better faulty encoding detection

copienator/commands/post_correction.py

@@ -16,6 +16,7 @@ from copienator import (
     read_json,
     workspace_from_args,
 )
+from copienator.latex_json import repair_decoded_latex_controls
 
 WORD_LIST_FILE = Path(__file__).resolve().parents[1] / "data" / "liste_francais.txt"
 ACCENT_PATTERN = re.compile(r"[éèêëàâäîïôöùûüçœÉÈÊËÀÂÄÎÏÔÖÙÛÜÇŒ]")
@@ -89,6 +90,7 @@ def fix_hex_corruption_safe(text: str) -> str:
 
 def repair_json_escape_corruption(text: str) -> str:
     r"""Restore observed LaTeX commands consumed as JSON control escapes."""
+    text = repair_decoded_latex_controls(text)
     replacements = (
         ("\x0crac", r"\frac"),
         ("\x0ceuille", r"\equiv"),
@@ -132,6 +134,33 @@ def clean_obj(value: Any, lookup: dict[str, str]) -> Any:
     return value
 
 
+def unresolved_math_control_paths(value: Any, path: str = "$") -> list[str]:
+    """Locate destructive JSON control escapes left in mathematical text."""
+    if isinstance(value, str):
+        return (
+            [path]
+            if any(
+                control in {"\b", "\f"}
+                for match in MATH_PATTERN.finditer(value)
+                for control in match.group(0)
+            )
+            else []
+        )
+    if isinstance(value, list):
+        return [
+            issue
+            for index, item in enumerate(value)
+            for issue in unresolved_math_control_paths(item, f"{path}[{index}]")
+        ]
+    if isinstance(value, dict):
+        return [
+            issue
+            for key, item in value.items()
+            for issue in unresolved_math_control_paths(item, f"{path}[{key!r}]")
+        ]
+    return []
+
+
 def run(
     workspace: EvaluationWorkspace,
     *,
@@ -141,11 +170,18 @@ def run(
     lookup = build_lookup_map(word_list_path)
     data = read_json(workspace.correction_file)
     cleaned = clean_obj(data, lookup)
+    unresolved = unresolved_math_control_paths(cleaned)
     backup = workspace.root / "correction_precleanup.json"
     atomic_write_bytes(backup, workspace.correction_file.read_bytes())
     print(f"Original JSON backed up to {backup}")
     atomic_write_json(workspace.correction_file, cleaned)
     print(f"Fixed JSON saved to {workspace.correction_file}")
+    if unresolved:
+        print(
+            "Warning: unresolved JSON control characters remain in mathematical "
+            "text at: " + ", ".join(unresolved)
+        )
+        return ExitCode.PARTIAL
     return ExitCode.SUCCESS