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

Add independent bank identity detection and APGB parser

Browse Source
This commit is contained in:
A R R R Associates
2026-08-04 13:50:52 +05:30
parent 79bfc808d3
commit 0f6e4126e8
3 changed files with 420 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files
app/modules/bank_statement_analyzer/parsers/registry.py
+43 -103
View File
@@ -1,9 +1,5 @@
from __future__ import annotations
import logging
logger = logging.getLogger(__name__)
from .idfc import IDFCFirstParser
from .axis import AxisParser
from .hdfc import HDFCParser
@@ -18,12 +14,15 @@ from .yes_bank import YesBankParser
from .city_union_bank import CityUnionBankParser
from .bank_of_baroda import BankOfBarodaParser
from .rbl_bank import RBLBankParser
from .andhra_pradesh_grameena_bank import AndhraPradeshGrameenaBankParser
from .bank_identity import apply_detected_bank_identity
from .base import extract_text
from .template_engine import parse_with_template
PARSERS = [
IDFCFirstParser,
AxisParser,
AndhraPradeshGrameenaBankParser,
CentralBankOfIndiaParser,
YesBankParser,
CityUnionBankParser,
@@ -43,6 +42,7 @@ PARSERS = [
BANK_OPTIONS = [
("auto", "Auto Detect"),
("axis", "Axis Bank"),
("andhra_pradesh_grameena_bank", "Andhra Pradesh Grameena Bank"),
("central_bank_of_india", "Central Bank of India"),
("yes_bank", "YES Bank"),
("city_union_bank", "City Union Bank"),
@@ -60,6 +60,7 @@ BANK_OPTIONS = [
BANK_PARSERS = {
"axis": [AxisParser],
"andhra_pradesh_grameena_bank": [AndhraPradeshGrameenaBankParser],
"central_bank_of_india": [CentralBankOfIndiaParser],
"yes_bank": [YesBankParser],
"city_union_bank": [CityUnionBankParser],
@@ -77,7 +78,11 @@ BANK_PARSERS = {
def detect_parser(text):
scored = sorted(((parser.detect(text), parser) for parser in PARSERS), key=lambda item: item[0], reverse=True)
scored = sorted(
((parser.detect(text), parser) for parser in PARSERS),
key=lambda item: item[0],
reverse=True,
)
if not scored or scored[0][0] <= 0:
return None, 0
return scored[0][1](), scored[0][0]
@@ -87,115 +92,50 @@ def _selected_parser(bank_key: str, text: str):
candidates = BANK_PARSERS.get(bank_key, [])
if not candidates:
return None, 0
scored = sorted(((parser.detect(text), parser) for parser in candidates), key=lambda item: item[0], reverse=True)
scored = sorted(
((parser.detect(text), parser) for parser in candidates),
key=lambda item: item[0],
reverse=True,
)
if scored and scored[0][0] > 0:
return scored[0][1](), scored[0][0]
return None, 0
def _usable_result(result) -> bool:
if result is None:
return False
try:
_meta, frame = result
return frame is not None and not frame.empty
except Exception:
return False
def _attempt(parser, path, text: str, route: str):
try:
result = parser.parse(path, text)
if _usable_result(result):
meta, frame = result
logger.info(
"Bank analyzer route accepted: route=%s parser=%s rows=%s file=%s",
route, getattr(meta, "parser_name", parser.__class__.__name__), len(frame), path,
)
return result
logger.warning(
"Bank analyzer route returned no rows: route=%s parser=%s file=%s",
route, parser.__class__.__name__, path,
)
except Exception as exc:
logger.warning(
"Bank analyzer route failed: route=%s parser=%s file=%s error=%s",
route, parser.__class__.__name__, path, exc, exc_info=True,
)
return None
def _template_first(path, text: str, hint: str):
try:
result = parse_with_template(path, text=text, bank_hint=hint)
if _usable_result(result):
meta, frame = result
logger.info(
"Bank analyzer template accepted: parser=%s rows=%s file=%s",
getattr(meta, "parser_name", "TemplateBasedStatementParser"), len(frame), path,
)
return result
logger.warning("Bank analyzer template returned no rows: file=%s", path)
except Exception as exc:
logger.warning(
"Bank analyzer template rejected: file=%s error=%s", path, exc, exc_info=True,
)
return None
def _ordered_bank_candidates(text: str):
scored = sorted(
((parser.detect(text), parser) for parser in PARSERS),
key=lambda item: item[0],
reverse=True,
)
return [parser for score, parser in scored if score > 0]
def _parse_and_identify(parser, path, text: str):
meta, df = parser.parse(path, text)
return apply_detected_bank_identity(meta, df, text)
def parse_pdf(path, bank_hint: str | None = None):
hint = (bank_hint or "auto").strip().lower()
text = extract_text(path)
template_result = _template_first(path, text, hint)
if template_result is not None:
return template_result
if hint == "hsbc":
meta, df = HSBCParser().parse(path, text)
return apply_detected_bank_identity(meta, df, text)
if hint and hint != "auto":
candidates = BANK_PARSERS.get(hint, [])
for parser_class in sorted(candidates, key=lambda cls: cls.detect(text), reverse=True):
result = _attempt(parser_class(), path, text, f"selected:{hint}")
if result is not None:
return result
label = dict(BANK_OPTIONS).get(hint, "the selected bank")
parser, score = _selected_parser(hint, text)
if parser is None:
label = dict(BANK_OPTIONS).get(hint, "the selected bank")
raise ValueError(
f"The uploaded statement does not match the selected {label} format. "
"Please verify the selected bank or choose Auto Detect."
)
return _parse_and_identify(parser, path, text)
parser, score = detect_parser(text)
if parser is None and not (text or "").strip():
# Image-only statements cannot be identified by pdftotext. HSBCParser
# performs its own OCR identification and raises a precise mismatch error.
try:
meta, df = HSBCParser().parse(path, text)
return apply_detected_bank_identity(meta, df, text)
except ValueError:
pass
if parser is None:
raise ValueError(
f"The uploaded statement could not be parsed by a validated layout template or the "
f"selected {label} parser. Please verify that the PDF is readable and supported."
"Unsupported statement format. Select the bank manually or add a bank-specific parser for this statement layout."
)
attempted: set[type] = set()
for parser_class in _ordered_bank_candidates(text):
attempted.add(parser_class)
result = _attempt(parser_class(), path, text, "auto-detected")
if result is not None:
return result
if not (text or "").strip():
result = _attempt(HSBCParser(), path, text, "image-only-ocr")
if result is not None:
return result
# Last-resort compatibility pass: a detector may miss a new layout even
# though the bank-specific parser can parse its visual table correctly.
for parser_class in PARSERS:
if parser_class in attempted:
continue
result = _attempt(parser_class(), path, text, "compatibility-fallback")
if result is not None:
return result
raise ValueError(
"The bank statement format was identified, but no transaction rows could be extracted. "
"The template engine and all bank-specific parsers were attempted. "
"Please verify that the PDF is text-readable and that this statement layout is supported."
)
return _parse_and_identify(parser, path, text)