For the move tool, only store the difference vector in undo/redo

Instead of storing copies of entire images. This is a bit slower on the CPU side, mostly on larger canvases, but it is a massive RAM optimization. Perhaps it could be move optimal to use a shader here.

src/Autoload/Global.gd

@@ -856,7 +856,10 @@ func path_join_array(basepaths: PackedStringArray, subpath: String) -> PackedStr
 	return res
 
 
-func undo_redo_compress_images(redo_data: Dictionary, undo_data: Dictionary, project := current_project) -> void:
+## Used by undo/redo operations to store compressed images in memory.
+func undo_redo_compress_images(
+	redo_data: Dictionary, undo_data: Dictionary, project := current_project
+) -> void:
 	for image in redo_data:
 		if not image is Image:
 			continue
@@ -873,8 +876,8 @@ func undo_redo_compress_images(redo_data: Dictionary, undo_data: Dictionary, pro
 		)
 
 
-## Decompresses the [param compressed_image_data] with [params buffer_size] to the [param image]
-## This is an opmization method used by [param Draw.gd] while performing undo/redo.
+## Decompresses the [param compressed_image_data] with [param buffer_size] to the [param image]
+## This is an optimization method used while performing undo/redo drawing operations.
 func undo_redo_draw_op(
 	image: Image, compressed_image_data: PackedByteArray, buffer_size: int
 ) -> void:
@@ -885,3 +888,13 @@ func undo_redo_draw_op(
 		image.get_format(),
 		compressed_image_data.decompress(buffer_size)
 	)
+
+
+## Used by the Move tool for undo/redo, moves all of the [Image]s in [param images]
+## by [param diff] pixels.
+func undo_redo_move(diff: Vector2i, images: Array[Image]) -> void:
+	for image in images:
+		var image_copy := Image.new()
+		image_copy.copy_from(image)
+		image.fill(Color(0, 0, 0, 0))
+		image.blit_rect(image_copy, Rect2i(Vector2i.ZERO, image.get_size()), diff)

src/Tools/Move.gd

@@ -1,6 +1,5 @@
 extends BaseTool
 
-var _undo_data := {}
 var _start_pos: Vector2i
 var _offset: Vector2i
 
@@ -41,7 +40,6 @@ func draw_start(pos: Vector2i) -> void:
 		return
 	_start_pos = pos
 	_offset = pos
-	_undo_data = _get_undo_data()
 	if Global.current_project.has_selection:
 		selection_node.transform_content_start()
 	_content_transformation_check = selection_node.is_moving_content
@@ -82,14 +80,8 @@ func draw_end(pos: Vector2i) -> void:
 		else:
 			var pixel_diff := pos - _start_pos
 			Global.canvas.move_preview_location = Vector2i.ZERO
-			var images := _get_selected_draw_images()
-			for image in images:
-				var image_copy := Image.new()
-				image_copy.copy_from(image)
-				image.fill(Color(0, 0, 0, 0))
-				image.blit_rect(image_copy, Rect2i(Vector2i.ZERO, project.size), pixel_diff)
 
-			commit_undo("Draw")
+			_commit_undo("Draw", pixel_diff)
 
 	_start_pos = Vector2.INF
 	_snap_to_grid = false
@@ -120,42 +112,19 @@ func _snap_position(pos: Vector2) -> Vector2:
 	return pos
 
 
-func commit_undo(action: String) -> void:
-	var redo_data := _get_undo_data()
+func _commit_undo(action: String, diff: Vector2i) -> void:
 	var project := Global.current_project
 	var frame := -1
 	var layer := -1
 	if Global.animation_timer.is_stopped() and project.selected_cels.size() == 1:
 		frame = project.current_frame
 		layer = project.current_layer
 
+	var images := _get_selected_draw_images()
 	project.undos += 1
 	project.undo_redo.create_action(action)
-	for image in redo_data:
-		project.undo_redo.add_do_property(image, "data", redo_data[image])
-	for image in _undo_data:
-		project.undo_redo.add_undo_property(image, "data", _undo_data[image])
+	project.undo_redo.add_do_method(Global.undo_redo_move.bind(diff, images))
 	project.undo_redo.add_do_method(Global.undo_or_redo.bind(false, frame, layer))
+	project.undo_redo.add_undo_method(Global.undo_redo_move.bind(-diff, images))
 	project.undo_redo.add_undo_method(Global.undo_or_redo.bind(true, frame, layer))
 	project.undo_redo.commit_action()
-
-	_undo_data.clear()
-
-
-func _get_undo_data() -> Dictionary:
-	var data := {}
-	var project := Global.current_project
-	var cels: Array[BaseCel] = []
-	if Global.animation_timer.is_stopped():
-		for cel_index in project.selected_cels:
-			cels.append(project.frames[cel_index[0]].cels[cel_index[1]])
-	else:
-		for frame in project.frames:
-			var cel: PixelCel = frame.cels[project.current_layer]
-			cels.append(cel)
-	for cel in cels:
-		if not cel is PixelCel:
-			continue
-		var image: Image = cel.image
-		data[image] = image.data
-	return data