miscs

copienator/commands/cutleft.py

@@ -284,6 +284,7 @@ class ImageReviewer:
         self.current_width_offset = 0
         self.default_max_per_file = default_max_per_file
         self.current_max_per_file = default_max_per_file
+        self.review_settings: dict[int, tuple[int, int, int]] = {}
         self.current_preview: Image.Image | None = None
         self.is_processing = False
         self.manual_queue: Queue[
@@ -298,6 +299,8 @@ class ImageReviewer:
         self.label_info = tk.Label(self.root, text="", font=("Arial", 12, "bold"))
         self.label_info.pack(pady=5)
         self.root.bind("<Return>", self.on_next)
+        self.root.bind("p", self.on_previous)
+        self.root.bind("P", self.on_previous)
         self.root.bind("!", self.on_skip)
         self.root.protocol("WM_DELETE_WINDOW", self.on_close)
         self.root.bind("<Left>", lambda _event: self.on_shift(-CROP_SHIFT_STEP))
@@ -343,8 +346,9 @@ class ImageReviewer:
             self.on_close()
             return
         self.is_processing = False
-        self.current_shift = 0
-        self.current_width_offset = 0
+        self.current_shift, self.current_width_offset, self.current_max_per_file = (
+            self.review_settings.get(self.index, (0, 0, self.default_max_per_file))
+        )
         self.current_result = None
         self.margin_analysis = None
         self.trigger_processing(self.files[self.index], self.current_shift)
@@ -426,7 +430,7 @@ class ImageReviewer:
                 f"Cols: {schema['columns_per_file']}\n"
                 f"{margin_info}"
                 "Arrows: left/right move by 50px | up/down widen/narrow by 50px\n"
-                "Enter: Save and next | !: flag error and skip | 1: use full pages"
+                "Enter: Save and next | P/p: previous copy | !: flag error and skip | 1: use full pages"
                 f"{personal_instructions}"
             ),
             fg="black",
@@ -474,11 +478,21 @@ class ImageReviewer:
         self.stop_prefetch.set()
         self.root.destroy()
 
+    def _remember_settings(self) -> None:
+        self.review_settings[self.index] = (
+            self.current_shift, self.current_width_offset, self.current_max_per_file,
+        )
+
+    def on_previous(self, _event=None) -> None:
+        if self.is_processing or self.index == 0:
+            return
+        self._remember_settings()
+        self.index -= 1
+        self.load_current_image()
+
     def _advance(self) -> None:
+        self._remember_settings()
         self.index += 1
-        self.current_shift = 0
-        self.current_width_offset = 0
-        self.current_max_per_file = self.default_max_per_file
         self.load_current_image()
 
 

copienator/commands/enonce_info.py

@@ -32,7 +32,7 @@ def fetch_and_save_sub_text(ex_id, indices, label, text_path):
     try:
         with urllib.request.urlopen(url) as response:
             content = response.read().decode('utf-8')
-            content = replace_dots(content.strip("\n"))
+            content = prepare_latex_fragment(content.strip("\n"))
             with open(os.path.join(text_path, f"{label}.tex"), 'w', encoding='utf-8') as f:
                 f.write(content)
             # Compile PDF
@@ -52,7 +52,7 @@ def fetch_and_save_sub_sol(ex_id, indices, label, sol_path):
     try:
         with urllib.request.urlopen(url) as response:
             content = response.read().decode('utf-8')
-            content = replace_dots(content.strip("\n"))
+            content = prepare_latex_fragment(content.strip("\n"))
             with open(os.path.join(sol_path, f"{label}.tex"), 'w', encoding='utf-8') as f:
                 f.write(content)
             # Compile PDF
@@ -70,13 +70,39 @@ def replace_dots(text):
     # (?m) enables multiline mode so ^ matches start of each line
     return re.sub(r"(?m)^(\s*.)\.", r"\1)", text)
 
-def replace_problem_labels(text):
+
+def prepare_latex_fragment(text):
+    """Keep counter-only extracted lists valid before an enumitem resume."""
+    # Extraction can omit every item before a context block while retaining
+    # the list and its counters for resume*. An unnumbered item makes that
+    # empty list legal without advancing the resumed question number.
+    return re.sub(
+        r'(\\begin\{enumerate\}(?:\[[^\]\n]*\])?)'
+        r'(\s*(?:\\stepcounter\{[^{}]+\}\s*)*)'
+        r'(\\end\{enumerate\})',
+        r'\1\\item[]\2\3',
+        replace_dots(text),
+    )
+
+def replace_problem_labels(text, numbering="problem"):
     """Replaces labels according to spaces depth when problem=True."""
     def repl(m):
         spaces = m.group(1)
         label = m.group(2)
         n = len(spaces)
         try:
+            if numbering == "problem_abc":
+                if n == 1 and label.isdigit():
+                    return f"{spaces}{ROMANS_CAP[int(label)]})"
+                if n == 4 and label.isalpha():
+                    return f"{spaces}{label.upper()})"
+                if n == 7 and label.isdigit():
+                    return m.group(0)
+                if n == 10 and label.isdigit():
+                    return f"{spaces}{chr(96 + int(label))})"
+                if n == 13 and label.isdigit():
+                    return f"{spaces}{ROMANS_LOW[int(label)]})"
+                return m.group(0)
             if n == 1 and label.isdigit():     # 1 space: 1) -> I)
                 return f"{spaces}{ROMANS_CAP[int(label)]})"
             elif n == 4 and label.isalpha():   # 4 spaces: a) -> 1)
@@ -92,7 +118,17 @@ def replace_problem_labels(text):
     # Matches start of line, spaces, alphanumeric label, and closing parenthesis
     return re.sub(r"(?m)^([ \t]+)([a-zA-Z0-9]+)\)", repl, text)
 
-def format_indices(indices, problem=False):
+def format_indices(indices, problem=False, *, numbering=None):
+    if numbering is not None:
+        if numbering not in {"exercise", "problem", "problem_abc"}:
+            raise ValueError(f"Unknown SHEETINFO numbering scheme: {numbering!r}")
+        problem = numbering != "exercise"
+    if numbering == "problem_abc":
+        formatters = (lambda n: ROMANS_CAP[n], lambda n: chr(64 + n), str,
+                      lambda n: chr(96 + n), lambda n: ROMANS_LOW[n])
+        if len(indices) > len(formatters):
+            raise ValueError("Too many question levels for problem_abc numbering")
+        return "".join(formatter(index) + ")" for formatter, index in zip(formatters, indices))
     if not indices: return ""
     if not problem:
         res = f"{indices[0]})"
@@ -200,6 +236,11 @@ def process_directory(workspace: EvaluationWorkspace) -> ExitCode:
 
             try:
                 data = json.loads(json_str)
+                numbering = data.get('numbering')
+                if numbering is not None:
+                    # Validate even exercises without question indexes.
+                    format_indices([], numbering=numbering)
+                    problem = numbering != 'exercise'
                 block_labels = []
                 # Construct 'ids' parameter
                 ex_id = str(data['id'])
@@ -222,7 +263,7 @@ def process_directory(workspace: EvaluationWorkspace) -> ExitCode:
                     fetch_and_save_sub_sol(ids, [], label, paths['Sol2'])
                 else:
                     for item in indexes:
-                        suffix = format_indices(item['indices'], problem)
+                        suffix = format_indices(item['indices'], problem, numbering=numbering)
                         label = f"Ex {current_ex_num}" + (f" : {suffix}" if suffix else "")
                         validate_windows_labels([label])
                         block_labels.append(label)
@@ -252,9 +293,9 @@ def process_directory(workspace: EvaluationWorkspace) -> ExitCode:
 
                 # Apply hierarchy depth replace if problem context
                 if problem:
-                    t_text = replace_problem_labels(t_text)
-                    s_text = replace_problem_labels(s_text)
-                    p_text = replace_problem_labels(p_text)
+                    t_text = replace_problem_labels(t_text, numbering or 'problem')
+                    s_text = replace_problem_labels(s_text, numbering or 'problem')
+                    p_text = replace_problem_labels(p_text, numbering or 'problem')
 
                 base_filename = f"Ex {current_ex_num}"
 

copienator/commands/gemini_for_enonce.py

@@ -381,6 +381,9 @@ def refine_existing(workspace: EvaluationWorkspace, mode: str, *, api_client=Non
         )
         groups = LabelGroups.model_validate_json(response.text).groups
         validate_groups(groups, labels)
+        label_order = {label: index for index, label in enumerate(labels)}
+        groups = [sorted(group, key=label_order.__getitem__) for group in groups]
+        groups.sort(key=lambda group: label_order[group[0]])
         atomic_write_text(workspace.label_groups_file,
                           "".join(", ".join(group) + "\n" for group in groups))
         print(f"Updated label_groups: {len(groups)} Gemini groups.")

copienator/commands/page_splitter.py

@@ -41,9 +41,13 @@ PAGE_SPLITTER_KB = {
 # Conversion factor: 1 cm to points (1 inch = 2.54 cm, 72 points = 1 inch)
 CM_TO_POINTS = (1 / 2.54) * 72
 
-def list_pdf_files(directory: str | Path) -> list[Path]:
+def list_pdf_files(
+    directory: str | Path, workspace: EvaluationWorkspace | None = None,
+) -> list[Path]:
+    workspace = workspace or EvaluationWorkspace(Path(directory))
     paths = sorted(Path(directory).glob("*.pdf"), key=lambda path: path.name.casefold())
-    return [path for path in paths if "enonce" not in path.name.casefold()]
+    return [path for path in paths if "enonce" not in path.name.casefold()
+            and not workspace.is_statement_pdf(workspace.root / path.name)]
 
 
 def _temporary_sibling(path: Path, purpose: str) -> Path:
@@ -177,7 +181,7 @@ class PDFPreviewer:
             f"'{fmt('rotate_page')}': Rotate page 180°, '{fmt('rotate_all_pages')}' : rotate all pages, '{fmt('rotate_all_files')}' : rotate all files\n"
             f"{fmt('keep_left')} {fmt('next_page')} {fmt('discard_page')} {fmt('keep_right')} {fmt('keep_as_is')}: keep left, next page, keep none, keep right, keep as is\n"
             f"{fmt('send_end')}: send page to end, '{fmt('arranger')}': pdf arranger, '{fmt('restart_file')}': restart file, '{fmt('prev_file')}': previous file\n"
-            f"{fmt('add_secondary_splitter')}: add secondary line, {fmt('delete_secondary_splitter')}: delete newest (arrows move newest; orange = active)\n"
+            f"{fmt('add_secondary_splitter')}: add secondary line, {fmt('delete_secondary_splitter')}: delete newest (arrows move newest; main line red, active secondary orange)\n"
             f"'{fmt('reverse_pages')}': reverse page order and restart from the new first page\n"
         )
 
@@ -302,7 +306,7 @@ class PDFPreviewer:
         # The line position is scaled by the same zoom factor
         line_x_scaled = self.current_line_x * self.current_zoom
         draw.line([(line_x_scaled, 0), (line_x_scaled, pix.height)],
-                  fill="red" if self.secondary_splitters else "orange", width=3)
+                  fill="red", width=3)
         for index, position in enumerate(self.secondary_splitters):
             x = position * self.current_zoom
             active = index == len(self.secondary_splitters) - 1
@@ -664,12 +668,14 @@ def _selected_inputs(
     if target.is_file():
         if target.suffix.casefold() != ".pdf":
             raise CliError(f"Target is not a PDF: {target}", ExitCode.INVALID_ARGUMENTS)
+        if workspace.is_statement_pdf(workspace.root / target.name):
+            raise CliError(f"Target is the statement PDF, not a copy: {target}", ExitCode.INVALID_ARGUMENTS)
         backup = workspace.original_copies_dir / target.name
         return [backup if backup.is_file() else target]
 
     directory = target
     if target == workspace.copies_dir:
-        candidates = list_pdf_files(workspace.copies_dir)
+        candidates = list_pdf_files(workspace.copies_dir, workspace)
         candidates = [
             (
                 workspace.original_copies_dir / path.name
@@ -679,7 +685,7 @@ def _selected_inputs(
             for path in candidates
         ]
     else:
-        candidates = list_pdf_files(directory)
+        candidates = list_pdf_files(directory, workspace)
     return list(reversed(candidates))
 
 

copienator/commands/plotting.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import argparse
 import copy
+import json
 import queue
 import re
 import threading
@@ -16,6 +17,7 @@ from copienator import (
     EvaluationWorkspace,
     ExitCode,
     atomic_write_json,
+    atomic_write_text,
     execute,
     read_json,
     target_parser,
@@ -31,6 +33,31 @@ MISSING_LABEL_COLOR = "orange"
 COMMON_MISSING_LABEL_COLOR = "#403a00"
 COMMON_MISSING_THRESHOLD = 0.66
 
+
+def prepare_label_json_for_edit(path: Path) -> None:
+    """Leave a trailing comma after the last label so another can be pasted."""
+    data = read_json(path, repair_trailing_commas=True)
+    if not isinstance(data, dict) or not isinstance(data.get("list"), list) or not data["list"]:
+        return
+    text = path.read_text(encoding="utf-8")
+    decoder = json.JSONDecoder()
+    whitespace = re.compile(r"\s*")
+    skip_space = lambda position: whitespace.match(text, position).end()
+    cursor = skip_space(text.index("{") + 1)
+    # Decode top-level values to locate this list without matching delimiters
+    # inside strings or nested bounding-box arrays. Keep the existing layout.
+    while text[cursor] != "}":
+        key, end = decoder.raw_decode(text, cursor)
+        start = skip_space(skip_space(end) + 1)  # Skip the key's colon.
+        _value, end = decoder.raw_decode(text, start)
+        if key == "list":
+            insertion = len(text[:end - 1].rstrip())
+            atomic_write_text(path, text[:insertion] + "," + text[insertion:])
+            return
+        cursor = skip_space(end)
+        if text[cursor] == ",":
+            cursor = skip_space(cursor + 1)
+
 try:
     font = ImageFont.truetype("DejaVuSans.ttf", size=30)
 except OSError:
@@ -100,7 +127,7 @@ def frequently_missing_labels(
         if match is None:
             continue
         try:
-            data = read_json(json_path)
+            data = read_json(json_path, repair_trailing_commas=True)
             entries = data["list"]
             if not isinstance(entries, list):
                 continue
@@ -217,11 +244,11 @@ def _worker_items(base_dir, files_to_process, all_labels, output_queue):
         nb_pages = json_schema["columns_per_file"][copie_part-1]
 
         if json_path.exists():
-            # Read strictly for visualization purposes
+            # Repair accidental trailing commas in user-edited label JSON.
             bb_list = []
             json_name = ""
             try:
-                json_result = read_json(json_path)
+                json_result = read_json(json_path, repair_trailing_commas=True)
                 bb_list = json_result.get("list", [])
                 json_name = json_result.get("name", "")
             except Exception as e:  # noqa: BLE001 - malformed user-editable JSON
@@ -439,7 +466,7 @@ class ImageViewer:
             previous_results = copy.deepcopy(self.accumulated_results)
 
             try:
-                current_data = read_json(self.current_json_path)
+                current_data = read_json(self.current_json_path, repair_trailing_commas=True)
                 items = current_data["list"]
 
                 # Perform the conversion now, post-edit
@@ -532,6 +559,11 @@ class ImageViewer:
 
     def on_edit(self, event):
         if self.is_viewing and self.current_json_path:
+            try:
+                prepare_label_json_for_edit(self.current_json_path)
+            except (OSError, ValueError) as error:
+                # Still let the user open and fix other JSON errors manually.
+                print(f"Could not prepare trailing comma: {error}")
             print(f"Opening {self.current_json_path}")
             open_path(self.current_json_path)
 

copienator/json_io.py

@@ -19,9 +19,40 @@ class JsonLockTimeout(TimeoutError):
     pass
 
 
-def read_json(path: str | Path, *, default: T | object = _MISSING) -> T:
+def read_json(
+    path: str | Path, *, default: T | object = _MISSING,
+    repair_trailing_commas: bool = False,
+) -> T:
     target = Path(path)
     try:
+        if repair_trailing_commas and target.name != "refaire.json":
+            original = text = target.read_text(encoding="utf-8")
+            while True:
+                try:
+                    value = json.loads(text)
+                except json.JSONDecodeError as error:
+                    # Only remove a comma immediately before a closing
+                    # delimiter reported by the parser, never inside strings.
+                    position = error.pos
+                    # Python 3.14 points at the comma; older versions point
+                    # at the closing bracket or brace.
+                    if error.msg.startswith("Illegal trailing comma"):
+                        position += 1
+                        while position < len(text) and text[position] in " \t\r\n":
+                            position += 1
+                    before = text[:position].rstrip()
+                    if (position >= len(text) or text[position] not in "]}"
+                            or not before.endswith(",")
+                            or not before[:-1].rstrip()
+                            or before[:-1].rstrip()[-1] in "[{:,"):
+                        raise
+                    comma = len(before) - 1
+                    text = text[:comma] + text[comma + 1:]
+                    continue
+                break
+            if text != original:
+                atomic_write_text(target, text)
+            return cast(T, value)
         with target.open("r", encoding="utf-8") as stream:
             if target.name == "refaire.json":
                 # This user-edited selection supports whole-line // comments.

copienator/left_margin.py

@@ -15,7 +15,8 @@ BODY_TOP = .30
 BODY_BOTTOM = .82
 DISCARD_LEFT_FRACTION = .10
 STANDARD_MARGIN_MM = 40
-MIN_MARGIN_MM = STANDARD_MARGIN_MM / 2
+MARGIN_CLEARANCE_MM = 20
+MIN_MARGIN_MM = MARGIN_CLEARANCE_MM
 
 
 def trimmed_left_edge(starts: list[float]) -> float | None:
@@ -89,7 +90,7 @@ def format_groups(images: list[Image.Image], reference_width: float) -> list[lis
 
 
 def analyse_margins(images: list[Image.Image], render_dpi: int, reference_width: float) -> dict:
-    """Pool occupied bands per format, trim outliers, and crop at half the edge."""
+    """Pool occupied bands and use the larger half-margin or 20 mm-clearance cut."""
     starts = [0] * len(images)
     groups = []
     page_estimates = {}
@@ -113,7 +114,8 @@ def analyse_margins(images: list[Image.Image], render_dpi: int, reference_width:
         left_mm = 0.0
         if len(values) >= min(2, len(indices)):
             status = 'near-edge' if edge < MIN_MARGIN_MM else 'margin'
-            left_mm = 0.0 if edge < MIN_MARGIN_MM else edge / 2
+            left_mm = (0.0 if edge < MIN_MARGIN_MM else
+                       max(edge / 2, edge - MARGIN_CLEARANCE_MM))
         # Secondary subpages can start in the middle of an answer. Preserve
         # their left edge even when the only visible writing is indented.
         if images[indices[0]].width < .8 * reference_width:
@@ -152,8 +154,9 @@ def analyse_margins(images: list[Image.Image], render_dpi: int, reference_width:
                       'status': 'unknown' if edge is None else
                       ('low' if edge < MIN_MARGIN_MM else 'normal')})
 
-    return {'version': 6, 'pages': pages, 'starts_px': starts, 'groups': groups,
+    return {'version': 8, 'pages': pages, 'starts_px': starts, 'groups': groups,
             'no_margin_from_page': no_margin_from_page,
-            'discard_left_fraction': DISCARD_LEFT_FRACTION, 'cut_fraction_of_margin': .5,
+            'discard_left_fraction': DISCARD_LEFT_FRACTION, 'margin_clearance_mm': MARGIN_CLEARANCE_MM,
+            'cut_fraction_of_margin': .5, 'cut_rule': 'maximum',
             'standard_margin_mm': STANDARD_MARGIN_MM, 'minimum_margin_mm': MIN_MARGIN_MM,
             'body_height_fraction': [BODY_TOP, BODY_BOTTOM]}

copienator/workspace.py

@@ -119,7 +119,10 @@ class EvaluationWorkspace:
         return self.resolve_file("correction.tex")
 
     def is_statement_pdf(self, path: Path) -> bool:
-        return path in self.input_candidates("enonce.pdf")
+        return path.parent == self.root and path.name.replace(" ", "") in {
+            candidate.name.replace(" ", "")
+            for candidate in self.input_candidates("enonce.pdf")
+        }
 
     @property
     def metadata_dir(self) -> Path: