Update app.py

app.py

@@ -59,6 +59,12 @@ Primary goals for reconcile rate:
             when the preceding DA3 table was under a deposits/credits heading. Only strips the
             contiguous prefix of the debit table whose rows appear in the credit table; requires
             ≥1 data row to remain; entire function wrapped in try/except (safe no-op on any error).
+- Fix 29: Move a misplaced debit/withdrawal heading to before its orphan DA3 table when OCR
+            places the heading AFTER the table instead of before it. Pattern detected:
+            [credit heading+table] ... [orphan DA3 table] ... [debit heading] — the heading is
+            relocated to just before the orphan table. Guards: no debit heading already before
+            table; credit context must precede table; no other table between orphan and heading;
+            try/except wrapper (safe no-op on any error).
 """
 
 # Patch asyncio first (before Gradio imports it) to suppress Python 3.13 cleanup noise
@@ -3510,6 +3516,122 @@ def _strip_credits_prefix_from_debit_da3_table(text: str) -> str:
         return text
 
 
+# --------------------------
+# Fix 29: Move misplaced debit heading to before its orphan DA3 table
+# --------------------------
+
+# Centered div heading pattern (BoA / generic style)
+_CENTER_DIV_HDR_RE = re.compile(
+    r'<div\s+align=["\']center["\']\s*>\s*(.*?)\s*</div>',
+    re.IGNORECASE | re.DOTALL,
+)
+_HEADING_SCAN_FORWARD = 800   # chars after table end to search for misplaced heading
+
+
+def _fix29_move_debit_heading_before_orphan_table(text: str) -> str:
+    """
+    Fix 29: When OCR places a debit/withdrawal section heading AFTER the withdrawals table
+    instead of before it, move the heading to its correct position (just before the table).
+
+    Pattern detected:
+      [credit/deposit DA3 table]  ...no debit heading...  [orphan DA3 table]  [debit heading]
+      where no other DA3 table separates the orphan table from the debit heading.
+
+    Action: move the debit heading to just before the orphan DA3 table.
+
+    Safety guards (ALL must pass):
+      1. Orphan table must be a valid DA3 (Date|Description|Amount).
+      2. A debit heading must appear AFTER the orphan table within _HEADING_SCAN_FORWARD chars.
+      3. No debit heading appears immediately before the orphan table (within _HEADING_SCAN_WINDOW).
+      4. A credit/deposit heading must exist before the orphan table (within 1200 chars).
+      5. No other table appears between the orphan table end and the debit heading.
+      6. Entire function wrapped in try/except -- returns input unchanged on any error.
+    """
+    if not text or "<table" not in text.lower():
+        return text
+    try:
+        tbl_pat = re.compile(r"<table[^>]*>.*?</table>", re.DOTALL | re.IGNORECASE)
+        matches = list(tbl_pat.finditer(text))
+        if len(matches) < 2:
+            return text
+
+        da3_matches = [m for m in matches if _parse_da3_grid_from_table_html(m.group(0))]
+        if len(da3_matches) < 2:
+            return text
+
+        moves = []
+
+        for i in range(1, len(da3_matches)):
+            orphan_m = da3_matches[i]
+            orphan_start = orphan_m.start()
+            orphan_end = orphan_m.end()
+
+            # Guard 3: no debit heading already before the orphan table
+            before_window = text[max(0, orphan_start - _HEADING_SCAN_WINDOW): orphan_start]
+            before_clean = re.sub(r"<[^>]+>", " ", before_window)
+            lines_before = [l.strip() for l in before_clean.splitlines() if l.strip()]
+            last_meaningful = ""
+            for l in reversed(lines_before):
+                if len(l) > 3 and "---" not in l:
+                    last_meaningful = l
+                    break
+            if _DEBIT_HEADING_KEYWORDS.search(last_meaningful):
+                continue  # already has debit heading before it
+
+            # Guard 4: credit context must precede the orphan table
+            big_before = text[max(0, orphan_start - 1200): orphan_start]
+            if not _CREDIT_HEADING_KEYWORDS.search(re.sub(r"<[^>]+>", " ", big_before)):
+                continue
+
+            # Guard 5: no other table between orphan end and the potential debit heading
+            next_tbl_m = tbl_pat.search(text, orphan_end)
+
+            # Guard 2: find debit heading after the orphan table (within forward window)
+            dh_start = dh_end = None
+            for div_m in _CENTER_DIV_HDR_RE.finditer(text, orphan_end, orphan_end + _HEADING_SCAN_FORWARD):
+                div_clean = re.sub(r"<[^>]+>", "", div_m.group(1)).strip()
+                if _DEBIT_HEADING_KEYWORDS.search(div_clean) and not _CREDIT_HEADING_KEYWORDS.search(div_clean):
+                    # Guard 5: no table should be between orphan_end and this heading
+                    if next_tbl_m and next_tbl_m.start() < div_m.start():
+                        break  # another table intervenes
+                    dh_start = div_m.start()
+                    dh_end = div_m.end()
+                    break
+
+            if dh_start is None:
+                continue
+
+            moves.append((orphan_start, dh_start, dh_end))
+
+        if not moves:
+            return text
+
+        # Apply moves in reverse order to preserve offsets
+        for orphan_start, dh_start, dh_end in reversed(moves):
+            orphan_m2 = next((m for m in tbl_pat.finditer(text) if m.start() == orphan_start), None)
+            if not orphan_m2:
+                continue
+            orphan_end = orphan_m2.end()
+
+            heading_html = text[dh_start: dh_end]
+            gap = text[orphan_end: dh_start]
+            orphan_html = text[orphan_start: orphan_end]
+            part_before = text[:orphan_start]
+            part_after = text[dh_end:]
+
+            text = part_before + heading_html + "\n\n" + orphan_html + gap.rstrip() + part_after
+            log.info(
+                "Fix 29: moved debit heading (was at offset %d) to before orphan DA3 table (offset %d)",
+                dh_start, orphan_start,
+            )
+
+        return text
+
+    except Exception as e:
+        log.warning("_fix29_move_debit_heading_before_orphan_table failed: %s", e)
+        return text
+
+
 def stabilize_tables_and_text(page_md: str) -> str:
     if not page_md:
         return page_md
@@ -3535,6 +3657,7 @@ def stabilize_tables_and_text(page_md: str) -> str:
     stabilized = _dedupe_subset_transaction_tables_html(stabilized)
     stabilized = _drop_da3_misplaced_debit_after_deposit_heading_tables(stabilized)
     stabilized = _strip_credits_prefix_from_debit_da3_table(stabilized)  # Fix 28
+    stabilized = _fix29_move_debit_heading_before_orphan_table(stabilized)  # Fix 29
     stabilized = _strip_orphan_continued_da3_flatline(stabilized)
     stabilized = _strip_standalone_continued_banner_line(stabilized)
     return close_unclosed_html(stabilized)
@@ -3662,6 +3785,7 @@ def run_ocr(uploaded_file):
                     stabilized = _dedupe_subset_transaction_tables_html(stabilized)
                     stabilized = _drop_da3_misplaced_debit_after_deposit_heading_tables(stabilized)
                     stabilized = _strip_credits_prefix_from_debit_da3_table(stabilized)  # Fix 28
+                    stabilized = _fix29_move_debit_heading_before_orphan_table(stabilized)  # Fix 29
                     stabilized = _strip_standalone_continued_banner_line(stabilized)
                 elif is_pdf and is_nfcu_doc:
                     nfcu_summary = _extract_nfcu_summary_table(path, page_num)
@@ -3737,6 +3861,7 @@ def run_ocr(uploaded_file):
             merged = _dedupe_subset_transaction_tables_html(merged)
             merged = _drop_da3_misplaced_debit_after_deposit_heading_tables(merged)
             merged = _strip_credits_prefix_from_debit_da3_table(merged)  # Fix 28
+            merged = _fix29_move_debit_heading_before_orphan_table(merged)  # Fix 29
             merged = _strip_orphan_continued_da3_flatline(merged)
             merged = _strip_standalone_continued_banner_line(merged)
         return merged