Rakseeibra/arrr-erp
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix modern ICICI template parsing and parser fallback

Browse Source
This commit is contained in:
A R R R Associates
2026-08-04 13:10:24 +05:30
parent 565425c914
commit 79bfc808d3
3 changed files with 213 additions and 44 deletions
Download Patch File Download Diff File Expand all files Collapse all files
app/modules/bank_statement_analyzer/parsers/icici.py
+90 -8
View File
@@ -10,10 +10,15 @@ from .base import BaseParser, StatementMeta, amount, extract_text, finalize, nor
from .common import date_iso, find, infer_mode
_DATE_RE = re.compile(r"^\d{2}-\d{2}-\d{4}$")
_DATE_RE = re.compile(r"^(?:\d{2}-\d{2}-\d{4}|\d{2}-[A-Za-z]{3}-\d{4})$")
_MONEY_RE = re.compile(r"^-?(?:\d{1,3}(?:,\d{2,3})+|\d+)\.\d{2}$")
def _clean_date_cell(value: str) -> str:
text = norm(value)
return re.sub(r"\s*([/-])\s*", r"\1", text)
class ICICIParser(BaseParser):
"""Parser for ICICI Bank retail/Privilege PDF account statements.
@@ -34,12 +39,83 @@ class ICICIParser(BaseParser):
score += 0.55
if "STATEMENT OF TRANSACTIONS IN SAVINGS ACCOUNT" in upper:
score += 0.20
if "ACCOUNT STATEMENT" in upper and "TRANSACTION ID" in upper:
score += 0.20
if "AVAILABLE BALANCE" in upper and "WITHDRAWAL" in upper and "DEPOSIT" in upper:
score += 0.15
if all(token in upper for token in ("DEPOSITS", "WITHDRAWALS", "BALANCE")):
score += 0.20
if "ACCOUNT RELATED OTHER INFORMATION" in upper:
score += 0.05
return min(score, 0.99)
@staticmethod
def _modern_header_mapping(row: list) -> dict[str, int] | None:
cells = [norm(cell).upper() for cell in (row or [])]
aliases = {
"serial": ("S.NO", "S NO", "SERIAL NO"),
"transaction_id": ("TRANSACTION ID", "TXN ID"),
"transaction_date": ("TRANSACTION DATE", "TXN DATE"),
"cheque_no": ("CHEQUE NO", "CHQ NO"),
"description": ("DESCRIPTION", "PARTICULARS"),
"debit": ("WITHDRAWAL (DR)", "WITHDRAWAL", "DEBIT"),
"credit": ("DEPOSIT (CR)", "DEPOSIT", "CREDIT"),
"balance": ("AVAILABLE BALANCE", "BALANCE"),
}
mapping: dict[str, int] = {}
for field, names in aliases.items():
for index, cell in enumerate(cells):
if any(name == cell or name in cell for name in names):
mapping[field] = index
break
required = {"transaction_date", "description", "debit", "credit", "balance"}
return mapping if required.issubset(mapping) else None
def _parse_modern_tables(self, pdf) -> list[dict]:
rows: list[dict] = []
active_mapping: dict[str, int] | None = None
for page_number, page in enumerate(pdf.pages, start=1):
for table in page.extract_tables() or []:
if not table:
continue
start_index = 0
header_mapping = self._modern_header_mapping(table[0])
if header_mapping:
active_mapping = header_mapping
start_index = 1
if active_mapping is None:
continue
mapping = active_mapping
for raw in table[start_index:]:
cells = list(raw or [])
max_index = max(mapping.values())
if len(cells) <= max_index:
cells.extend([None] * (max_index + 1 - len(cells)))
transaction_date = _clean_date_cell(cells[mapping["transaction_date"]])
if not _DATE_RE.fullmatch(transaction_date):
continue
narration = norm(cells[mapping["description"]])
transaction_id = norm(cells[mapping.get("transaction_id", -1)]) if "transaction_id" in mapping else ""
cheque_no = norm(cells[mapping.get("cheque_no", -1)]) if "cheque_no" in mapping else ""
debit = amount(cells[mapping["debit"]])
credit = amount(cells[mapping["credit"]])
balance_value = amount(cells[mapping["balance"]])
if balance_value is None or (debit is None and credit is None):
continue
reference = cheque_no if cheque_no and cheque_no != "-" else transaction_id
rows.append({
"transaction_date": date_iso(transaction_date),
"value_date": date_iso(transaction_date),
"narration": narration,
"reference_no": reference,
"debit": debit,
"credit": credit,
"balance": balance_value,
"source_page": page_number,
"mode": infer_mode(narration),
})
return rows
@staticmethod
def _table_geometry(page, words: list[dict]) -> tuple[float, list[float]] | None:
required = {"DATE", "PARTICULARS", "DEPOSITS", "WITHDRAWALS", "BALANCE"}
@@ -188,10 +264,14 @@ class ICICIParser(BaseParser):
name_match = re.search(r"(?m)^\s*((?:MS|MR|MRS)\.?\s*[^\n]+)$", text, re.I)
meta.customer_name = norm(re.split(r"\s{2,}", name_match.group(1).strip())[0]) if name_match else ""
meta.customer_id = find(r"Cust ID\s*:\s*([0-9]+)", text)
meta.account_number = find(r"Savings Account Number\s*:\s*([0-9]+)", text)
meta.customer_id = find(r"(?:Cust ID|Customer ID)\s*:\s*([0-9]+)", text)
meta.account_number = find(r"(?:Savings Account Number|Account number)\s*:\s*([0-9]+)", text, flags=re.I)
if not meta.account_number:
meta.account_number = find(r"Savings\s+A/c\s+([0-9]+)", text)
if not meta.customer_name:
meta.customer_name = norm(find(r"Account name\s*:\s*([^\n]+)", text, flags=re.I))
if not meta.ifsc:
meta.ifsc = find(r"IFSC code\s*:\s*([A-Z0-9]+)", text, flags=re.I)
meta.ifsc = find(r"IFSC CODE\s+NAME OF NOMINEE.*?Savings\s+[0-9]+\s+[0-9]+\s+([A-Z0-9]+)", text, flags=re.I | re.S)
period = re.search(
@@ -205,11 +285,13 @@ class ICICIParser(BaseParser):
all_rows: list[dict] = []
with pdfplumber.open(str(pdf_path)) as pdf:
for page_number, page in enumerate(pdf.pages, start=1):
page_rows, page_opening = self._parse_page(page, page_number)
if meta.opening_balance is None and page_opening is not None:
meta.opening_balance = page_opening
all_rows.extend(page_rows)
all_rows = self._parse_modern_tables(pdf)
if not all_rows:
for page_number, page in enumerate(pdf.pages, start=1):
page_rows, page_opening = self._parse_page(page, page_number)
if meta.opening_balance is None and page_opening is not None:
meta.opening_balance = page_opening
all_rows.extend(page_rows)
data = pd.DataFrame(all_rows)
if not data.empty: