Does not work with rotation. It might be a good idea to replace this whole system with one that uses a transformation matrix, to handle rotation and skewing more easily in the future. I'm not a huge fan of this current code.
Related to the PoolVectorArray locking issue (#331). This still doesn't fix the issue itself, but at least Pixelorama does not crash anymore. It is still possible, though, to fail to load a guide when opening a project. Opening it again may solve the issue.
* Basic move tool
* Added marching ants effect on the selection borders
* Rename SelectionRectangle to SelectionShape, make it have non-rectangular shape and multiple SelectionShapes can exist
- Create multiple selection rectangles
- Merge them together if they intersect
- Move the selections (without contents as of right now)
- Gizmos are being drawn but they are not functional yet
Code is very ugly.
* Sort vectors counter-clockwise to be used as polygon borders
I did this, no idea if it works properly, probably won't be used but I thought I'd keep it saved somewhere
* More experiments I may or may not need
Trying to generate a polygon from the individual selected pixels
* Change default rectangle select behavior and ability to clip polygons using Control
* Fix rectangle selection clipping
* Split polygon into two with selection subtracting
* Move selection with contents with the move tool
Code is still a mess, don't bother looking.
* Move some methods from SelectionShape.gd to Selection.gd
The purpose of this is to generalize some selection code, so that it applies to all polygons, the entire selection. More will follow.
* UndoRedo for border moving
Nothing else in the selections system works properly in UndoRedo right now. Needs:
- UR support for creating selections
- UR support for modifying selections (merging and cutting selections together)
- UR support for removing selection
- UR support for moving content
& for all the rest of the remaining features
* Moving all of the selection shape logic to Selection.gd
Handle all of the polygons there instead of having them as individual nodes. Should be easier to handle undo/redo this way. This commit probably breaks move tool + selection tool and undo/redo. Code is still a mess. For your sanity, I hope you are not reading this. I promise I will clean up.
* Move tool works again
Buggy and messy, of course.
* Remove unneeded code and restore selection move undoredo logic
* Made Selection.gd have one big preview_image for when moving content, instead of each polygon having its own image
Could be further optimized for some specific cases. We could also remove selected_pixels from SelectionPolygon.
* UndoRedo support for creating, deleting, merging and clipping selections
UndoRedo support for moving content not added in this commit. Should work but needs more testing. This PR also removes selected_pixels from the SelectionPolygon class.
* Confirm & cancel selection movement, should support undoredo properly too
Press Enter or do any editing to confirm movement, Escape to cancel. I will most likely add UI buttons for confirm and cancel too.
* Mirror View affects selection
* Restore Cut, Copy, Paste and Clear Selection
Pasting now no longer requires a pre-existing selection and instead copies the selections themselves too.
* Created a new Select menu, which has Select All and Clear Selection as options
Clear Selection now also confirms content moving. TopMenuContainer code has changed to no longer rely on Global for the menu buttons.
* Draw gizmos as rectangles
No functionality yet. They may need to be turned to nodes, so that they can easily resize based on zoom level and check for mouse enter/exit events.
* Made gizmos get drawn in the sides and corners of the big bounding rectangle instead of individual selection parts
Still no functionality yet.
* Restore label text
* Minor optimization when clipping selections
This will execute the for loop less times
* Made a Gizmo class, cursor change on hover, has_focus = false on mouse click
Now I should actually make them resize when dragged, aye?
* Very basic gizmo resizing, still a WIP, does not work properly yet
* Start replacing the array of selected pixels with a BitMap
This should optimize the selection making a lot, and it also allows for easy border drawing without having to deal with polygons, thanks to the MarchingAntsOutline.shader
Still commit is still a WIP, image effects and brushes may not work properly yet.
Because the BitMap has a fixed size, the size of the project, moving the selection outside of canvas boundaries has proven to be a bit tricky. I did implement a hacky way of handling it, but it may be buggy and problematic. I'm still unsure whether this is the best way to handle the situation.
* Selection works with mirror view
* Draw a black rectangle when the user is making a rectangular selection
After they release the mouse, the black rectangle becomes the selection
* Make Selection.gd update when undoing/redoing
* Fix brushes not working properly with non-rectangular selections
* Added invert selection
* Cache has_selection as a variable for a speedup
* Fix conflict issues with the shape tools
* Made the bitmap image squared so the marching ants effect will be the same on both dimensions
There may be a better way to fix the issue, perhaps inside the shader itself.
* Some optimizations to call selection_bitmap_changed() less times
* Restored almost all of the image effects
Left to do:
- Change gradient's behavior. Unsure of how it will work with non-rectangular selections yet, but it should not generate pixels outside of the selection.
- Restore rotation
- Resize bitmap on image resize
- Remove the `pixels` array from the ImageEffect
* Fix Selection.gd not updating when changing project
* Resize the selection bitmap along with image resize
* Restored rotation's old behavior and finally got rid of the selected_pixels array
The rotation does not yet work properly with selections, but at least it now "works".
* Resize selection too when using gizmos
Left to do for gizmos:
- Proper cancel transformation
- Begin transformation
(currently named move_content_start but it should be renamed to something more general) when resizing gizmos
- Keep the original image and selection in memory and resize them. Meaning, gizmos should not resize the already resized data, but only resize the original. This is less destructive as there is no danger of data loss.
- Always resize on InputEventMouseMotion. This is going to be worse for performance, but it will look better for the user.
* Image and bitmap resizing now uses the original data and begin transformation on gizmo click
No matter how many times the user resizes on the current transformation, the original data will not be lost until they either confirm or cancel, so there is no data loss before confirmation/cancel.
* Cancel transformation now works properly when the selection has been resized
* Made gizmos resize on mouse motion, fix issues with negative bounding rectangle and when combined with the move tool
* Resizing can now get out of positive bounds, clearing and inverting now gets limited to the canvas bounds
Resizing currently does not work properly with negative (left & up) canvas boundaries
* Flip image when resizing and the bounding rectangle gets flipped
* Call move_content_confirm() when inverting selection
* Attempt to implement selection resizing that goes outside of the canvas boundaries (not working properly yet)
* Flip selection when resizing to negative bounding rectangle sizes
And fix preview_image vertical flipping
* Fix rotation so that it works (almost) properly with selections
Rotation algorithms now accept and only work with a given image, and the pivot has been added as a parameter
* Experimental gizmo rotation - does not work properly yet
Transforming the selection outside of the canvas is still broken.
* Fix some issues with moving selection out of canvas bounds
* Fix more issues with selection getting resized outside of canvas bounds
* Update marching ants effect properly when switching between projects
And make sure the frequency of the marching ants effect always looks roughly the same on all project sizes
* Made the rotation gizmo part of the gizmos array and resize them based on camera zoom
* Remove unneeded parameter from move_bitmap_values()
* Remove more unneeded parameters
* Move the selection only if the cursor is above it and neither shift nor control are currently pressed
* Gradient generation now works on non-rectangular selections
Although this behavior might not be the intended one
* Copy/paste marching ants effect offset
Useful for when the selection is in negative coords
* Fix issue with clear selection & UndoRedo
* Restore the ability to move selection when it's in negative coords
* Made the marching ants offset a Project variable
This fixes the issue of project switching and keeping the previous project's offset. Again, this is only relevant for when the selection is in negative coords.
* Made the "from current selection" palette preset work with the new selection system
* Fix out of bounds error when using the rectangular select tool on negative coords
* Some code cleanup
* Comment out the rotation gizmo for now, since it does not work properly
* Update marching ants shader params and gizmo sizes when the bitmap changes
* Move some methods around in Selection.gd
A performance boost for when opening large images. Also fixed an issue with pixels being selected outside of the canvas boundaries, when the selection rectangle was outside the canvas and its size got reduced.
* Cache tile mode rects
* Make TransparentChecker don't calculate tile mode rect on its own
* Minor fixes
- typo
- formatting
- moved unlock() call to the method where matching lock() was called
* Move get_tile_mode_rect() method to the Project class
* Refactored isometric grid
Now it should be drawn properly in the target rect.
Settings (in pixels):
- cell's AABB size,
- offset from the canvas origin.
* Draw grid only over the Canvas (when in tiling mode)
* Replace some magic numbers with enums.
It's too easy to break something when adding something new in here. Should be a little harder now.
* Added Pixel Grid.
- Pixel grid is shown only when it's enabled and camera is zoomed close enough.
- Settings: pixel_grid_show_at_zoom (as a percentage because that's what's shown in the settings panel), pixel_grid_color. Default values might need changing.
- To distinguish between grid and pixel grid default settings for grid width, grid height are changed.
- Now both grid and pixel grid are drawn above (after) tile mode. Grid is drawn above (after) pixel grid.
It's Cel's property and thus it changes when frame_changed is called (also layer_changed calls frame_changed on its own).
Co-authored-by: kleonc <kleonc@users.noreply.github.com>
Instead of having a frame_duration[] array in the Project class. This makes the code much more readable and understandable, using less lines of code and, by avoiding an extra array, we also avoid potential out-of-bound array crashes.
The functionality for the user has not changed, and by testing so far I didn't find any issues.
Export.was_reported was keeping its value when switching projects, instead of being project-specific. Also fixed issue where imported images were not remembering the directory_path and file_name when switching projects.
It now first attempts to serialize the project data and turn them to JSON, and then opens the file. This is a lto safer because, if a crash happens in the serialization, the file is not already open, and therefore it does not get corrupt. Earlier, if a crash happened, the file would be empty. This meant that if that file existed previously, all of the data would be lost.
* Sync for my local files to my repository
* This is frame properties update, it works but I can't be stored while
saving or loading and that makes crashes just the beggining :)
* I forgot this files :P
* Frame Properties update.
* Updating frame properties
* Update Translations.pot
* Changes to CanvasPreview and CelButton change
Instead of drawing individual lines, execute draw_isometric_tile() multiple times. What this method does is basically draws a single isometric tile, which then gets repeated for the entire canvas size. Also moved all the grid logic to a new script, which gets updated less often that Canvas.gd, for optimization purposes.
Works better than the previous method, although it might be a bit slower, and currently does not work well if the canvas size is not divisible by the grid size. For now, it only gets affected by the grid width. Changing grid height has no effect.
