Upload fix_run_ocr.py

fix_run_ocr.py

@@ -0,0 +1,177 @@
+"""
+Applies three fixes to smoke_signal_tab.py:
+  1. font_name reset per page in second pass loop (line ~2205)
+  2. Top-level try/except inside run_ocr per book so errors surface in log
+  3. Phase log lines: book start, render complete, batch OCR complete, queue write complete
+
+Run with:
+  python fix_run_ocr.py
+"""
+from pathlib import Path
+
+TARGET = Path.home() / "Codex_Extractor" / "smoke_signal_tab.py"
+content = TARGET.read_text()
+original = content  # keep for diff check
+
+fixes_applied = []
+
+# ── Fix 1: Reset font_name at top of second pass loop ────────────────────────
+# The second pass loop at ~line 2205 never resets font_name per page,
+# so it carries stale value from previous page or first pass.
+
+old1 = (
+    '        # Second pass: build page-level OCR output and review queue.\n'
+    '        for page_data in profile.get("pages", []):\n'
+    '            page_num = page_data["page_number"]\n'
+    '            route = page_data["route"]\n'
+    '            font_name = page_data.get("font_name")\n'  # already added by earlier fix
+)
+
+# Check if fix 1 already applied
+if old1 in content:
+    fixes_applied.append("Fix 1: already applied (font_name reset in second pass)")
+else:
+    # Version without font_name reset (original)
+    old1_orig = (
+        '        # Second pass: build page-level OCR output and review queue.\n'
+        '        for page_data in profile.get("pages", []):\n'
+        '            page_num = page_data["page_number"]\n'
+        '            route = page_data["route"]\n'
+    )
+    new1 = (
+        '        # Second pass: build page-level OCR output and review queue.\n'
+        '        for page_data in profile.get("pages", []):\n'
+        '            page_num = page_data["page_number"]\n'
+        '            route = page_data["route"]\n'
+        '            font_name = page_data.get("font_name")  # reset per page — prevent stale carry-over\n'
+    )
+    if old1_orig in content:
+        content = content.replace(old1_orig, new1, 1)
+        fixes_applied.append("Fix 1: APPLIED — font_name reset per page in second pass")
+    else:
+        fixes_applied.append("Fix 1: SKIPPED — pattern not found (may already have font_name line)")
+
+# ── Fix 2 + 3: Per-book try/except + phase logs ───────────────────────────────
+# Find the per-book iteration and wrap it with try/except and phase logs.
+# We look for the for loop over eligible books and the profile load inside it.
+
+old2 = '    for _, book_row in eligible.iterrows():\n'
+
+new2 = '    for _, book_row in eligible.iterrows():\n'  # same — we patch inside
+
+# What we actually want: find the book_id assignment and wrap the book body.
+# Safer approach: inject phase logs at known anchor points.
+
+# Phase log: book start — inject after book_id is assigned
+old_book_start = (
+    '    for _, book_row in eligible.iterrows():\n'
+    '        book_id = str(book_row.get("book_id", "")).strip()\n'
+)
+new_book_start = (
+    '    for _, book_row in eligible.iterrows():\n'
+    '        book_id = str(book_row.get("book_id", "")).strip()\n'
+    '        log.append(log_line(f"▶ Starting OCR for {book_id}"))\n'
+)
+
+if old_book_start in content:
+    content = content.replace(old_book_start, new_book_start, 1)
+    fixes_applied.append("Fix 3a: APPLIED — phase log: book start")
+elif new_book_start in content:
+    fixes_applied.append("Fix 3a: already applied — book start log")
+else:
+    fixes_applied.append("Fix 3a: SKIPPED — book_id assignment pattern not found")
+
+# Phase log: after renders complete — find render dir creation anchor
+old_render_done = (
+    '            log.append(log_line(f"  ✓ {book_id}: rendered {len(items_to_ocr)} page(s)"))\n'
+)
+new_render_done = (
+    '            log.append(log_line(f"  ✓ {book_id}: rendered {len(items_to_ocr)} page(s)"))\n'
+    '            log.append(log_line(f"  → {book_id}: render complete — starting OCR batches"))\n'
+)
+if old_render_done in content and new_render_done not in content:
+    content = content.replace(old_render_done, new_render_done, 1)
+    fixes_applied.append("Fix 3b: APPLIED — phase log: render complete")
+else:
+    fixes_applied.append("Fix 3b: SKIPPED — render log anchor not found or already applied")
+
+# Phase log: after batch OCR loop — before second pass
+old_second_pass_header = (
+    '        # Second pass: build page-level OCR output and review queue.\n'
+)
+new_second_pass_header = (
+    '        log.append(log_line(f"  → {book_id}: batch OCR complete — building review queue"))\n'
+    '        # Second pass: build page-level OCR output and review queue.\n'
+)
+if old_second_pass_header in content and new_second_pass_header not in content:
+    content = content.replace(old_second_pass_header, new_second_pass_header, 1)
+    fixes_applied.append("Fix 3c: APPLIED — phase log: batch OCR complete")
+else:
+    fixes_applied.append("Fix 3c: SKIPPED — second pass header not found or already applied")
+
+# ── Fix 2: Per-book try/except wrapper ───────────────────────────────────────
+# Wrap the entire per-book block in try/except so errors surface in the log
+# instead of silently hanging or crashing the whole run_ocr call.
+# Find the queue_rows append at end of book processing as the anchor.
+
+old_queue_write = (
+            '        if queue_rows_for_book:\n'
+            '            processed_books.append(book_id)\n'
+)
+new_queue_write = (
+            '        if queue_rows_for_book:\n'
+            '            log.append(log_line(f"  ✓ {book_id}: queue write complete — {len(queue_rows_for_book)} page(s) added"))\n'
+            '            processed_books.append(book_id)\n'
+)
+if old_queue_write in content and new_queue_write not in content:
+    content = content.replace(old_queue_write, new_queue_write, 1)
+    fixes_applied.append("Fix 3d: APPLIED — phase log: queue write complete")
+else:
+    fixes_applied.append("Fix 3d: SKIPPED — queue write anchor not found or already applied")
+
+# Top-level per-book except — find end of book loop body to add except clause.
+# We look for the `finally: doc.close()` pattern which ends the book block.
+old_finally = (
+    '        finally:\n'
+    '            if doc is not None:\n'
+    '                try:\n'
+    '                    doc.close()\n'
+    '                except Exception:\n'
+    '                    pass\n'
+    '            doc = None\n'
+)
+new_finally = (
+    '        except Exception as _book_err:\n'
+    '            import traceback\n'
+    '            log.append(log_line(f"  ✗ {book_id}: UNHANDLED ERROR — {_book_err}"))\n'
+    '            log.append(log_line(traceback.format_exc()))\n'
+    '        finally:\n'
+    '            if doc is not None:\n'
+    '                try:\n'
+    '                    doc.close()\n'
+    '                except Exception:\n'
+    '                    pass\n'
+    '            doc = None\n'
+)
+
+# Only add except if there's already a try: wrapping the book body.
+# Check for existing try: at book level.
+if 'except Exception as _book_err:' in content:
+    fixes_applied.append("Fix 2: already applied — per-book except block exists")
+elif old_finally in content:
+    # Check there's a matching try: for this finally — if so, add except before finally
+    content = content.replace(old_finally, new_finally, 1)
+    fixes_applied.append("Fix 2: APPLIED — per-book except block added before finally")
+else:
+    fixes_applied.append("Fix 2: SKIPPED — finally block pattern not found; add manually")
+
+# ── Write file if changed ─────────────────────────────────────────────────────
+if content != original:
+    TARGET.write_text(content)
+    print(f"✓ Wrote {TARGET}")
+else:
+    print("⚠ No changes written — all fixes were already applied or skipped")
+
+print("\nFix summary:")
+for f in fixes_applied:
+    print(f"  {f}")