Initial commit for finalisation web interface

copienator/review_snapshot.py

@@ -0,0 +1,71 @@
+"""Rendering inputs captured after handwriting interpretation, before user overrides.
+
+Optional for the legacy readers. The web renderer consumes this state without
+rescanning annotation PDFs. A selective redo merges only the selected labels.
+"""
+from copy import deepcopy
+from hashlib import sha256
+import json
+from pathlib import Path
+from PIL import Image
+from copienator import atomic_write_json, read_json
+
+FILENAME = 'review_snapshot.json'
+
+
+def stage(output, staging, data, notes, *, engine, bnote=None, selected=None, force_filtered=()):
+    previous = read_json(output / FILENAME, default=None)
+    if selected and previous is None:
+        # Old preserved images cannot tell us their original feedback and ink.
+        return
+    if previous and previous['engine'] == 'standard' and engine == 'grouped':
+        geometry = {item['label']: item for item in previous['bnote']['images']}
+        for label, note in previous['notes'].items():
+            note['old_header_h'] = geometry.get(label, {}).get('header_height', 0)
+            if 'layout' in geometry.get(label, {}):
+                note['layout'] = geometry[label]['layout']
+    retained = set(force_filtered)
+    if selected:
+        base = read_json(output / 'annotation_baseline.json', default={})
+        retained.update(set(base.get('filtered', [])) - set(selected))
+    selected = set(selected or data)
+    if previous:
+        # Preserve raw inputs for the labels not re-read in a selective redo.
+        merged_data = previous['data'] if selected != set(data) else {}
+        merged_notes = previous['notes'] if selected != set(data) else {}
+    else:
+        merged_data, merged_notes = {}, {}
+    for label in selected:
+        if label in data:
+            content = deepcopy(data[label])
+            path = Path(content['pdf_path']).resolve()
+            root = output.parent.parent.resolve()
+            content['pdf_path'] = str(path.relative_to(root)) if path.is_relative_to(root) else str(path)
+            merged_data[label] = content
+        merged_notes.pop(label, None)
+        note = notes.get(label)
+        if note is not None:
+            filename = '.review-note-' + sha256(label.encode()).hexdigest() + '.png'
+            note['img'].save(staging / filename)
+            merged_notes[label] = {k: v for k, v in note.items() if k != 'img'} | {'file': filename}
+    payload = {'version': 1, 'engine': engine, 'data': merged_data, 'notes': merged_notes, 'bnote': bnote, 'force_filtered': sorted(retained)}
+    # Roundtrip normalizes tuples while rejecting unsupported rendering data.
+    atomic_write_json(staging / FILENAME, json.loads(json.dumps(payload)))
+
+
+def load(root, output):
+    saved = read_json(output / FILENAME)
+    if saved.get('version') != 1:
+        raise ValueError('Version du résultat de lecture inconnue. Relisez les annotations.')
+    data = saved['data']
+    for content in data.values():
+        path = Path(content['pdf_path'])
+        content['pdf_path'] = path if path.is_absolute() else root / path
+    notes = {}
+    for label, metadata in saved['notes'].items():
+        filename = metadata['file']
+        if Path(filename).name != filename:
+            raise ValueError('Image manuscrite invalide.')
+        with Image.open(output / filename) as image:
+            notes[label] = {k: v for k, v in metadata.items() if k != 'file'} | {'img': image.copy()}
+    return saved, data, notes