Use per-page annotation spacing with inferred margin fallback

copienator/commands/annotating.py

@@ -63,9 +63,13 @@ def make_base_image(pdf_path):
     base_img = Image.new("RGBA", (max_w, total_h), "white")
 
     current_y = 0
+    page_regions = []
     for page in pages:
-        base_img.paste(page.convert("RGBA"), ((max_w - page.width) // 2, current_y))
+        x = (max_w - page.width) // 2
+        base_img.paste(page.convert("RGBA"), (x, current_y))
+        page_regions.append({'x': x, 'y': current_y, 'width': page.width, 'height': page.height})
         current_y += page.height
+    base_img.info['page_regions'] = page_regions
     return (base_img, total_h, max_w)
 
 def normalize_mathtext(text):
@@ -352,6 +356,10 @@ def compose_label_image(base_img, label, result, hmin,
 
     layout = dict(layout if layout is not None else layout_for(source_pdf))
     margin_left = layout['margin_left']
+    page_regions = base_img.info.get('page_regions', [
+        {'x': 0, 'y': 0, 'width': base_img.width, 'height': base_img.height}])
+    page_margins = layout.get('page_margins', [])
+    per_page = layout.get('version', 1) >= 2
     left_pad = 0
     if base_img.width < TARGET_MIN_WIDTH:
         total_missing = TARGET_MIN_WIDTH - base_img.width
@@ -411,7 +419,11 @@ def compose_label_image(base_img, label, result, hmin,
     total_height = base_img.height + header_height
 
     # Create Canvas
-    final_img = Image.new("RGB", (base_img.width + margin_left, total_height), "white")
+    final_width = (max(TARGET_MIN_WIDTH + margin_left,
+                      max(region['width'] + (page_margins[i] if i < len(page_margins) else margin_left)
+                          for i, region in enumerate(page_regions)))
+                   if per_page else base_img.width + margin_left)
+    final_img = Image.new("RGB", (final_width, total_height), "white")
 
     # Draw Headers
     current_y = 0
@@ -434,7 +446,13 @@ def compose_label_image(base_img, label, result, hmin,
 
     # Paste Base Image
     image_offset_y = current_y
-    final_img.paste(base_img, (margin_left, image_offset_y))
+    if per_page:
+        for i, region in enumerate(page_regions):
+            x, y, w, h = (region[key] for key in ('x', 'y', 'width', 'height'))
+            margin = page_margins[i] if i < len(page_margins) else margin_left
+            final_img.paste(base_img.crop((x, y, x + w, y + h)), (margin, image_offset_y + y))
+    else:
+        final_img.paste(base_img, (margin_left, image_offset_y))
 
     # 2. Draw Local Annotations
     draw = ImageDraw.Draw(final_img, "RGBA") # Refresh draw object
@@ -451,12 +469,24 @@ def compose_label_image(base_img, label, result, hmin,
 
         # Draw Rectangle (if not suppressed)
         if "norectangle" not in fb:
-            draw.rectangle([target_xmin, target_ymin, target_xmax, target_ymax], outline="red", width=3)
+            boxes = [[target_xmin, target_ymin, target_xmax, target_ymax]]
+            if per_page:
+                boxes = []
+                for i, region in enumerate(page_regions):
+                    top = max(ymin - hmin, region['y'])
+                    bottom = min(ymax - hmin, region['y'] + region['height'])
+                    if bottom <= top:
+                        continue
+                    margin = page_margins[i] if i < len(page_margins) else margin_left
+                    shift = margin - region['x']
+                    boxes.append([xmin + shift, top + image_offset_y,
+                                  xmax + shift, bottom + image_offset_y])
+            for box in boxes:
+                draw.rectangle(box, outline="red", width=3)
 
-            if draw_callback:
-                callback_items.append(("local_rect",
-                                       {"box": [target_xmin, target_ymin, target_xmax, target_ymax]},
-                                       {"data": fb, "index": idx}))
+                if draw_callback:
+                    callback_items.append(("local_rect", {"box": box},
+                                           {"data": fb, "index": idx}))
 
         # Render Text
         txt_img = render_fn(fb['text'], width_px=ANNOT_WIDTH,