Update app.py

app.py

@@ -48,6 +48,10 @@ Primary goals for reconcile rate:
             (≤6 rows), append those rows to the preceding debits table and drop the extra ED heading.
             Skip the *first* Electronic Debits + table on the page when the previous table is ITM/credits-only
             and the fragment is clearly debits (so we do not treat the main debits section as a duplicate).
+- Fix 17h: UCB sometimes OCRs Electronic Credits (+ optional Electronic Debits) then the *main debits* DA3
+            table first and a *second* DA3 table whose rows start with ITM DEPOSIT (credits) plus duplicate
+            debits. Reorder to Credits + ITM table + Electronic Debits + merged debits; drop dup keys from the
+            second table against the first; merge remaining orphan rows into the debits table.
 - Fix 17e: Repeat EC/ED fused + 17d passes until stable (handles chained replacements). Centered section titles
             match both align=\"center\" and style=\"text-align:center\" on divs.
 - Post-pass: dedupe 3-col Date|Description|Amount tables when one is a duplicate fragment
@@ -2944,6 +2948,17 @@ def _is_strong_ucb_electronic_debit_row(grid, r: int) -> bool:
     return False
 
 
+def _ucb_grid_has_any_strong_debit_row(grid) -> bool:
+    if not grid:
+        return False
+    for r in range(1, len(grid)):
+        if not any(str(c or "").strip() for c in grid[r]):
+            continue
+        if _is_strong_ucb_electronic_debit_row(grid, r):
+            return True
+    return False
+
+
 def _ucb_has_itm_deposit_data_row(grid) -> bool:
     for r in range(1, len(grid)):
         if _is_itm_deposit_row_ucb(grid, r):
@@ -3088,6 +3103,111 @@ _UCB_EC_ED_SINGLE_FUSED_TABLE = re.compile(
     re.DOTALL | re.IGNORECASE,
 )
 
+# Fix 17h: debits DA3 first, then DA3 with leading ITM + duplicate tail (no ED between tables).
+_UCB_EC_TWO_DA3_DEBITS_BEFORE_ITM_TAIL = re.compile(
+    r'(?P<lead>(?:' + _UCB_CENTER_DIV + r'\s*</div>\s*)?)'
+    r'(?P<ecdiv>' + _UCB_CENTER_DIV + r'\s*Electronic\s+Credits\s*</div>\s*)'
+    r'(?P<ed_maybe>(?:' + _UCB_CENTER_DIV + r'\s*Electronic\s+Debits\s*</div>\s*)?)'
+    r'(?P<t1><table[^>]*>.*?</table>)'
+    r'\s*'
+    r'(?P<t2><table[^>]*>.*?</table>)',
+    re.DOTALL | re.IGNORECASE,
+)
+
+
+def _ucb_reorder_ec_ed_debits_first_itm_tail_html(text: str) -> str:
+    """
+    Fix 17h: Electronic Credits (+ optional Electronic Debits) then two DA3 tables where the first is
+    clearly debits and the second begins with a run of ITM DEPOSIT rows (mis-ordered vs credits section).
+    """
+    if not text or "electronic credits" not in text.lower():
+        return text
+    low = text.lower()
+    if "itm deposit" not in low and "itm dep" not in low:
+        return text
+    pat = _UCB_EC_TWO_DA3_DEBITS_BEFORE_ITM_TAIL
+    _dn = re.compile(r"^\d{1,2}[/\-]\d{2}")
+    _amt_ok = re.compile(r"^-?\d+(?:\.\d{1,4})?-?$")
+
+    def _key_cells(cells):
+        parts = [str(c or "").strip() for c in cells[:3]]
+        while len(parts) < 3:
+            parts.append("")
+        date_s = re.sub(r"\s+", "", parts[0])
+        amt_s = re.sub(r"[\s$,]", "", parts[2]).lower()
+        amt_s = amt_s.replace("\u2212", "-").replace("\u2013", "-")
+        return date_s, amt_s
+
+    def _row_loose_key(grid, r):
+        cells = [str(c or "").strip() for c in grid[r]]
+        if not any(cells):
+            return None
+        k = _key_cells(cells)
+        if not (_dn.match(k[0]) and _amt_ok.match(k[1]) and "." in k[1]):
+            return None
+        return k
+
+    m = pat.search(text)
+    if not m:
+        return text
+    g1 = _parse_da3_grid_from_table_html(m.group("t1"))
+    g2 = _parse_da3_grid_from_table_html(m.group("t2"))
+    if not g1 or not g2 or not _is_da3_header(g1) or not _is_da3_header(g2):
+        return text
+    if not _ucb_grid_has_any_strong_debit_row(g1):
+        return text
+    itm_end = 1
+    while itm_end < len(g2):
+        if not any(str(c or "").strip() for c in g2[itm_end]):
+            itm_end += 1
+            continue
+        if not _is_itm_deposit_row_ucb(g2, itm_end):
+            break
+        itm_end += 1
+    if itm_end <= 1:
+        return text
+    cb = Counter()
+    for r in range(1, len(g1)):
+        kk = _row_loose_key(g1, r)
+        if kk:
+            cb[kk] += 1
+    new_tail = []
+    for r in range(itm_end, len(g2)):
+        if not any(str(c or "").strip() for c in g2[r]):
+            continue
+        if _is_itm_deposit_row_ucb(g2, r):
+            continue
+        kk = _row_loose_key(g2, r)
+        if kk and cb.get(kk, 0) > 0:
+            cb[kk] -= 1
+            continue
+        new_tail.append(list(g2[r]))
+    g_itm = [list(g2[0])] + [list(g2[r]) for r in range(1, itm_end)]
+    merged = [list(g1[0])] + [list(r) for r in g1[1:]] + new_tail
+    max_c = max(len(r) for r in merged)
+    for row in merged:
+        while len(row) < max_c:
+            row.append("")
+    max_ci = max(len(r) for r in g_itm)
+    for row in g_itm:
+        while len(row) < max_ci:
+            row.append("")
+    sect = '<div align="center">\n\n{title}\n\n</div>\n\n'
+    parts = []
+    lead = m.group("lead") or ""
+    if lead.strip():
+        parts.append(lead.rstrip())
+    parts.extend(
+        [
+            sect.format(title="Electronic Credits"),
+            _grid_to_html(g_itm),
+            sect.format(title="Electronic Debits"),
+            _grid_to_html(merged),
+        ]
+    )
+    new_blob = "\n\n".join(parts)
+    return text[: m.start()] + new_blob + text[m.end() :]
+
 
 def _ucb_strip_redundant_ed_subset_fragment(text: str) -> str:
     """
@@ -3247,14 +3367,6 @@ def _ucb_prune_ed_fragment_dupes_vs_prev_table(text: str) -> str:
                 return None
             return k
 
-        def _grid_has_strong_debit_row(grid):
-            for r in range(1, len(grid)):
-                if not any(str(c or "").strip() for c in grid[r]):
-                    continue
-                if _is_strong_ucb_electronic_debit_row(grid, r):
-                    return True
-            return False
-
         for _ in range(12):
             changed = False
             for m in ed_block.finditer(text):
@@ -3274,7 +3386,7 @@ def _ucb_prune_ed_fragment_dupes_vs_prev_table(text: str) -> str:
                 ):
                     continue
                 # Legitimate layout: ITM/credits-only table, then Electronic Debits + main debits — do not prune.
-                if not _grid_has_strong_debit_row(g_big) and _grid_has_strong_debit_row(g_frag):
+                if not _ucb_grid_has_any_strong_debit_row(g_big) and _ucb_grid_has_any_strong_debit_row(g_frag):
                     continue
                 cb = Counter()
                 for r in range(1, len(g_big)):
@@ -3389,9 +3501,10 @@ def _split_ucb_deposits_electronic_sections_html(text: str) -> str:
             return new
 
         text = _UCB_DEP_CONT_BLOCK.sub(_repl, text)
-    # Fix 17e: repeat fused EC/ED fix + duplicate ED strip + 17g prune until stable.
+    # Fix 17e: repeat 17h reorder + fused EC/ED fix + duplicate ED strip + 17g prune until stable.
     for _ in range(16):
-        nxt = _split_ucb_fused_ec_ed_headings_single_table_html(text)
+        nxt = _ucb_reorder_ec_ed_debits_first_itm_tail_html(text)
+        nxt = _split_ucb_fused_ec_ed_headings_single_table_html(nxt)
         nxt = _ucb_strip_redundant_ed_subset_fragment(nxt)
         nxt = _ucb_prune_ed_fragment_dupes_vs_prev_table(nxt)
         if nxt == text: