When no margin detected, reserve a bit more space for comments

copienator/commands/splitting_int.py

@@ -22,6 +22,7 @@ from copienator import (
     workspace_from_target,
 )
 from copienator.filesystem import staged_directory
+from copienator.annotation_layout import fingerprint, metadata_path, save_parts
 
 SQUARE = 1000 // 38
 ANSWER_TOP_PADDING_POINTS = 4 * 72 / 25.4
@@ -129,6 +130,11 @@ def _render_split_outputs(
         visible_bounds = _prepare_split_pages(document)
         parsed = _parse_coordinates(coords_list)
         parts_by_label: defaultdict[str, list[Path]] = defaultdict(list)
+        provenance: defaultdict[str, list[dict]] = defaultdict(list)
+        schema = read_json(input_pdf.parent.parent / 'Cutleft' / f'{input_pdf.stem}_schema.json', default={})
+        analysis = schema.get('auto_margin') or {}
+        measurements = ({item['page']: item for item in analysis.get('pages', [])}
+                        if schema.get('source_sha256') == fingerprint(input_pdf) else {})
         with tempfile.TemporaryDirectory(prefix="copienator-split-") as temp_directory:
             temporary = Path(temp_directory)
             for index, item in enumerate(parsed):
@@ -200,6 +206,15 @@ def _render_split_outputs(
                         part_path,
                     )
                     parts_by_label[clean_label].append(part_path)
+                    edge = measurements.get(page_number + 1, {}).get('edge_mm')
+                    remaining = None if edge is None else max(0, edge - clip.x0 * 25.4 / 72)
+                    provenance[clean_label].append({
+                        'source': input_pdf.name, 'source_sha256': schema.get('source_sha256'),
+                        'page': page_number + 1, 'crop_points': list(clip),
+                        'margin_mm': remaining,
+                        'margin_status': ('unknown' if remaining is None else
+                                          'low' if remaining < 20 else 'normal'),
+                    })
 
             generated: set[str] = set()
             for label, parts in parts_by_label.items():
@@ -212,6 +227,8 @@ def _render_split_outputs(
                 finally:
                     merger.close()
                 generated.add(filename)
+                save_parts(staging / filename, provenance[label])
+                generated.add(metadata_path(staging / filename).name)
             return generated
     finally:
         document.close()