This document provides an overview of the changes between Terminal.Gui v1 and v2. It is intended to help developers migrate their applications from v1 to v2.
For detailed breaking change documentation check out this Discussion: https://github.com/gui-cs/Terminal.Gui/discussions/2448
In v1, View and most sub-classes, had multiple constructors that took a variety of parameters. In v2, the constructors have been replaced with initializers. This change was made to simplify the API and make it easier to use. In addition, the v1 constructors drove a false (and needlessly complex) distinction between "Absoulte" and "Computed" layout. In v2, the layout system is much simpler and more intuitive.
Replace the constructor calls with initializer calls.
- var myView = new View (new Rect (10, 10, 40, 10));
+ var myView = new View { X = 10, Y = 10, Width = 40, Height = 10 };
Terminal.Gui v2 now supports 24-bit color by default. This means that the colors you use in your application will be more accurate and vibrant. If you are using custom colors in your application, you may need to update them to use the new 24-bit color format.
The Attribute class has been simplified. Color names now match the ANSI standard ('Brown' is now called 'Yellow')
Static class Attribute.Make
has been removed. Use constructor instead
- var c = Attribute.Make(Color.BrightMagenta, Color.Blue);
+ var c = new Attribute(Color.BrightMagenta, Color.Blue);
- var c = Color.Brown;
+ var c = Color.Yellow;
Rect
-> Rectangle
Point
-> Point
Size
-> Size
Rect
with Rectangle
NStack.string
has been removed. Use System.Rune
instead.See Unicode for details.
Replace using
statements with the System.Text
namespace
- using NStack;
+ using System.Text;
Anywhere you have an implicit cast from char
to Rune
, replace with a constructor call
- myView.AddRune(col, row, '▄');
+ myView.AddRune(col, row, new Rune('▄'));
When measuring the screen space taken up by a Rune
use GetColumns()
- Rune.ColumnWidth(rune);
+ rune.GetColumns();
When measuring the screen space taken up by a string
you can use the extension method GetColumns()
- myString.Sum(c=>Rune.ColumnWidth(c));
+ myString.GetColumns();
In v1, View was derived from Responder
which supported IDisposable
. In v2, Responder
has been removed and View is the base-class supporting IDisposable
.
In v1, Application.Init automatically created a toplevel view and set Applicaton.Top. In v2, Application.Init no longer automatically creates a toplevel or sets Applicaton.Top; app developers must explicitly create the toplevel view and pass it to Appliation.Run (or use Application.Run<myTopLevel>
). Developers are responsible for calling Dispose
on any toplevel they create before exiting.
Responder
with ViewApplication.Init
automatically created a toplevel view and set Applicaton.Top
.Application.Init
automatically disposed of the toplevel view when the application exited.Pos.At
, was renamed to Pos.Absolute for consistency.Dim.Sized
, was renamed to Dim.Absolute for consistency.Pos.Pos
-> Pos
.Dim.Dim
-> Dim
.Pos.At
-> Pos.Absolute
Dim.Sized
-> Dim.Absolute
Dim.Anchor
-> Dim.GetAnchor
Pos.Anchor
-> Pos.GetAnchor
In v2, the layout system has been improved to make it easier to create complex user interfaces. If you are using custom layouts in your application, you may need to update them to use the new layout system.
Absoulte Layout
and Computed Layout
has been removed, as has the LayoutStyle
enum. v1 drew a false distinction between these styles.Frame
property is of type Rectangle
.Viewport
property represents the visible area of the view in its own coordinate system. The Viewport
property is of type Rectangle
.ScrollView
and ScrollBarView
in v1. See more below.Bounds
-> Viewport
LayoutStyle
.Bounds
to Viewport
. The Location
property of Bounds
can now have non-zero values.Bounds.Location
was always Point.Empty
.Bounds
to refer to the size of the view's content. Use GetContentSize()
instead.Bounds.Size
was the same as Frame.Size
. Frame.Size
defines the size of the view in the superview's coordinate system, while Viewport.Size
defines the visible area of the view in its own coordinate system.Viewport
. View subclasses should not implement their own concept of padding or margins but leverage these Adornments
instead.Viewport
not the Frame
.View.AutoSize
has been removed. Use Dim.Auto for width or height instead.In v1, View.AutoSize
was used to size a view to its Text
. In v2, View.AutoSize
has been removed. Use Dim.Auto for width or height instead.
View.AutoSize = true
with View.Width = Dim.Auto
or View.Height = Dim.Auto
as needed. See the DimAuto Deep Dive for more information.In v2, the Border
, Margin
, and Padding
properties have been added to all views. This simplifies view development and enables a sophisticated look and feel. If you are using custom borders, margins, or padding in your application, you may need to update them to use the new properties.
View.Border
is now of type Adornment. View.BorderStyle is provided as a convenience property to set the border style (myView.BorderStyle = LineStyle.Double
).In v1, scrolling was enabled by using ScrollView
or ScrollBarView
. In v2, the base View class supports scrolling inherently. The area of a view visible to the user at a given moment was previously a rectangle called Bounds
. Bounds.Location
was always Point.Empty
. In v2 the visible area is a rectangle called Viewport
which is a protal into the Views content, which can be bigger (or smaller) than the area visible to the user. Causing a view to scroll is as simple as changing View.Viewport.Location
. The View's content described by View.GetContentSize(). See Layout for details.
ScrollView
with View and use Viewport
and View.GetContentSize() to control scrolling.Bounds.Location
was always Point.Empty
.Bounds
to refer to the size of the view's content. Use View.GetContentSize() instead.Bounds.Size
was the same as Frame.Size
. Frame.Size
defines the size of the view in the superview's coordinate system, while Viewport.Size
defines the visible area of the view in its own coordinate system.The API for handling keyboard input is significantly improved. See Keyboard API.
KeyEvent
struct and provides a platform-independent abstraction for common keyboard operations. It is used for processing keyboard input and raising keyboard events. This class provides a high-level abstraction with helper methods and properties for common keyboard operations. Use this class instead of the low-level KeyCode enum when possible. See Key for more details.View.AddCommand()
. Use the View.Keybindings to configure the key bindings.Toplevel
is now Esc
(it was previously Ctrl+Q
).KeyEvent
with Key
Command
s.OnKeyPressed
etc...Ctrl+Q
was hard-coded as the "quit key", replace with Application.QuitKey
.The API for mouse input is now internally consistent and easier to use.
MouseEventEventArgs
.View.WantContinousButtonPresses = true
to have their Command.Accept command be invoked repeatedly as the user holds a mouse button down on the view.Viewport
not the Screen
.MouseEventEventArgs
with MouseEvent
View.WantContinousButtonPresses = true
to have the Command.Accept command be invoked repeatedly as the user holds a mouse button down on the view.Screen
.Cursor
, Focus
, TabStop
etc...The cursor and focus system has been redesigned in v2 to be more consistent and easier to use. If you are using custom cursor or focus logic in your application, you may need to update it to use the new system.
In v1, whether the cursor (the flashing caret) was visible or not was controlled by View.CursorVisibility
which was an enum extracted from Ncruses/Terminfo. It only works in some cases on Linux, and only partially with WindowsDriver
. The position of the cursor was the same as ConsoleDriver.Row
/Col
and determined by the last call to ConsoleDriver.Move
. View.PositionCursor()
could be overridden by views to cause Application
to call ConsoleDriver.Move
on behalf of the app and to manage setting CursorVisiblity
. This API was confusing and bug-prone.
In v2, the API is (NOT YET IMPLEMENTED) simplified. A view simply reports the style of cursor it wants and the Viewport-relative location:
public Point? CursorPosition
null
the cursor is not visible{}
the cursor is visible at the Point
.public event EventHandler<LocationChangedEventArgs>? CursorPositionChanged
public int? CursorStyle
null
the default cursor style is used.{}
specifies the style of cursor. See cursor.md for more.Application
now has APIs for querying available cursor styles.ConsoleDriver
are no longer available to applications.null
to hide the cursor.OnEnter
and OnLeave
that explicitly change the cursor.See navigation.md for more details. See also Keyboard where HotKey is covered more deeply...
HasFocus
as a get-only property. In v2, view.HasFocus
can be set as well. Setting to true
is equivalent to calling view.SetFocus
. Setting to false
is equivalent to calling view.SuperView.AdvanceFocus
(which might not actually cause view
to stop having focus).super.Add (view)
where view.CanFocus == true
caused all views up the hierarchy (all SuperViews) to get CanFocus
set to true
as well. In v2, developers need to explicitly set CanFocus
for any view in the view-hierarchy where focus is desired. This simplifies the implementation and removes confusing automatic behavior.view.CanFocus == true
, Add
would automatically set TabStop
. In v2, the automatic setting of TabStop
in Add
is retained because it is not overly complex to do so and is a nice convenience for developers to not have to set both Tabstop
and CanFocus
. Note v2 does NOT automatically change CanFocus
if TabStop
is changed.view.TabStop
now describes the behavior of a view in the focus-chain. the TabBehavior
enum includes NoStop
(the view may be focusable, but not via next/prev keyboard nav), TabStop
(the view may be focusable, and NextTabStop
/PrevTabStop
keyboard nav will stop), TabGroup
(the view may be focusable, and NextTabGroup
/PrevTabGroup
keyboard nav will stop).View.Focused
property was a cache of which view in SubViews/TabIndexes
had HasFocus == true
. There was a lot of logic for keeping this property in sync. In v2, View.Focused
is a get-only, computed property.View.MostFocused
property recursed down the subview-hierarchy on each get. In addition, because only one View in an application can be the "most focused", it doesn't make sense for this property to be on every View. In v2, this API is removed. Use Application.Navigation.GetFocused()
instead.View.EnsureFocus
/FocusNext
/FocusPrev
/FocusFirst
/FocusLast
are replaced in v2 with these APIs that accomplish the same thing, more simply.
public bool AdvanceFocus (NavigationDirection direction, TabBehavior? behavior)
public bool FocusDeepest (NavigationDirection direction, TabBehavior? behavior)
View.OnEnter/Enter
and View.OnLeave/Leave
virtual methods/events could be used to notify that a view had gained or lost focus, but had confusing semantics around what it mean to override (requiring calling base
) and bug-ridden behavior on what the return values signified. The "Enter" and "Leave" terminology was confusing. In v2, View.OnHasFocusChanging/HasFocusChanging
and View.OnHasFocusChanged/HasFocusChanged
replace View.OnEnter/Enter
and View.OnLeave/Leave
. These virtual methods/events follow standard Terminal.Gui event patterns. The View.OnHasFocusChanging/HasFocusChanging
event supports being cancelled.Mdi
views included a large amount of complex code (in Toplevel
and Application
) for dealing with navigation across overlapped Views. This has all been radically simplified in v2. Any View can work in an "overlapped" or "tiled" way. See navigation.md for more details.View.TabIndex
and View.TabIndexes
have been removed. Change the order of the views in View.Subviews
to change the navigation order.In v2, HotKey
s can be used to navigate across the entire application view-hierarchy. They work independently of Focus
. This enables a user to navigate across a complex UI of nested subviews if needed (even in overlapped scenarios). An example use-case is the AllViewsTester
scenario.
In v2, unlike v1, multiple Views in an application (even within the same SuperView) can have the same HotKey
. Each press of the HotKey
will invoke the next HotKey
across the View hierarchy (NOT IMPLEMENTED YET)*
In v1, the keys used for navigation were both hard-coded and configurable, but in an inconsistent way. Tab
and Shift+Tab
worked consistently for navigating between Subviews, but were not configurable. Ctrl+Tab
and Ctrl+Shift+Tab
navigated across Overlapped
views and had configurable "alternate" versions (Ctrl+PageDown
and Ctrl+PageUp
).
In v2, this is made consistent and configurable:
Application.NextTabStopKey
(Key.Tab
) - Navigates to the next subview that is a TabStop
(see below). If there is no next, the first subview that is a TabStop
will gain focus.Application.PrevTabStopKey
(Key.Tab.WithShift
) - Opposite of Application.NextTabStopKey
.Key.CursorRight
- Operates identically to Application.NextTabStopKey
.Key.CursorDown
- Operates identically to Application.NextTabStopKey
.Key.CursorLeft
- Operates identically to Application.PrevTabStopKey
.Key.CursorUp
- Operates identically to Application.PrevTabStopKey
.Application.NextTabGroupKey
(Key.F6
) - Navigates to the next view in the view-hierarchy that is a TabGroup
(see below). If there is no next, the first view that is a TabGroup
will gain focus.Application.PrevTabGroupKey
(Key.F6.WithShift
) - Opposite of Application.NextTabGroupKey
.F6
was chosen to match Windows
These keys are all registered as KeyBindingScope.Application
key bindings by Application
. Because application-scoped key bindings have the lowest priority, Views can override the behaviors of these keys (e.g. TextView
overrides Key.Tab
by default, enabling the user to enter \t
into text). The AllViews_AtLeastOneNavKey_Leaves
unit test ensures all built-in Views have at least one of the above keys that can advance.
...
object sender, EventArgs args
signaturePreviously events in Terminal.Gui used a mixture of Action
(no arguments), Action<string>
(or other raw datatype) and Action<EventArgs>
. Now all events use the EventHandler<EventArgs>
standard .net design pattern.
For example, event Action
TimeoutAddedhas become
event EventHandler TimeoutAdded`
This change was made for the following reasons:
For example:
public class TimeoutEventArgs : EventArgs {
/// <summary>
/// Gets the <see cref="DateTime.Ticks"/> in UTC time when the
/// <see cref="Timeout"/> will next execute after.
/// </summary>
public long Ticks { get; }
[...]
}
If you previously had a lamda expression, you can simply add the extra arguments:
- btnLogin.Clicked += () => { /*do something*/ };
+ btnLogin.Clicked += (s,e) => { /*do something*/ };
If you have used a named method instead of a lamda you will need to update the signature e.g.
- private void MyButton_Clicked ()
+ private void MyButton_Clicked (object sender, EventArgs e)
ReDraw
is now Draw
ReDraw
with Draw
Viewport
not the Frame
.All public classes that were previously nested classes are now in the root namespace as their own classes.
Replace references to to nested types with the new standalone version
- var myTab = new TabView.Tab();
+ var myTab = new Tab();
In v1, both TextAlignment
and VerticalTextAlignment
enums were used to align text in views. In v2, these enums have been replaced with the Alignment enum. The View.TextAlignment property controls horizontal text alignment and the View.VerticalTextAlignment property controls vertical text alignment.
v2 now supports Pos.Align which enables views to be easily aligned within their Superview.
The Aligner class makes it easy to align elements (text, Views, etc...) within a container.
VerticalAlignment.Middle
is now Alignment.Center.StatusBar
- StatusItem
is replaced by Shortcut
StatusBar has been upgraded to utilize Shortcut.
- var statusBar = new StatusBar (
- new StatusItem []
- {
- new (
- Application.QuitKey,
- $"{Application.QuitKey} to Quit",
- () => Quit ()
- )
- }
- );
+ var statusBar = new StatusBar (new Shortcut [] { new (Application.QuitKey, "Quit", Quit) });
CheckBox
- API renamed and simplifiedIn v1 CheckBox
used bool?
to represent the 3 states. To support consistent behavior for the Accept
event, CheckBox
was refactored to use the new CheckState
enum instead of bool?
.
Additionally the Toggle
event was renamed CheckStateChanging
and made cancelable. The Toggle
method was renamed to AdvanceCheckState
.
-var cb = new CheckBox ("_Checkbox", true); {
- X = Pos.Right (label) + 1,
- Y = Pos.Top (label) + 2
- };
- cb.Toggled += (e) => {
- };
- cb.Toggle ();
+
+var cb = new CheckBox ()
+{
+ Title = "_Checkbox",
+ CheckState = CheckState.Checked
+}
+cb.CheckStateChanging += (s, e) =>
+{
+ e.Cancel = preventChange;
+}
+preventChange = false;
+cb.AdvanceCheckState ();