fix sorting

copienator/commands/gemini_for_labels.py

@@ -340,6 +340,34 @@ def group_images(image_files: list[Path]) -> dict[str, list[Path]]:
     return dict(groups)
 
 
+def sort_boxes_for_image(
+    workspace: EvaluationWorkspace,
+    image_file: Path,
+    boxes: list[BoxItem],
+) -> list[BoxItem]:
+    """Sort boxes in the image's page-column reading order when schema exists."""
+    match = re.match(r"(.+)_(\d+)$", image_file.stem)
+    if not match:
+        return boxes
+    schema_path = workspace.cutleft_dir / f"{match.group(1)}_schema.json"
+    try:
+        schema = read_json(schema_path)
+        columns_per_file = schema["columns_per_file"]
+        column_count = int(columns_per_file[int(match.group(2)) - 1])
+        if column_count < 1:
+            return boxes
+    except (OSError, KeyError, IndexError, TypeError, ValueError):
+        return boxes
+
+    def position(item: BoxItem) -> tuple[int, int, int]:
+        ymin, xmin, _ymax, xmax = item.box_2d
+        center_x = (xmin + xmax) // 2
+        column = min(column_count - 1, max(0, center_x * column_count // 1000))
+        return column, ymin, xmin
+
+    return sorted(boxes, key=position)
+
+
 def _existing_context(output_json: Path) -> list[str]:
     try:
         loaded = read_json(output_json)
@@ -452,6 +480,9 @@ def process_copy_group(
                         continue
                     annotation.name = "Unknown"
 
+                annotation.list = sort_boxes_for_image(
+                    workspace, image_file, annotation.list
+                )
                 atomic_write_json(output_json, annotation.model_dump())
                 accumulated_labels.extend(box.label for box in annotation.list)
                 generated += 1

copienator/commands/plotting.py

@@ -333,28 +333,16 @@ class ImageViewer:
 
             try:
                 current_data = read_json(self.current_json_path)
-                nb_pages = self.current_meta["schema"]["columns_per_file"][
-                    self.current_meta["part"] - 1
-                ]
-                original_items = current_data["list"]
-                ordered_items = sort_bounding_boxes(original_items, nb_pages)
-
-                if ordered_items != original_items:
-                    current_data["list"] = ordered_items
-                    atomic_write_json(self.current_json_path, current_data)
-                    print(
-                        f"Reordered labels by column in "
-                        f"{self.current_json_path.name}."
-                    )
+                items = current_data["list"]
 
                 # Perform the conversion now, post-edit
                 converted_items = convert_list(
-                    ordered_items,
+                    items,
                     self.current_meta["part"],
                     self.current_meta["schema"]
                 )
 
-                labels = normalized_labels(ordered_items)
+                labels = normalized_labels(items)
                 false_labels = [
                     label for label in labels if label not in self.valid_labels
                 ]