ssl_context_t::pem);
+            ssl_context_.use_private_key_file(key_filename, ssl_context_t::pem);
+            ssl_context_.set_options(
+              asio::ssl::context::default_workarounds | asio::ssl::context::no_sslv2 | asio::ssl::context::no_sslv3);
+            return *this;
+        }
+
+        /// \brief Use `.pem` file for SSL
+        self_t& ssl_file(const std::string& pem_filename)
+        {
+            ssl_used_ = true;
+            ssl_context_.set_verify_mode(asio::ssl::verify_peer);
+            ssl_context_.set_verify_mode(asio::ssl::verify_client_once);
+            ssl_context_.load_verify_file(pem_filename);
+            ssl_context_.set_options(
+              asio::ssl::context::default_workarounds | asio::ssl::context::no_sslv2 | asio::ssl::context::no_sslv3);
+            return *this;
+        }
+
+        /// \brief Use certificate chain and key files for SSL
+        self_t& ssl_chainfile(const std::string& crt_filename, const std::string& key_filename)
+        {
+            ssl_used_ = true;
+            ssl_context_.set_verify_mode(asio::ssl::verify_peer);
+            ssl_context_.set_verify_mode(asio::ssl::verify_client_once);
+            ssl_context_.use_certificate_chain_file(crt_filename);
+            ssl_context_.use_private_key_file(key_filename, ssl_context_t::pem);
+            ssl_context_.set_options(
+              asio::ssl::context::default_workarounds | asio::ssl::context::no_sslv2 | asio::ssl::context::no_sslv3);
+            return *this;
+        }
+
+        self_t& ssl(asio::ssl::context&& ctx)
+        {
+            ssl_used_ = true;
+            ssl_context_ = std::move(ctx);
+            return *this;
+        }
+
+        bool ssl_used() const
+        {
+            return ssl_used_;
+        }
+#else
+        
+        template<typename T, typename... Remain>
+        self_t& ssl_file(T&&, Remain&&...)
+        {
+            // We can't call .ssl() member function unless CROW_ENABLE_SSL is defined.
+            static_assert(
+              // make static_assert dependent to T; always false
+              std::is_base_of<T, void>::value,
+              "Define CROW_ENABLE_SSL to enable ssl support.");
+            return *this;
+        }
+
+        template<typename T, typename... Remain>
+        self_t& ssl_chainfile(T&&, Remain&&...)
+        {
+            // We can't call .ssl() member function unless CROW_ENABLE_SSL is defined.
+            static_assert(
+              // make static_assert dependent to T; always false
+              std::is_base_of<T, void>::value,
+              "Define CROW_ENABLE_SSL to enable ssl support.");
+            return *this;
+        }
+
+        template<typename T>
+        self_t& ssl(T&&)
+        {
+            // We can't call .ssl() member function unless CROW_ENABLE_SSL is defined.
+            static_assert(
+              // make static_assert dependent to T; always false
+              std::is_base_of<T, void>::value,
+              "Define CROW_ENABLE_SSL to enable ssl support.");
+            return *this;
+        }
+
+        bool ssl_used() const
+        {
+            return false;
+        }
+#endif
+
+        // middleware
+        using context_t = detail::context<Middlewares...>;
+        using mw_container_t = std::tuple<Middlewares...>;
+        template<typename T>
+        typename T::context& get_context(const request& req)
+        {
+            static_assert(black_magic::contains<T, Middlewares...>::value, "App doesn't have the specified middleware type.");
+            auto& ctx = *reinterpret_cast<context_t*>(req.middleware_context);
+            return ctx.template get<T>();
+        }
+
+        template<typename T>
+        T& get_middleware()
+        {
+            return utility::get_element_by_type<T, Middlewares...>(middlewares_);
+        }
+
+        /// \brief Wait until the server has properly started
+        void wait_for_server_start()
+        {
+            {
+                std::unique_lock<std::mutex> lock(start_mutex_);
+                if (!server_started_)
+                    cv_started_.wait(lock);
+            }
+            if (server_)
+                server_->wait_for_start();
+#ifdef CROW_ENABLE_SSL
+            else if (ssl_server_)
+                ssl_server_->wait_for_start();
+#endif
+        }
+
+    private:
+        template<typename... Ts>
+        std::tuple<Middlewares...> make_middleware_tuple(Ts&&... ts)
+        {
+            auto fwd = std::forward_as_tuple((ts)...);
+            return std::make_tuple(
+              std::forward<Middlewares>(
+                black_magic::tuple_extract<Middlewares, decltype(fwd)>(fwd))...);
+        }
+
+        /// \brief Notify anything using \ref wait_for_server_start() to proceed
+        void notify_server_start()
+        {
+            std::unique_lock<std::mutex> lock(start_mutex_);
+            server_started_ = true;
+            cv_started_.notify_all();
+        }
+
+        void set_static_routes_added() {
+            static_routes_added_ = true;
+        }
+
+        bool are_static_routes_added() {
+            return static_routes_added_;
+        }
+
+    private:
+        std::uint8_t timeout_{5};
+        uint16_t port_ = 80;
+        uint16_t concurrency_ = 2;
+        uint64_t max_payload_{UINT64_MAX};
+        std::string server_name_ = std::string("Crow/") + VERSION;
+        std::string bindaddr_ = "0.0.0.0";
+        size_t res_stream_threshold_ = 1048576;
+        Router router_;
+        bool static_routes_added_{false};
+
+#ifdef CROW_ENABLE_COMPRESSION
+        compression::algorithm comp_algorithm_;
+        bool compression_used_{false};
+#endif
+
+        std::chrono::milliseconds tick_interval_;
+        std::function<void()> tick_function_;
+
+        std::tuple<Middlewares...> middlewares_;
+
+#ifdef CROW_ENABLE_SSL
+        std::unique_ptr<ssl_server_t> ssl_server_;
+        bool ssl_used_{false};
+        ssl_context_t ssl_context_{asio::ssl::context::sslv23};
+#endif
+
+        std::unique_ptr<server_t> server_;
+
+        std::vector<int> signals_{SIGINT, SIGTERM};
+
+        bool server_started_{false};
+        std::condition_variable cv_started_;
+        std::mutex start_mutex_;
+        std::vector<crow::websocket::connection*> websockets_;
+    };
+
+    /// \brief Alias of Crow<Middlewares...>. Useful if you want
+    /// a instance of an Crow application that require Middlewares
+    template<typename... Middlewares>
+    using App = Crow<Middlewares...>;
+
+    /// \brief Alias of Crow<>. Useful if you want a instance of
+    /// an Crow application that doesn't require of Middlewares
+    using SimpleApp = Crow<>;
+} // namespace crow
+

benchmark/crow/main.cpp

@@ -0,0 +1,17 @@
+#include "crow_all.h"
+
+class CustomLogger : public crow::ILogHandler {
+public:
+  void log(std::string, crow::LogLevel) {}
+};
+
+int main() {
+  CustomLogger logger;
+  crow::logger::setHandler(&logger);
+
+  crow::SimpleApp app;
+
+  CROW_ROUTE(app, "/")([]() { return "Hello world!"; });
+
+  app.port(8080).multithreaded().run();
+}

benchmark/flask/main.py

@@ -0,0 +1,9 @@
+from flask import Flask
+app = Flask(__name__)
+
+import logging
+logging.getLogger('werkzeug').disabled = True
+
[email protected]('/')
+def hello_world():
+    return 'Hello, World!'