Update app.py

app.py

@@ -1730,6 +1730,84 @@ def _split_fused_multicolumn_header(grid):
     return new_grid
 
 
+def _normalize_daily_balance_table(grid):
+    """
+    Fix OCR artifact in DAILY BALANCE SUMMARY tables where:
+    - OCR produces N DATE columns followed by N BALANCE columns
+      instead of interleaved DATE|BALANCE|DATE|BALANCE pairs
+    - Multiple dates are stacked in one cell "01/02\n01/03\n01/04\n01/05"
+
+    Reorders columns and expands stacked date cells into proper rows.
+    Only fires when ALL of these hold:
+    - Header has exactly 2N columns where N >= 2
+    - First N columns are all DATE, last N columns are all BALANCE
+    - At least one data cell contains a newline (stacked dates)
+    """
+    if not grid or len(grid) < 2:
+        return grid
+
+    header = [str(c or "").strip().upper() for c in grid[0]]
+    ncols = len(header)
+
+    if ncols < 4 or ncols % 2 != 0:
+        return grid
+
+    half = ncols // 2
+    date_half  = header[:half]
+    bal_half   = header[half:]
+
+    if not all(h == "DATE" for h in date_half):
+        return grid
+    if not all(h == "BALANCE" for h in bal_half):
+        return grid
+
+    # Build new interleaved header: DATE BALANCE DATE BALANCE ...
+    new_header = []
+    for i in range(half):
+        new_header.append("DATE")
+        new_header.append("BALANCE")
+
+    # Expand each data row
+    new_rows = []
+    _DATE_RE3 = re.compile(r"^\d{1,2}/\d{2}(?:/\d{2,4})?$")
+
+    for row in grid[1:]:
+        cells = [str(c or "").strip() for c in row]
+        while len(cells) < ncols:
+            cells.append("")
+
+        # Dates are all stacked in cells[0]; balances follow immediately in cells[1..N]
+        # (OCR layout: stacked_dates | bal1 | bal2 | bal3 | bal4 | empty...)
+        # NOT: date | date | date | date | bal | bal | bal | bal (despite header order)
+        date_cells = [cells[0]]
+        bal_cells  = cells[1:]
+
+        # Check if dates are stacked in first cell (newline OR space-separated)
+        # e.g. "01/02 01/03 01/04 01/05" or "01/02\n01/03\n01/04\n01/05"
+        first = date_cells[0]
+        tokens = re.split(r"[\n\r\s]+", first)
+        date_tokens = [t.strip() for t in tokens if t.strip() and _DATE_RE3.match(t.strip())]
+        if len(date_tokens) >= 2:
+            stacked = date_tokens
+        else:
+            stacked = date_cells  # already one date per cell
+
+        # Build ONE interleaved row for this group of stacked dates
+        new_row = [""] * ncols
+        for i, date_val in enumerate(stacked):
+            if not _DATE_RE3.match(date_val):
+                continue
+            bal_val = bal_cells[i] if i < len(bal_cells) else ""
+            new_row[i * 2]     = date_val
+            new_row[i * 2 + 1] = bal_val
+        new_rows.append(new_row)
+
+    if not new_rows:
+        return grid
+
+    return [new_header] + new_rows
+
+
 def normalize_html_tables(text: str) -> str:
     """
     For every <table>...</table>:
@@ -1759,8 +1837,9 @@ def normalize_html_tables(text: str) -> str:
 
             # Suppress known junk tables (OCR artifacts from page headers)
             if _is_junk_table(grid):
+                # Append nothing (suppress the table), update last, skip rest of pipeline
                 out.append("")
-                last = m.end()
+                last = m.end()  # must set here before continue since line below is outside try
                 continue
 
             grid = _drop_truly_empty_columns(grid)
@@ -1780,6 +1859,9 @@ def normalize_html_tables(text: str) -> str:
             # Fix 8: split fused multi-keyword header cell (First Horizon Bank)
             grid = _split_fused_multicolumn_header(grid)
 
+            # Fix 9: normalize DAILY BALANCE SUMMARY tables with stacked dates
+            grid = _normalize_daily_balance_table(grid)
+
             # Fix 6: merge split rows where description wraps to next line
             # (Hardin County Bank and similar monospace statement formats)
             grid = _merge_split_rows(grid)