Update app.py

app.py

@@ -977,10 +977,11 @@ def _extract_textlayer_rows(pdf_path: str, page_num: int):
             line_words = sorted(lines_by_y[y], key=lambda t: t[0])
             sorted_lines.append([w for _, w in line_words])
 
-        # Date patterns — numeric (MM/DD variants) or month-name (Jan 16)
+        # Date patterns — numeric (MM/DD, MM-DD variants) or month-name (Jan 16)
         _MONTHS = r"(?:Jan|Feb|Mar|Apr|May|Jun|Jul|Aug|Sep|Oct|Nov|Dec)"
         date_re = re.compile(
             r"^\d{1,2}/\d{2}(?:/\d{2,4})?$"          # MM/DD, MM/DD/YY, MM/DD/YYYY
+            r"|^\d{1,2}-\d{2}(?:-\d{2,4})?$"           # MM-DD, MM-DD-YY (East West Bank)
             r"|^\d{4}-\d{2}-\d{2}$"                    # YYYY-MM-DD
             r"|^" + _MONTHS + r"$",                        # "Jan", "Feb" etc (month-name date part 1)
             re.IGNORECASE,
@@ -1553,6 +1554,8 @@ def run_ocr(uploaded_file):
                     ftr = ocr_zone(page_images[page_num], fs, 1.0)
                 if ftr and ftr.strip():
                     ftr_clean = normalize_money_glyphs(ftr.strip())
+
+                    # Guard 1: first-line dedup (original check)
                     ftr_first_line = next(
                         (ln.strip().lower() for ln in ftr_clean.splitlines() if ln.strip()),
                         ""
@@ -1560,7 +1563,19 @@ def run_ocr(uploaded_file):
                     already_present = ftr_first_line and any(
                         ftr_first_line in part.lower() for part in parts
                     )
-                    if not already_present:
+
+                    # Guard 2: suppress footer that is a raw text-layer dump of
+                    # transaction rows (e.g. East West Bank two-column layout).
+                    # Detected when the footer contains 3+ date tokens (MM/DD or MM-DD)
+                    # AND the page already has an OCR table — the table already
+                    # captured this data correctly, so the footer text is redundant.
+                    _footer_date_re = re.compile(r"\b\d{1,2}[-/]\d{2}\b")
+                    is_txn_dump = (
+                        len(_footer_date_re.findall(ftr_clean)) >= 3
+                        and any("<table" in part.lower() for part in parts)
+                    )
+
+                    if not already_present and not is_txn_dump:
                         parts.append(ftr_clean)
 
             if parts: