Essentially we are going to remove calls to SetItemAllowOverlap() and standardize the fact that only 'HoveredId == id' test from it is performed.
# Conflicts:
# imgui_widgets.cpp
Tagging #6344#6003#2694#1688 as it relates to scoring, however this doesn't technically fix any of them fully yet.
But e.g. once we restore axial path for #2694 this commit will allow going back and forth to initial location.
+ Demo: use BeginDisabled() block in BackendFlags section.
I'd still consider this undefined behavior as some combination may not work properly, but let's fix things while we can as we encounter them.
(There are two additional unhandled flags that only affect padding: '-' and ' '. Formatting flags don't make sense in a SliderInt's format string, so I've omitted them)
Altered ItemAdd() clipping rule to keep previous-frame ActiveId unclipped to support that late commit.
Also, MarkItemEdited() may in theory need to do:
if (g.ActiveIdPreviousFrame == id)
g.ActiveIdPreviousFrameHasBeenEditedBefore = true;
But this should already be set so not adding now.
This commit is a preparation toward adding ImGui apis with explicit context
and making ImGui applications being able to use multiple context at the same time
whatever their concurrency model.
This commit modifies ImGuiInputTextCallback, ImGuiListClipper and ImGuiStackSize so those classes do not to depend on GImGui context anymore.
About ImGuiInputTextCallback:
- ImGuiInputTextCallback depends on ImGuiContext because it has a
`InsertChars` method adding character to `g.InputTextState`
- To make ImGuiInputTextCallback aware of which context to use, the
appropriate context is given as argument of ImGuiInputTextCallback
constructor.
About ImGuiListClipper:
- ImGuiListClipper apply to a context through its `Begin`, `End`, and `Step`
method.
- To make ImGuiListClipper aware of which context to use, the
appropriate context is given as argument of ImGuiListClipper
constructor.
- Since the behavior is different than previously the class has been
renamed ImGuiListClipperEx
- In order to preserve backward compatibility, a subclass of ImGuiListClipperEx
named ImGuiListClipper has been defined and forward the implicit context
to ImGuiListClipperEx parent.
About ImGuiTextFilter:
- ImGuiTextFilter depends on the implicit context because the Draw(..)
method call ImGui::InputText(...)
- Instead from that commit the Draw(...) method takes an explicit context
as first argument
- Since the behavior is different than previously the class has been
renamed ImGuiTextFilterEx
- In order to preserve backward compatibility, a subclass of ImGuiTextFilterEx
named ImGuiTextFilter has been defined. This subclass has a draw method
override which and forward the implicit context to the parent class Draw(...)
About ImGuiStackSizes:
- ImGuiStackSizes was depending on ImGuiContext because of its
`SetToCurrentState` and `CompareWithCurrentState` method
- ImGuiStackSizes is an helper object use
for comparing state of context. It does not necessarily need to
compare the same context. For that reason, as opposed to previous
classes it takes the context it wants to compare to as argument of
its method.
- For this occasion `SetToCurrentState` and `CompareWithCurrentState`
have been renamed `SetToContextState` and `CompareWithContextState`
to match the new method signature.
ImGuiListClipper
ImGuiInputTextCallbackData
In particular:
- imgui.cpp : move UpdateInputEvents() higher in NewFrame() to match docking + update RenderMouseCursor() to match.
- imgui_draw.cpp: ImDrawList::_ResetForNewFrame() change from c807192ab
- Backends: SDL2. Add MouseWindowID + change SDL_CaptureMouse() test to match docking branch. Not strictly necessary but aimed at reducing drift because we go on and fork this file.
+ moved responsability of checking valid names to TabBarGetTabName() to simplify both branches.
Needlessly introduced in baae057a from WIP tables branch at the time, ended up unused by Tables.
A comment "I find it counter intuitive that hovering supersedes activation." in #3516 led me to this however this is not the cause of said issue.
This commit is a preparation toward adding ImGui apis with explicit context
and making ImGui applications being able to use multiple context at the same time
whatever their concurrency model.
--
Prior to this commit ImGuiInputTextState::OnKeyPressed was depending on the
global context to know which font and font size to use, and if it should
follow MacOSX behaviors or not (c.f ConfigMacOSXBehaviors).
Instead of using the global context, this commit store the context as
attribute of ImGuiInputTextState. Since this state is forwarded to most
of text edit related function, it possible to access font, font size and
ConfigMacOSXBehaviors from everywhere.
NOTE: I have noticed a bug prior to that commit: if the font or font size
change while editing the same widget, the ImGuiInputTextState become invalid
and there is no code to handle this invalidation. Fixing this bug is out
of scope of current pull request.
# Conflicts:
# imgui_internal.h
- Added SetKeyOwner(), SetItemKeyOwner(), TestKeyOwner().
- Added new IsKeyXXX IsMouseXXX functions with ImGuID owner_id and flags.
- Obsoleted SetItemUsingMouseWheel(). (#2891)
- Removed IsKeyPresseedEx() which was a recent internal addition 2022-07-08 deemed to be temporary exactly for this.
- Added ImGuiButtonFlags_NoSetKeyOwner, ImGuiButtonFlags_NoTestKeyOwner
- Added ImGuiSelectableFlags_NoSetKeyOwner.
- Added ImGuiInputFlags_LockThisFrame, ImGuiInputFlags_LockUntilRelease for for SetKeyOwner(), SetItemKeyOwner().
- Added ImGuiInputFlags_CondXXX values for SetItemKeyOwner().
Intended as part of work for input routing + blind menu processing shortcuts. Some of this commit will be stripped by next commit.
Intent was to sort windows along with focus scope to build a hierarchy, but for our needs we'd need a persistant one, so scrapping the idea. Not squashing this with next commit to keep a bit of history for future references.
This put an extra flag check in ItemAdd() but essentially reduce inconsistency with windows decorations not using this. Useful for debugging.
It however buries the info/blurs the line about what it means to not use ItemAdd() since they are now doing it much less.
+ TabBar: starts displaying the unsaved document marker with a frame delay to match how close button is processed, otherwise the transition would be noticeable.
Replace BeginMenu/MenuItem swapping g.NavWindow with a more adequate ImGuiItemFlags_NoWindowHoverableCheck.
Expecting more subtle issues to stem from this.
Note that NoWindowHoverableCheck is not supported by IsItemHovered() but then IsItemHovered() on BeginMenu() never worked: fix should be easy in BeginMenu() + add test is IsItemHovered(), will do later
As it turns out, functions like IsItemHovered() won't work on an open BeginMenu() because LastItemData is overriden by BeginPopup(). Probably an easy fix.
Changed signature of GetKeyChordName() to use ImGuiKeyChord.
Additionally SetActiveIdUsingAllKeyboardKeys() doesn't set ImGuiKey_ModXXX but we never need/use those and the system will be changed in upcoming commits.
Amend bda2cde6
Fixes the case where following menu hover sequence results in incorrect RestoreNavWindow (previously SourceWindow) pointing to opened sub-menu:
1. Hover "Menu -> Options" - Open ##Menu_01 window
2. Hover "Menu -> Colors" - SourceWindow incorrectly points to ##Menu_01 window
Curiously very old, amend 83efdce and bdbb2b2. Using stb_ functions updated ->CurLenA without updating ->TextA, leading to `buf_display_end = buf_display + state->CurLenA;` in the display.
Since f3ab5e62 they are 1 case out of 4 which didn't apply back to ->TextA and this is essentially the one where we ensure appliance. Another solution would be to alter the lower display code, but applying to TextA makes things more consistent.
Note that this widen the small gap between polling g.HoveredID and using IsItemHovered() the later does various filtering.
Added IsItemHovered(ImGuiHoveredFlags_NoNavOverride) to disable that specific state redirect/filter.
Side-effect: fix EndGroup() use of combining HoveredId values when gamepad/keyboard nav is active. Unlikely to have user-visible side effect since IsItemHovered() would have filtered out anyway.
Side-effect: fix IsAnyItemHovered() when gamepad/keyboard is active (but this wasn't the primary intent of this change).
Side-effect: fix using SetItemUsingMouseWheel() while hovering an item and gamepad/keyboard is active. (#2891)
