Update sqlite3

SquiLu-ext/sqlite3.c

@@ -233,7 +233,7 @@ extern "C" {
 */
 #define SQLITE_VERSION        "3.8.8"
 #define SQLITE_VERSION_NUMBER 3008008
-#define SQLITE_SOURCE_ID      "2014-11-20 02:58:10 2d7c8da5f16e64eaa7b0c2d66898682ea3d102a0"
+#define SQLITE_SOURCE_ID      "2014-11-27 11:36:36 f095cde579e7417306e11b5c1d2dd90b6bb547d5"
 
 /*
 ** CAPI3REF: Run-Time Library Version Numbers
@@ -6065,34 +6065,34 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 **
 ** The SQLite source code contains multiple implementations
 ** of these mutex routines.  An appropriate implementation
-** is selected automatically at compile-time.  ^(The following
+** is selected automatically at compile-time.  The following
 ** implementations are available in the SQLite core:
 **
 ** <ul>
 ** <li>   SQLITE_MUTEX_PTHREADS
 ** <li>   SQLITE_MUTEX_W32
 ** <li>   SQLITE_MUTEX_NOOP
-** </ul>)^
+** </ul>
 **
-** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
+** The SQLITE_MUTEX_NOOP implementation is a set of routines
 ** that does no real locking and is appropriate for use in
-** a single-threaded application.  ^The SQLITE_MUTEX_PTHREADS and
+** a single-threaded application.  The SQLITE_MUTEX_PTHREADS and
 ** SQLITE_MUTEX_W32 implementations are appropriate for use on Unix
 ** and Windows.
 **
-** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
+** If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
 ** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
 ** implementation is included with the library. In this case the
 ** application must supply a custom mutex implementation using the
 ** [SQLITE_CONFIG_MUTEX] option of the sqlite3_config() function
 ** before calling sqlite3_initialize() or any other public sqlite3_
-** function that calls sqlite3_initialize().)^
+** function that calls sqlite3_initialize().
 **
 ** ^The sqlite3_mutex_alloc() routine allocates a new
-** mutex and returns a pointer to it. ^If it returns NULL
-** that means that a mutex could not be allocated.  ^SQLite
-** will unwind its stack and return an error.  ^(The argument
-** to sqlite3_mutex_alloc() is one of these integer constants:
+** mutex and returns a pointer to it. ^The sqlite3_mutex_alloc()
+** routine returns NULL if it is unable to allocate the requested
+** mutex.  The argument to sqlite3_mutex_alloc() must one of these
+** integer constants:
 **
 ** <ul>
 ** <li>  SQLITE_MUTEX_FAST
@@ -6105,7 +6105,8 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** <li>  SQLITE_MUTEX_STATIC_PMEM
 ** <li>  SQLITE_MUTEX_STATIC_APP1
 ** <li>  SQLITE_MUTEX_STATIC_APP2
-** </ul>)^
+** <li>  SQLITE_MUTEX_STATIC_APP3
+** </ul>
 **
 ** ^The first two constants (SQLITE_MUTEX_FAST and SQLITE_MUTEX_RECURSIVE)
 ** cause sqlite3_mutex_alloc() to create
@@ -6113,14 +6114,14 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** is used but not necessarily so when SQLITE_MUTEX_FAST is used.
 ** The mutex implementation does not need to make a distinction
 ** between SQLITE_MUTEX_RECURSIVE and SQLITE_MUTEX_FAST if it does
-** not want to.  ^SQLite will only request a recursive mutex in
-** cases where it really needs one.  ^If a faster non-recursive mutex
+** not want to.  SQLite will only request a recursive mutex in
+** cases where it really needs one.  If a faster non-recursive mutex
 ** implementation is available on the host platform, the mutex subsystem
 ** might return such a mutex in response to SQLITE_MUTEX_FAST.
 **
 ** ^The other allowed parameters to sqlite3_mutex_alloc() (anything other
 ** than SQLITE_MUTEX_FAST and SQLITE_MUTEX_RECURSIVE) each return
-** a pointer to a static preexisting mutex.  ^Six static mutexes are
+** a pointer to a static preexisting mutex.  ^Nine static mutexes are
 ** used by the current version of SQLite.  Future versions of SQLite
 ** may add additional static mutexes.  Static mutexes are for internal
 ** use by SQLite only.  Applications that use SQLite mutexes should
@@ -6129,16 +6130,13 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 **
 ** ^Note that if one of the dynamic mutex parameters (SQLITE_MUTEX_FAST
 ** or SQLITE_MUTEX_RECURSIVE) is used then sqlite3_mutex_alloc()
-** returns a different mutex on every call.  ^But for the static
+** returns a different mutex on every call.  ^For the static
 ** mutex types, the same mutex is returned on every call that has
 ** the same type number.
 **
 ** ^The sqlite3_mutex_free() routine deallocates a previously
-** allocated dynamic mutex.  ^SQLite is careful to deallocate every
-** dynamic mutex that it allocates.  The dynamic mutexes must not be in
-** use when they are deallocated.  Attempting to deallocate a static
-** mutex results in undefined behavior.  ^SQLite never deallocates
-** a static mutex.
+** allocated dynamic mutex.  Attempting to deallocate a static
+** mutex results in undefined behavior.
 **
 ** ^The sqlite3_mutex_enter() and sqlite3_mutex_try() routines attempt
 ** to enter a mutex.  ^If another thread is already within the mutex,
@@ -6146,23 +6144,21 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** SQLITE_BUSY.  ^The sqlite3_mutex_try() interface returns [SQLITE_OK]
 ** upon successful entry.  ^(Mutexes created using
 ** SQLITE_MUTEX_RECURSIVE can be entered multiple times by the same thread.
-** In such cases the,
+** In such cases, the
 ** mutex must be exited an equal number of times before another thread
-** can enter.)^  ^(If the same thread tries to enter any other
-** kind of mutex more than once, the behavior is undefined.
-** SQLite will never exhibit
-** such behavior in its own use of mutexes.)^
+** can enter.)^  If the same thread tries to enter any mutex other
+** than an SQLITE_MUTEX_RECURSIVE more than once, the behavior is undefined.
 **
 ** ^(Some systems (for example, Windows 95) do not support the operation
 ** implemented by sqlite3_mutex_try().  On those systems, sqlite3_mutex_try()
-** will always return SQLITE_BUSY.  The SQLite core only ever uses
-** sqlite3_mutex_try() as an optimization so this is acceptable behavior.)^
+** will always return SQLITE_BUSY. The SQLite core only ever uses
+** sqlite3_mutex_try() as an optimization so this is acceptable 
+** behavior.)^
 **
 ** ^The sqlite3_mutex_leave() routine exits a mutex that was
-** previously entered by the same thread.   ^(The behavior
+** previously entered by the same thread.   The behavior
 ** is undefined if the mutex is not currently entered by the
-** calling thread or is not currently allocated.  SQLite will
-** never do either.)^
+** calling thread or is not currently allocated.
 **
 ** ^If the argument to sqlite3_mutex_enter(), sqlite3_mutex_try(), or
 ** sqlite3_mutex_leave() is a NULL pointer, then all three routines
@@ -6183,9 +6179,9 @@ SQLITE_API void sqlite3_mutex_leave(sqlite3_mutex*);
 ** used to allocate and use mutexes.
 **
 ** Usually, the default mutex implementations provided by SQLite are
-** sufficient, however the user has the option of substituting a custom
+** sufficient, however the application has the option of substituting a custom
 ** implementation for specialized deployments or systems for which SQLite
-** does not provide a suitable implementation. In this case, the user
+** does not provide a suitable implementation. In this case, the application
 ** creates and populates an instance of this structure to pass
 ** to sqlite3_config() along with the [SQLITE_CONFIG_MUTEX] option.
 ** Additionally, an instance of this structure can be used as an
@@ -6226,13 +6222,13 @@ SQLITE_API void sqlite3_mutex_leave(sqlite3_mutex*);
 ** (i.e. it is acceptable to provide an implementation that segfaults if
 ** it is passed a NULL pointer).
 **
-** The xMutexInit() method must be threadsafe.  ^It must be harmless to
+** The xMutexInit() method must be threadsafe.  It must be harmless to
 ** invoke xMutexInit() multiple times within the same process and without
 ** intervening calls to xMutexEnd().  Second and subsequent calls to
 ** xMutexInit() must be no-ops.
 **
-** ^xMutexInit() must not use SQLite memory allocation ([sqlite3_malloc()]
-** and its associates).  ^Similarly, xMutexAlloc() must not use SQLite memory
+** xMutexInit() must not use SQLite memory allocation ([sqlite3_malloc()]
+** and its associates).  Similarly, xMutexAlloc() must not use SQLite memory
 ** allocation for a static mutex.  ^However xMutexAlloc() may use SQLite
 ** memory allocation for a fast or recursive mutex.
 **
@@ -6258,29 +6254,29 @@ struct sqlite3_mutex_methods {
 ** CAPI3REF: Mutex Verification Routines
 **
 ** The sqlite3_mutex_held() and sqlite3_mutex_notheld() routines
-** are intended for use inside assert() statements.  ^The SQLite core
+** are intended for use inside assert() statements.  The SQLite core
 ** never uses these routines except inside an assert() and applications
-** are advised to follow the lead of the core.  ^The SQLite core only
+** are advised to follow the lead of the core.  The SQLite core only
 ** provides implementations for these routines when it is compiled
-** with the SQLITE_DEBUG flag.  ^External mutex implementations
+** with the SQLITE_DEBUG flag.  External mutex implementations
 ** are only required to provide these routines if SQLITE_DEBUG is
 ** defined and if NDEBUG is not defined.
 **
-** ^These routines should return true if the mutex in their argument
+** These routines should return true if the mutex in their argument
 ** is held or not held, respectively, by the calling thread.
 **
-** ^The implementation is not required to provide versions of these
+** The implementation is not required to provide versions of these
 ** routines that actually work. If the implementation does not provide working
 ** versions of these routines, it should at least provide stubs that always
 ** return true so that one does not get spurious assertion failures.
 **
-** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
+** If the argument to sqlite3_mutex_held() is a NULL pointer then
 ** the routine should return 1.   This seems counter-intuitive since
 ** clearly the mutex cannot be held if it does not exist.  But
 ** the reason the mutex does not exist is because the build is not
 ** using mutexes.  And we do not want the assert() containing the
 ** call to sqlite3_mutex_held() to fail, so a non-zero return is
-** the appropriate thing to do.  ^The sqlite3_mutex_notheld()
+** the appropriate thing to do.  The sqlite3_mutex_notheld()
 ** interface should also return 1 when given a NULL pointer.
 */
 #ifndef NDEBUG
@@ -53809,6 +53805,11 @@ static void ptrmapPutOvflPtr(MemPage *pPage, u8 *pCell, int *pRC){
 ** end of the page and all free space is collected into one
 ** big FreeBlk that occurs in between the header and cell
 ** pointer array and the cell content area.
+**
+** EVIDENCE-OF: R-44582-60138 SQLite may from time to time reorganize a
+** b-tree page so that there are no freeblocks or fragment bytes, all
+** unused bytes are contained in the unallocated space region, and all
+** cells are packed tightly at the end of the page.
 */
 static int defragmentPage(MemPage *pPage){
   int i;                     /* Loop counter */
@@ -53932,6 +53933,8 @@ static u8 *pageFindSlot(MemPage *pPg, int nByte, int *pRc, int *pbDefrag){
       testcase( x==4 );
       testcase( x==3 );
       if( x<4 ){
+        /* EVIDENCE-OF: R-11498-58022 In a well-formed b-tree page, the total
+        ** number of bytes in fragments may not exceed 60. */
         if( aData[hdr+7]>=60 ){
           if( pbDefrag ) *pbDefrag = 1;
           return 0;
@@ -53986,19 +53989,13 @@ static int allocateSpace(MemPage *pPage, int nByte, int *pIdx){
   assert( pPage->cellOffset == hdr + 12 - 4*pPage->leaf );
   gap = pPage->cellOffset + 2*pPage->nCell;
   assert( gap<=65536 );
-  top = get2byte(&data[hdr+5]);
-  if( gap>top ){
-    if( top==0 ){
-      /* EVIDENCE-OF: R-29356-02391 If the database uses a 65536-byte page size
-      ** and the reserved space is zero (the usual value for reserved space)
-      ** then the cell content offset of an empty page wants to be 65536.
-      ** However, that integer is too large to be stored in a 2-byte unsigned
-      ** integer, so a value of 0 is used in its place. */
-      top = 65536;
-    }else{
-      return SQLITE_CORRUPT_BKPT;
-    }
-  }
+  /* EVIDENCE-OF: R-29356-02391 If the database uses a 65536-byte page size
+  ** and the reserved space is zero (the usual value for reserved space)
+  ** then the cell content offset of an empty page wants to be 65536.
+  ** However, that integer is too large to be stored in a 2-byte unsigned
+  ** integer, so a value of 0 is used in its place. */
+  top = get2byteNotZero(&data[hdr+5]);
+  if( gap>top ) return SQLITE_CORRUPT_BKPT;
 
   /* If there is enough space between gap and top for one more cell pointer
   ** array entry offset, and if the freelist is not empty, then search the
@@ -54167,18 +54164,32 @@ static int decodeFlags(MemPage *pPage, int flagByte){
   pPage->childPtrSize = 4-4*pPage->leaf;
   pBt = pPage->pBt;
   if( flagByte==(PTF_LEAFDATA | PTF_INTKEY) ){
+    /* EVIDENCE-OF: R-03640-13415 A value of 5 means the page is an interior
+    ** table b-tree page. */
+    assert( (PTF_LEAFDATA|PTF_INTKEY)==5 );
+    /* EVIDENCE-OF: R-20501-61796 A value of 13 means the page is a leaf
+    ** table b-tree page. */
+    assert( (PTF_LEAFDATA|PTF_INTKEY|PTF_LEAF)==13 );
     pPage->intKey = 1;
     pPage->intKeyLeaf = pPage->leaf;
     pPage->noPayload = !pPage->leaf;
     pPage->maxLocal = pBt->maxLeaf;
     pPage->minLocal = pBt->minLeaf;
   }else if( flagByte==PTF_ZERODATA ){
+    /* EVIDENCE-OF: R-27225-53936 A value of 2 means the page is an interior
+    ** index b-tree page. */
+    assert( (PTF_ZERODATA)==2 );
+    /* EVIDENCE-OF: R-16571-11615 A value of 10 means the page is a leaf
+    ** index b-tree page. */
+    assert( (PTF_ZERODATA|PTF_LEAF)==10 );
     pPage->intKey = 0;
     pPage->intKeyLeaf = 0;
     pPage->noPayload = 0;
     pPage->maxLocal = pBt->maxLocal;
     pPage->minLocal = pBt->minLocal;
   }else{
+    /* EVIDENCE-OF: R-47608-56469 Any other value for the b-tree page type is
+    ** an error. */
     return SQLITE_CORRUPT_BKPT;
   }
   pPage->max1bytePayload = pBt->max1bytePayload;
@@ -54218,21 +54229,33 @@ static int btreeInitPage(MemPage *pPage){
 
     hdr = pPage->hdrOffset;
     data = pPage->aData;
+    /* EVIDENCE-OF: R-28594-02890 The one-byte flag at offset 0 indicating
+    ** the b-tree page type. */
     if( decodeFlags(pPage, data[hdr]) ) return SQLITE_CORRUPT_BKPT;
     assert( pBt->pageSize>=512 && pBt->pageSize<=65536 );
     pPage->maskPage = (u16)(pBt->pageSize - 1);
     pPage->nOverflow = 0;
     usableSize = pBt->usableSize;
-    pPage->cellOffset = cellOffset = hdr + 12 - 4*pPage->leaf;
+    pPage->cellOffset = cellOffset = hdr + 8 + pPage->childPtrSize;
     pPage->aDataEnd = &data[usableSize];
     pPage->aCellIdx = &data[cellOffset];
+    /* EVIDENCE-OF: R-58015-48175 The two-byte integer at offset 5 designates
+    ** the start of the cell content area. A zero value for this integer is
+    ** interpreted as 65536. */
     top = get2byteNotZero(&data[hdr+5]);
+    /* EVIDENCE-OF: R-37002-32774 The two-byte integer at offset 3 gives the
+    ** number of cells on the page. */
     pPage->nCell = get2byte(&data[hdr+3]);
     if( pPage->nCell>MX_CELL(pBt) ){
       /* To many cells for a single page.  The page must be corrupt */
       return SQLITE_CORRUPT_BKPT;
     }
     testcase( pPage->nCell==MX_CELL(pBt) );
+    /* EVIDENCE-OF: R-24089-57979 If a page contains no cells (which is only
+    ** possible for a root page of a table that contains no rows) then the
+    ** offset to the cell content area will equal the page size minus the
+    ** bytes of reserved space. */
+    assert( pPage->nCell>0 || top==usableSize || CORRUPT_DB );
 
     /* A malformed database page might cause us to read past the end
     ** of page when parsing a cell.  
@@ -54266,13 +54289,20 @@ static int btreeInitPage(MemPage *pPage){
     }  
 #endif
 
-    /* Compute the total free space on the page */
+    /* Compute the total free space on the page
+    ** EVIDENCE-OF: R-23588-34450 The two-byte integer at offset 1 gives the
+    ** start of the first freeblock on the page, or is zero if there are no
+    ** freeblocks. */
     pc = get2byte(&data[hdr+1]);
-    nFree = data[hdr+7] + top;
+    nFree = data[hdr+7] + top;  /* Init nFree to non-freeblock free space */
     while( pc>0 ){
       u16 next, size;
       if( pc<iCellFirst || pc>iCellLast ){
-        /* Start of free block is off the page */
+        /* EVIDENCE-OF: R-55530-52930 In a well-formed b-tree page, there will
+        ** always be at least one cell before the first freeblock.
+        **
+        ** Or, the freeblock is off the end of the page
+        */
         return SQLITE_CORRUPT_BKPT; 
       }
       next = get2byte(&data[pc]);
@@ -58629,9 +58659,17 @@ static void dropCell(MemPage *pPage, int idx, int sz, int *pRC){
     return;
   }
   pPage->nCell--;
-  memmove(ptr, ptr+2, 2*(pPage->nCell - idx));
-  put2byte(&data[hdr+3], pPage->nCell);
-  pPage->nFree += 2;
+  if( pPage->nCell==0 ){
+    memset(&data[hdr+1], 0, 4);
+    data[hdr+7] = 0;
+    put2byte(&data[hdr+5], pPage->pBt->usableSize);
+    pPage->nFree = pPage->pBt->usableSize - pPage->hdrOffset
+                       - pPage->childPtrSize - 8;
+  }else{
+    memmove(ptr, ptr+2, 2*(pPage->nCell - idx));
+    put2byte(&data[hdr+3], pPage->nCell);
+    pPage->nFree += 2;
+  }
 }
 
 /*
@@ -58885,6 +58923,14 @@ static int pageFreeArray(
 }
 
 /*
+** apCell[] and szCell[] contains pointers to and sizes of all cells in the
+** pages being balanced.  The current page, pPg, has pPg->nCell cells starting
+** with apCell[iOld].  After balancing, this page should hold nNew cells
+** starting at apCell[iNew].
+**
+** This routine makes the necessary adjustments to pPg so that it contains
+** the correct cells after being balanced.
+**
 ** The pPg->nFree field is invalid when this function returns. It is the
 ** responsibility of the caller to set it correctly.
 */
@@ -58925,12 +58971,13 @@ static void editPage(
     );
   }
 
-  pData = &aData[get2byte(&aData[hdr+5])];
+  pData = &aData[get2byteNotZero(&aData[hdr+5])];
   if( pData<pBegin ) goto editpage_fail;
 
   /* Add cells to the start of the page */
   if( iNew<iOld ){
-    int nAdd = iOld-iNew;
+    int nAdd = MIN(nNew,iOld-iNew);
+    assert( (iOld-iNew)<nNew || nCell==0 || CORRUPT_DB );
     pCellptr = pPg->aCellIdx;
     memmove(&pCellptr[nAdd*2], pCellptr, nCell*2);
     if( pageInsertArray(
@@ -61279,8 +61326,14 @@ static int checkTreePage(
     assert( contentOffset<=usableSize );  /* Enforced by btreeInitPage() */
     memset(hit+contentOffset, 0, usableSize-contentOffset);
     memset(hit, 1, contentOffset);
+    /* EVIDENCE-OF: R-37002-32774 The two-byte integer at offset 3 gives the
+    ** number of cells on the page. */
     nCell = get2byte(&data[hdr+3]);
+    /* EVIDENCE-OF: R-23882-45353 The cell pointer array of a b-tree page
+    ** immediately follows the b-tree page header. */
     cellStart = hdr + 12 - 4*pPage->leaf;
+    /* EVIDENCE-OF: R-02776-14802 The cell pointer array consists of K 2-byte
+    ** integer offsets to the cell contents. */
     for(i=0; i<nCell; i++){
       int pc = get2byte(&data[cellStart+i*2]);
       u32 size = 65536;
@@ -61296,6 +61349,9 @@ static int checkTreePage(
         for(j=pc+size-1; j>=pc; j--) hit[j]++;
       }
     }
+    /* EVIDENCE-OF: R-20690-50594 The second field of the b-tree page header
+    ** is the offset of the first freeblock, or zero if there are no
+    ** freeblocks on the page. */
     i = get2byte(&data[hdr+1]);
     while( i>0 ){
       int size, j;
@@ -61303,7 +61359,13 @@ static int checkTreePage(
       size = get2byte(&data[i+2]);
       assert( i+size<=usableSize );  /* Enforced by btreeInitPage() */
       for(j=i+size-1; j>=i; j--) hit[j]++;
+      /* EVIDENCE-OF: R-58208-19414 The first 2 bytes of a freeblock are a
+      ** big-endian integer which is the offset in the b-tree page of the next
+      ** freeblock in the chain, or zero if the freeblock is the last on the
+      ** chain. */
       j = get2byte(&data[i]);
+      /* EVIDENCE-OF: R-06866-39125 Freeblocks are always connected in order of
+      ** increasing offset. */
       assert( j==0 || j>i+size );  /* Enforced by btreeInitPage() */
       assert( j<=usableSize-4 );   /* Enforced by btreeInitPage() */
       i = j;
@@ -61317,6 +61379,11 @@ static int checkTreePage(
         break;
       }
     }
+    /* EVIDENCE-OF: R-43263-13491 The total number of bytes in all fragments
+    ** is stored in the fifth field of the b-tree page header.
+    ** EVIDENCE-OF: R-07161-27322 The one-byte integer at offset 7 gives the
+    ** number of fragmented free bytes within the cell content area.
+    */
     if( cnt!=data[hdr+7] ){
       checkAppendMsg(pCheck,
           "Fragmentation of %d bytes reported as %d on page %d",
@@ -77172,6 +77239,13 @@ SQLITE_API int sqlite3_blob_reopen(sqlite3_blob *pBlob, sqlite3_int64 iRow){
 # define SQLITE_DEBUG_SORTER_THREADS 1
 #endif
 
+/*
+** Hard-coded maximum amount of data to accumulate in memory before flushing
+** to a level 0 PMA. The purpose of this limit is to prevent various integer
+** overflows. 512MiB.
+*/
+#define SQLITE_MAX_MXPMASIZE    (1<<29)
+
 /*
 ** Private objects used by the sorter
 */
@@ -77870,7 +77944,7 @@ SQLITE_PRIVATE int sqlite3VdbeSorterInit(
       pSorter->mnPmaSize = SORTER_MIN_WORKING * pgsz;
       mxCache = db->aDb[0].pSchema->cache_size;
       if( mxCache<SORTER_MIN_WORKING ) mxCache = SORTER_MIN_WORKING;
-      pSorter->mxPmaSize = mxCache * pgsz;
+      pSorter->mxPmaSize = MIN((i64)mxCache*pgsz, SQLITE_MAX_MXPMASIZE);
 
       /* EVIDENCE-OF: R-26747-61719 When the application provides any amount of
       ** scratch memory using SQLITE_CONFIG_SCRATCH, SQLite avoids unnecessary
@@ -88320,8 +88394,6 @@ static void decodeIntArray(
   if( z==0 ) z = "";
 #else
   assert( z!=0 );
-  pIndex->bUnordered = 0;
-  pIndex->noSkipScan = 0;
 #endif
   for(i=0; *z && i<nOut; i++){
     v = 0;
@@ -88341,25 +88413,28 @@ static void decodeIntArray(
     if( *z==' ' ) z++;
   }
 #ifndef SQLITE_ENABLE_STAT3_OR_STAT4
-  assert( pIndex!=0 );
+  assert( pIndex!=0 ); {
 #else
-  if( pIndex )
+  if( pIndex ){
 #endif
-  while( z[0] ){
-    if( sqlite3_strglob("unordered*", z)==0 ){
-      pIndex->bUnordered = 1;
-    }else if( sqlite3_strglob("sz=[0-9]*", z)==0 ){
-      pIndex->szIdxRow = sqlite3LogEst(sqlite3Atoi(z+3));
-    }else if( sqlite3_strglob("noskipscan*", z)==0 ){
-      pIndex->noSkipScan = 1;
-    }
+    pIndex->bUnordered = 0;
+    pIndex->noSkipScan = 0;
+    while( z[0] ){
+      if( sqlite3_strglob("unordered*", z)==0 ){
+        pIndex->bUnordered = 1;
+      }else if( sqlite3_strglob("sz=[0-9]*", z)==0 ){
+        pIndex->szIdxRow = sqlite3LogEst(sqlite3Atoi(z+3));
+      }else if( sqlite3_strglob("noskipscan*", z)==0 ){
+        pIndex->noSkipScan = 1;
+      }
 #ifdef SQLITE_ENABLE_COSTMULT
-    else if( sqlite3_strglob("costmult=[0-9]*",z)==0 ){
-      pIndex->pTable->costMult = sqlite3LogEst(sqlite3Atoi(z+9));
-    }
+      else if( sqlite3_strglob("costmult=[0-9]*",z)==0 ){
+        pIndex->pTable->costMult = sqlite3LogEst(sqlite3Atoi(z+9));
+      }
 #endif
-    while( z[0]!=0 && z[0]!=' ' ) z++;
-    while( z[0]==' ' ) z++;
+      while( z[0]!=0 && z[0]!=' ' ) z++;
+      while( z[0]==' ' ) z++;
+    }
   }
 }
 
@@ -95394,8 +95469,8 @@ static void absFunc(sqlite3_context *context, int argc, sqlite3_value **argv){
     default: {
       /* Because sqlite3_value_double() returns 0.0 if the argument is not
       ** something that can be converted into a number, we have:
-      ** IMP: R-57326-31541 Abs(X) return 0.0 if X is a string or blob that
-      ** cannot be converted to a numeric value. 
+      ** IMP: R-01992-00519 Abs(X) returns 0.0 if X is a string or blob
+      ** that cannot be converted to a numeric value.
       */
       sqlite_double rVal = sqlite3_value_double(argv[0]);
       if( rVal<0 ) rVal = -rVal;
@@ -113505,7 +113580,12 @@ SQLITE_PRIVATE void sqlite3VtabBeginParse(
   addModuleArgument(db, pTable, sqlite3NameFromToken(db, pModuleName));
   addModuleArgument(db, pTable, 0);
   addModuleArgument(db, pTable, sqlite3DbStrDup(db, pTable->zName));
-  pParse->sNameToken.n = (int)(&pModuleName->z[pModuleName->n] - pName1->z);
+  assert( (pParse->sNameToken.z==pName2->z && pName2->z!=0)
+       || (pParse->sNameToken.z==pName1->z && pName2->z==0)
+  );
+  pParse->sNameToken.n = (int)(
+      &pModuleName->z[pModuleName->n] - pParse->sNameToken.z
+  );
 
 #ifndef SQLITE_OMIT_AUTHORIZATION
   /* Creating a virtual table invokes the authorization callback twice.
@@ -119046,10 +119126,30 @@ static int whereLoopInsert(WhereLoopBuilder *pBuilder, WhereLoop *pTemplate){
 ** Adjust the WhereLoop.nOut value downward to account for terms of the
 ** WHERE clause that reference the loop but which are not used by an
 ** index.
-**
-** In the current implementation, the first extra WHERE clause term reduces
-** the number of output rows by a factor of 10 and each additional term
-** reduces the number of output rows by sqrt(2).
+*
+** For every WHERE clause term that is not used by the index
+** and which has a truth probability assigned by one of the likelihood(),
+** likely(), or unlikely() SQL functions, reduce the estimated number
+** of output rows by the probability specified.
+**
+** TUNING:  For every WHERE clause term that is not used by the index
+** and which does not have an assigned truth probability, heuristics
+** described below are used to try to estimate the truth probability.
+** TODO --> Perhaps this is something that could be improved by better
+** table statistics.
+**
+** Heuristic 1:  Estimate the truth probability as 93.75%.  The 93.75%
+** value corresponds to -1 in LogEst notation, so this means decrement
+** the WhereLoop.nOut field for every such WHERE clause term.
+**
+** Heuristic 2:  If there exists one or more WHERE clause terms of the
+** form "x==EXPR" and EXPR is not a constant 0 or 1, then make sure the
+** final output row estimate is no greater than 1/4 of the total number
+** of rows in the table.  In other words, assume that x==EXPR will filter
+** out at least 3 out of 4 rows.  If EXPR is -1 or 0 or 1, then maybe the
+** "x" column is boolean or else -1 or 0 or 1 is a common default value
+** on the "x" column and so in that case only cap the output row estimate
+** at 1/2 instead of 1/4.
 */
 static void whereLoopOutputAdjust(
   WhereClause *pWC,      /* The WHERE clause */
@@ -119058,9 +119158,10 @@ static void whereLoopOutputAdjust(
 ){
   WhereTerm *pTerm, *pX;
   Bitmask notAllowed = ~(pLoop->prereq|pLoop->maskSelf);
-  int i, j;
-  int nEq = 0;    /* Number of = constraints not within likely()/unlikely() */
+  int i, j, k;
+  LogEst iReduce = 0;    /* pLoop->nOut should not exceed nRow-iReduce */
 
+  assert( (pLoop->wsFlags & WHERE_AUTO_INDEX)==0 );
   for(i=pWC->nTerm, pTerm=pWC->a; i>0; i--, pTerm++){
     if( (pTerm->wtFlags & TERM_VIRTUAL)!=0 ) break;
     if( (pTerm->prereqAll & pLoop->maskSelf)==0 ) continue;
@@ -119073,20 +119174,26 @@ static void whereLoopOutputAdjust(
     }
     if( j<0 ){
       if( pTerm->truthProb<=0 ){
+        /* If a truth probability is specified using the likelihood() hints,
+        ** then use the probability provided by the application. */
         pLoop->nOut += pTerm->truthProb;
       }else{
+        /* In the absence of explicit truth probabilities, use heuristics to
+        ** guess a reasonable truth probability. */
         pLoop->nOut--;
-        if( pTerm->eOperator&WO_EQ ) nEq++;
+        if( pTerm->eOperator&WO_EQ ){
+          Expr *pRight = pTerm->pExpr->pRight;
+          if( sqlite3ExprIsInteger(pRight, &k) && k>=(-1) && k<=1 ){
+            k = 10;
+          }else{
+            k = 20;
+          }
+          if( iReduce<k ) iReduce = k;
+        }
       }
     }
   }
-  /* TUNING:  If there is at least one equality constraint in the WHERE
-  ** clause that does not have a likelihood() explicitly assigned to it
-  ** then do not let the estimated number of output rows exceed half 
-  ** the number of rows in the table. */
-  if( nEq && pLoop->nOut>nRow-10 ){
-    pLoop->nOut = nRow - 10;
-  }
+  if( pLoop->nOut > nRow-iReduce )  pLoop->nOut = nRow - iReduce;
 }
 
 /*
@@ -127016,13 +127123,20 @@ static int binCollFunc(
 ){
   int rc, n;
   n = nKey1<nKey2 ? nKey1 : nKey2;
+  /* EVIDENCE-OF: R-65033-28449 The built-in BINARY collation compares
+  ** strings byte by byte using the memcmp() function from the standard C
+  ** library. */
   rc = memcmp(pKey1, pKey2, n);
   if( rc==0 ){
     if( padFlag
      && allSpaces(((char*)pKey1)+n, nKey1-n)
      && allSpaces(((char*)pKey2)+n, nKey2-n)
     ){
-      /* Leave rc unchanged at 0 */
+      /* EVIDENCE-OF: R-31624-24737 RTRIM is like BINARY except that extra
+      ** spaces at the end of either string do not change the result. In other
+      ** words, strings will compare equal to one another as long as they
+      ** differ only in the number of spaces at the end.
+      */
     }else{
       rc = nKey1 - nKey2;
     }
@@ -129023,14 +129137,21 @@ static int openDatabase(
   /* Add the default collation sequence BINARY. BINARY works for both UTF-8
   ** and UTF-16, so add a version for each to avoid any unnecessary
   ** conversions. The only error that can occur here is a malloc() failure.
+  **
+  ** EVIDENCE-OF: R-52786-44878 SQLite defines three built-in collating
+  ** functions:
   */
   createCollation(db, "BINARY", SQLITE_UTF8, 0, binCollFunc, 0);
   createCollation(db, "BINARY", SQLITE_UTF16BE, 0, binCollFunc, 0);
   createCollation(db, "BINARY", SQLITE_UTF16LE, 0, binCollFunc, 0);
+  createCollation(db, "NOCASE", SQLITE_UTF8, 0, nocaseCollatingFunc, 0);
   createCollation(db, "RTRIM", SQLITE_UTF8, (void*)1, binCollFunc, 0);
   if( db->mallocFailed ){
     goto opendb_out;
   }
+  /* EVIDENCE-OF: R-08308-17224 The default collating function for all
+  ** strings is BINARY. 
+  */
   db->pDfltColl = sqlite3FindCollSeq(db, SQLITE_UTF8, "BINARY", 0);
   assert( db->pDfltColl!=0 );
 

SquiLu-ext/sqlite3.h

@@ -109,7 +109,7 @@ extern "C" {
 */
 #define SQLITE_VERSION        "3.8.8"
 #define SQLITE_VERSION_NUMBER 3008008
-#define SQLITE_SOURCE_ID      "2014-11-20 02:58:10 2d7c8da5f16e64eaa7b0c2d66898682ea3d102a0"
+#define SQLITE_SOURCE_ID      "2014-11-27 11:36:36 f095cde579e7417306e11b5c1d2dd90b6bb547d5"
 
 /*
 ** CAPI3REF: Run-Time Library Version Numbers
@@ -5941,34 +5941,34 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 **
 ** The SQLite source code contains multiple implementations
 ** of these mutex routines.  An appropriate implementation
-** is selected automatically at compile-time.  ^(The following
+** is selected automatically at compile-time.  The following
 ** implementations are available in the SQLite core:
 **
 ** <ul>
 ** <li>   SQLITE_MUTEX_PTHREADS
 ** <li>   SQLITE_MUTEX_W32
 ** <li>   SQLITE_MUTEX_NOOP
-** </ul>)^
+** </ul>
 **
-** ^The SQLITE_MUTEX_NOOP implementation is a set of routines
+** The SQLITE_MUTEX_NOOP implementation is a set of routines
 ** that does no real locking and is appropriate for use in
-** a single-threaded application.  ^The SQLITE_MUTEX_PTHREADS and
+** a single-threaded application.  The SQLITE_MUTEX_PTHREADS and
 ** SQLITE_MUTEX_W32 implementations are appropriate for use on Unix
 ** and Windows.
 **
-** ^(If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
+** If SQLite is compiled with the SQLITE_MUTEX_APPDEF preprocessor
 ** macro defined (with "-DSQLITE_MUTEX_APPDEF=1"), then no mutex
 ** implementation is included with the library. In this case the
 ** application must supply a custom mutex implementation using the
 ** [SQLITE_CONFIG_MUTEX] option of the sqlite3_config() function
 ** before calling sqlite3_initialize() or any other public sqlite3_
-** function that calls sqlite3_initialize().)^
+** function that calls sqlite3_initialize().
 **
 ** ^The sqlite3_mutex_alloc() routine allocates a new
-** mutex and returns a pointer to it. ^If it returns NULL
-** that means that a mutex could not be allocated.  ^SQLite
-** will unwind its stack and return an error.  ^(The argument
-** to sqlite3_mutex_alloc() is one of these integer constants:
+** mutex and returns a pointer to it. ^The sqlite3_mutex_alloc()
+** routine returns NULL if it is unable to allocate the requested
+** mutex.  The argument to sqlite3_mutex_alloc() must one of these
+** integer constants:
 **
 ** <ul>
 ** <li>  SQLITE_MUTEX_FAST
@@ -5981,7 +5981,8 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** <li>  SQLITE_MUTEX_STATIC_PMEM
 ** <li>  SQLITE_MUTEX_STATIC_APP1
 ** <li>  SQLITE_MUTEX_STATIC_APP2
-** </ul>)^
+** <li>  SQLITE_MUTEX_STATIC_APP3
+** </ul>
 **
 ** ^The first two constants (SQLITE_MUTEX_FAST and SQLITE_MUTEX_RECURSIVE)
 ** cause sqlite3_mutex_alloc() to create
@@ -5989,14 +5990,14 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** is used but not necessarily so when SQLITE_MUTEX_FAST is used.
 ** The mutex implementation does not need to make a distinction
 ** between SQLITE_MUTEX_RECURSIVE and SQLITE_MUTEX_FAST if it does
-** not want to.  ^SQLite will only request a recursive mutex in
-** cases where it really needs one.  ^If a faster non-recursive mutex
+** not want to.  SQLite will only request a recursive mutex in
+** cases where it really needs one.  If a faster non-recursive mutex
 ** implementation is available on the host platform, the mutex subsystem
 ** might return such a mutex in response to SQLITE_MUTEX_FAST.
 **
 ** ^The other allowed parameters to sqlite3_mutex_alloc() (anything other
 ** than SQLITE_MUTEX_FAST and SQLITE_MUTEX_RECURSIVE) each return
-** a pointer to a static preexisting mutex.  ^Six static mutexes are
+** a pointer to a static preexisting mutex.  ^Nine static mutexes are
 ** used by the current version of SQLite.  Future versions of SQLite
 ** may add additional static mutexes.  Static mutexes are for internal
 ** use by SQLite only.  Applications that use SQLite mutexes should
@@ -6005,16 +6006,13 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 **
 ** ^Note that if one of the dynamic mutex parameters (SQLITE_MUTEX_FAST
 ** or SQLITE_MUTEX_RECURSIVE) is used then sqlite3_mutex_alloc()
-** returns a different mutex on every call.  ^But for the static
+** returns a different mutex on every call.  ^For the static
 ** mutex types, the same mutex is returned on every call that has
 ** the same type number.
 **
 ** ^The sqlite3_mutex_free() routine deallocates a previously
-** allocated dynamic mutex.  ^SQLite is careful to deallocate every
-** dynamic mutex that it allocates.  The dynamic mutexes must not be in
-** use when they are deallocated.  Attempting to deallocate a static
-** mutex results in undefined behavior.  ^SQLite never deallocates
-** a static mutex.
+** allocated dynamic mutex.  Attempting to deallocate a static
+** mutex results in undefined behavior.
 **
 ** ^The sqlite3_mutex_enter() and sqlite3_mutex_try() routines attempt
 ** to enter a mutex.  ^If another thread is already within the mutex,
@@ -6022,23 +6020,21 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*);
 ** SQLITE_BUSY.  ^The sqlite3_mutex_try() interface returns [SQLITE_OK]
 ** upon successful entry.  ^(Mutexes created using
 ** SQLITE_MUTEX_RECURSIVE can be entered multiple times by the same thread.
-** In such cases the,
+** In such cases, the
 ** mutex must be exited an equal number of times before another thread
-** can enter.)^  ^(If the same thread tries to enter any other
-** kind of mutex more than once, the behavior is undefined.
-** SQLite will never exhibit
-** such behavior in its own use of mutexes.)^
+** can enter.)^  If the same thread tries to enter any mutex other
+** than an SQLITE_MUTEX_RECURSIVE more than once, the behavior is undefined.
 **
 ** ^(Some systems (for example, Windows 95) do not support the operation
 ** implemented by sqlite3_mutex_try().  On those systems, sqlite3_mutex_try()
-** will always return SQLITE_BUSY.  The SQLite core only ever uses
-** sqlite3_mutex_try() as an optimization so this is acceptable behavior.)^
+** will always return SQLITE_BUSY. The SQLite core only ever uses
+** sqlite3_mutex_try() as an optimization so this is acceptable 
+** behavior.)^
 **
 ** ^The sqlite3_mutex_leave() routine exits a mutex that was
-** previously entered by the same thread.   ^(The behavior
+** previously entered by the same thread.   The behavior
 ** is undefined if the mutex is not currently entered by the
-** calling thread or is not currently allocated.  SQLite will
-** never do either.)^
+** calling thread or is not currently allocated.
 **
 ** ^If the argument to sqlite3_mutex_enter(), sqlite3_mutex_try(), or
 ** sqlite3_mutex_leave() is a NULL pointer, then all three routines
@@ -6059,9 +6055,9 @@ SQLITE_API void sqlite3_mutex_leave(sqlite3_mutex*);
 ** used to allocate and use mutexes.
 **
 ** Usually, the default mutex implementations provided by SQLite are
-** sufficient, however the user has the option of substituting a custom
+** sufficient, however the application has the option of substituting a custom
 ** implementation for specialized deployments or systems for which SQLite
-** does not provide a suitable implementation. In this case, the user
+** does not provide a suitable implementation. In this case, the application
 ** creates and populates an instance of this structure to pass
 ** to sqlite3_config() along with the [SQLITE_CONFIG_MUTEX] option.
 ** Additionally, an instance of this structure can be used as an
@@ -6102,13 +6098,13 @@ SQLITE_API void sqlite3_mutex_leave(sqlite3_mutex*);
 ** (i.e. it is acceptable to provide an implementation that segfaults if
 ** it is passed a NULL pointer).
 **
-** The xMutexInit() method must be threadsafe.  ^It must be harmless to
+** The xMutexInit() method must be threadsafe.  It must be harmless to
 ** invoke xMutexInit() multiple times within the same process and without
 ** intervening calls to xMutexEnd().  Second and subsequent calls to
 ** xMutexInit() must be no-ops.
 **
-** ^xMutexInit() must not use SQLite memory allocation ([sqlite3_malloc()]
-** and its associates).  ^Similarly, xMutexAlloc() must not use SQLite memory
+** xMutexInit() must not use SQLite memory allocation ([sqlite3_malloc()]
+** and its associates).  Similarly, xMutexAlloc() must not use SQLite memory
 ** allocation for a static mutex.  ^However xMutexAlloc() may use SQLite
 ** memory allocation for a fast or recursive mutex.
 **
@@ -6134,29 +6130,29 @@ struct sqlite3_mutex_methods {
 ** CAPI3REF: Mutex Verification Routines
 **
 ** The sqlite3_mutex_held() and sqlite3_mutex_notheld() routines
-** are intended for use inside assert() statements.  ^The SQLite core
+** are intended for use inside assert() statements.  The SQLite core
 ** never uses these routines except inside an assert() and applications
-** are advised to follow the lead of the core.  ^The SQLite core only
+** are advised to follow the lead of the core.  The SQLite core only
 ** provides implementations for these routines when it is compiled
-** with the SQLITE_DEBUG flag.  ^External mutex implementations
+** with the SQLITE_DEBUG flag.  External mutex implementations
 ** are only required to provide these routines if SQLITE_DEBUG is
 ** defined and if NDEBUG is not defined.
 **
-** ^These routines should return true if the mutex in their argument
+** These routines should return true if the mutex in their argument
 ** is held or not held, respectively, by the calling thread.
 **
-** ^The implementation is not required to provide versions of these
+** The implementation is not required to provide versions of these
 ** routines that actually work. If the implementation does not provide working
 ** versions of these routines, it should at least provide stubs that always
 ** return true so that one does not get spurious assertion failures.
 **
-** ^If the argument to sqlite3_mutex_held() is a NULL pointer then
+** If the argument to sqlite3_mutex_held() is a NULL pointer then
 ** the routine should return 1.   This seems counter-intuitive since
 ** clearly the mutex cannot be held if it does not exist.  But
 ** the reason the mutex does not exist is because the build is not
 ** using mutexes.  And we do not want the assert() containing the
 ** call to sqlite3_mutex_held() to fail, so a non-zero return is
-** the appropriate thing to do.  ^The sqlite3_mutex_notheld()
+** the appropriate thing to do.  The sqlite3_mutex_notheld()
 ** interface should also return 1 when given a NULL pointer.
 */
 #ifndef NDEBUG