32k pixel as hard group limit, better manual resolution support

copienator/commands/annotating_by_label.py

@@ -27,7 +27,7 @@ from copienator.commands import annotating, annotating_with_checks
 from copienator.filesystem import staged_directory
 from copienator.utils import natural_key
 
-MAX_HEIGHT_PX = 25000
+MAX_HEIGHT_PX = 32000
 
 
 def render_item(item):
@@ -63,6 +63,11 @@ def save_batch(batch, prefix, group_id, output_root: Path) -> None:
     print(f"Generating group PDF: {prefix} G{group_id} ({len(batch)} elements)")
     max_width = max(item[2].width for item in batch)
     total_height = sum(item[2].height for item in batch)
+    if total_height > MAX_HEIGHT_PX:
+        raise CliError(
+            f"{prefix} G{group_id} would be {total_height} px high; "
+            f"the maximum is {MAX_HEIGHT_PX} px."
+        )
     concatenated = Image.new("RGB", (max_width, total_height), "white")
     draw = ImageDraw.Draw(concatenated)
     checkbox_map: list[dict[str, Any]] = []
@@ -268,32 +273,44 @@ def _existing_group_state(
 
 
 def split_batches(rendered):
-    def split(maximum_height: float):
-        batches = []
-        current = []
-        current_height = 0
-        previous_student = None
-        for item in rendered:
-            student_id = item[0]
-            image_height = item[2].height
-            if (
-                current
-                and current_height + image_height > maximum_height
-                and student_id != previous_student
-            ):
-                batches.append(current)
-                current = []
-                current_height = 0
-            current.append(item)
-            current_height += image_height
-            previous_student = student_id
+    batches = []
+    current = []
+    current_height = 0
+
+    def flush() -> None:
+        nonlocal current, current_height
         if current:
             batches.append(current)
-        return batches
+            current = []
+            current_height = 0
 
-    strict = split(MAX_HEIGHT_PX)
-    relaxed = split(1.1 * MAX_HEIGHT_PX)
-    return relaxed if len(relaxed) < len(strict) else strict
+    index = 0
+    while index < len(rendered):
+        student_id = rendered[index][0]
+        end = index + 1
+        while end < len(rendered) and rendered[end][0] == student_id:
+            end += 1
+        student_items = rendered[index:end]
+        student_height = sum(item[2].height for item in student_items)
+
+        if student_height > MAX_HEIGHT_PX:
+            labels = ", ".join(item[1] for item in student_items)
+            raise CliError(
+                f"Copie{student_id} ({labels}) is {student_height} px high; "
+                f"one student's labels cannot be split across groups and the "
+                f"maximum group height is {MAX_HEIGHT_PX} px."
+            )
+
+        # Consider the student's complete block before adding its first image.
+        # Otherwise later images from that same student could overflow the batch.
+        if current and current_height + student_height > MAX_HEIGHT_PX:
+            flush()
+        current.extend(student_items)
+        current_height += student_height
+        index = end
+
+    flush()
+    return batches
 
 
 def _generate_groups(