A recent commit broke the pattern editor popups for fin/replace and goto
line. The problem was cause by changes to the function that returns the
name of the currently focused subwindow using a function that only
updates when ImHex main window losses focus. The commit was aimed at
fixing evaluation of shortcuts in pattern data view and pattern editor
simultaneously but missed to fix some shortcuts like cut and paste.
The fix substitutes how the subwindow is first selected by using the
result of the subwindow selection used by imhex to insure that menus and
other ui components don't steal focus from views. The function that
returns the name of the current focused subwindow was changed to use
this value. This fixes both the window popups of pattern editor and all
the shortcut duplications.
Currently, interactions with the user interface, like changing providers, opening menus or even resizing windows, take the focus away from the main views. This PR resets focus to the child (if view has no children then the view's window itself is used) of the view that had focus before the interaction took place.
It was tested by interacting with menus, changing providers, using toolbar icons, using command palette, resizing windows or widgets of the view itself that are not children windows of the main view (e.g. running a pattern and having focus return to the pattern editor when evaluation ends. or using the icons in the hex editor) and also by selecting the main view itself. To clarify this last item, if you click on the view tab (not the provider tab but the view tab itself) the focus will be restored to the child that had focused before the tab was clicked
There is no attempt to recover the active status of widgets within the window but it can be easily recovered by clicking the Tab key once. Some views, like the pattern data view, can set other views to focus depending on where they are clicked.
The implementation saves the the child sub-window in a pointer of the view and is only changed if another child is given focus. Then various UI interactions are detected with care not to change focus while the interaction occurs. The end of the interaction is detected by checking if undesired items are the ones that have focus (these undesired values only occur when the UI interaction ends) and if they are, then the focus is restored to the window that is stored in the view pointer.
This is a trivial change which adds virtual methods to View, `onOpen()`
and `onClose()`, which are called when the view is opened or closed.
This information is already tracked inside the View, but not exposed via
the API. There is `didWindowJustOpen()` and `didWindowJustClose()`, but
these fetch and then reset the flag, so they can't be used more than
once in a frame (and are sometimes called by the frame, meaning the flag
has already been consumed by the time the View's draw callback gets
called).
The use case here is that I have a View which needs to do some work
every time it's shown.
Where one door opens, another one closes (said someone) and in this
situation i mean view, because there's no event for closing a view only
opening one, so please add this :c
### Problem description
Simply adds an **EventViewClosed** to match the pre-existing
**EventViewOpened**. Also fixes the issue where EventViewOpened wouldn't
trigger for windows that were popups (even if they are technically a
View.
### Implementation description
I mirrored the current functionality of EventViewOpened for closing.
---------
Co-authored-by: paxcut <53811119+paxcut@users.noreply.github.com>
### Problem description
This PR addresses issue #2013 that described a cluttered Event Manager.
This is a DX issue and should not impact the users whatsoever.
### Implementation description
The changes revolve around three main points:
1. the Event Manager (`event_manager.hpp`) was split into four
categories: GUI, Interaction, Lifecycle, and Provider, and two types:
Events, and Requests. This results in the following files:
- `events_gui.hpp`
- `events_interaction.hpp`
- `events_lifecycle.hpp`
- `events_provider.hpp`
- `requests_gui.hpp`
- `requests_interaction.hpp`
- `requests_lifecycle.hpp`
- `requests_provider.hpp`
2. Every event and request now has its own piece of documentation, with
a `@brief`, accompanied by a longer comment if needed, and gets its
`@param`s described.
3. The old `event_manager.hpp` import was removed and replaced by the
correct imports wherever needed, as to reduce spread of those files only
to where they are truly useful.
### Additional things
The commits have been split into (chrono-)logical steps:
- `feat`: split the Event Manager, and replace the imports
- `refactor`, `chore`: make various small changes to match the required
structure
- `docs`: add documentation for events and requests
Hopefully, this will help to review the PR.
*Note: Beware of very long rebuild times in between the commits, use
them sparingly! The Actions will ensure this PR builds anyways*
Closes#2013
---------
Signed-off-by: BioTheWolff <47079795+BioTheWolff@users.noreply.github.com>
Co-authored-by: Nik <werwolv98@gmail.com>
This PR updates imgui and its dependencies from the last time, ~1 year
ago
(deabacbd50)
Commits will be refactored before merging
Things you might ask :
- why did you remove `ImGui_ImplGlfw_SetBorderlessWindowMode()` ?
-> Where is it used ? The only usage of it I see is commented
(cb9a3b1f55/lib/external/imgui/source/imgui_impl_glfw.cpp (L757))
- why did you remove the implot anti aliasing flag ?
-> They.. seem to have removed it altogether ?
https://github.com/epezent/implot/issues/479
* patterns: Initial work to refactor pattern language to use smart pointers
* patterns: Fixed remaining issues, moved patterns to unique files
* sys: Added missing includes for macOS
