feat(pdf): extract statement header (account + period) + date format

Two related additions for the accountant workflow:

**1. Statement header extraction.** New
``extract_statement_metadata(pages)`` pulls the account number
and statement period out of the first page (falls back to
page 1+2 if either is missing on page 1 — Wells Fargo business
accounts put header info on page 2). Detected fields are
stamped onto EVERY transaction row so a multi-statement CSV is
self-attributing per row::

    {
      "date": "20250113",
      "description": "Coffee Shop",
      "amount_1": -4.50,
      "account_number": "****5678",
      "statement_period_start": "20250101",
      "statement_period_end": "20250131",
      ...
    }

Account-number regex is tolerant of masks (``****1234``),
hyphens (``1234-5678-9012``), and spaces. Period regex looks
for "Statement Period" / "From" / "Period Covered" labels plus
the first 1-2 full-year dates that follow. If only one date is
present near the label, it's used for both start and end (some
statements show only the closing date).

**2. Year inference for short dates.** When the row date is a
short ``01/13`` or ``Jan 13`` without a year, the scanner now
binds the year from the statement period's end date BEFORE
formatting. Doesn't handle the December-in-January-statement
cross-year case (rare; user can edit in the table).

**3. Configurable output date format.** New
``output_date_format`` parameter on ``scan_pdf_for_transactions``
defaults to ``%Y%m%d``. Applied to: the transaction date column
AND the statement period start/end fields. The page surfaces a
dropdown in Scan options with common presets (YYYYMMDD,
YYYY-MM-DD, MM/DD/YYYY, DD/MM/YYYY, ``Mon DD, YYYY``) plus a
Custom option that accepts a raw strftime string.

New helper: ``format_date(iso_str, fmt)`` converts ISO
``YYYY-MM-DD`` to any strftime; passes invalid input through
unchanged so the user can see what was actually there rather
than getting silent empties.

20 new tests cover: format_date, account-number extraction
(masked / hyphenated / spaced / no-label / short), period
extraction (standard / from-to / single-date / no-label),
metadata orchestrator (full header / no pages / page-2
fallback), year inference (US / dash / month-name / no-period /
unparseable), plus an end-to-end class that builds a header'd
PDF with short-date transactions and confirms metadata
attribution + year inference + format round-trip.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-05-20 00:20:46 +00:00
parent 3cf935c999
commit 155dd30746
4 changed files with 499 additions and 13 deletions

View File

@@ -15,9 +15,14 @@ from __future__ import annotations
from src.pdf_extract import (
Page,
WordBox,
_extract_account_number,
_extract_statement_period,
_find_amount_tokens,
_find_dates_in_words,
_infer_year_for_short_date,
cluster_rows,
extract_statement_metadata,
format_date,
parse_amount,
parse_date,
)
@@ -207,3 +212,134 @@ class TestFindAmountTokens:
# test module — they need ``scan_pdf_for_transactions`` which in
# turn uses ``extract_pages_auto``. The unit-test layer here pins
# the building blocks; smoke tests pin the wiring.
class TestFormatDate:
def test_yyyymmdd(self):
assert format_date("2026-01-13", "%Y%m%d") == "20260113"
def test_iso_passthrough(self):
assert format_date("2026-01-13", "%Y-%m-%d") == "2026-01-13"
def test_us(self):
assert format_date("2026-01-13", "%m/%d/%Y") == "01/13/2026"
def test_invalid_input_passes_through(self):
# Non-ISO input — return as-is so the user sees what was
# actually there rather than a silent empty string.
assert format_date("01/13", "%Y%m%d") == "01/13"
def test_none_or_empty(self):
assert format_date(None) == ""
assert format_date("") == ""
class TestExtractAccountNumber:
def test_masked(self):
text = "Customer Name\nAccount Number: ****1234\nBalance"
assert _extract_account_number(text) == "****1234"
def test_with_hyphens(self):
text = "Account #: 1234-5678-9012"
assert _extract_account_number(text) == "1234-5678-9012"
def test_with_spaces(self):
text = "Account: 1234 5678 9012"
assert _extract_account_number(text) == "1234 5678 9012"
def test_no_label_no_match(self):
text = "Just some text with 1234567890 in it"
assert _extract_account_number(text) is None
def test_requires_at_least_four_digits(self):
# An "account" label followed by only XX shouldn't count.
text = "Account: XX"
assert _extract_account_number(text) is None
class TestExtractStatementPeriod:
def test_standard_period(self):
text = "Statement Period: 01/01/2025 - 01/31/2025\nBalance"
start, end = _extract_statement_period(text)
assert start == "2025-01-01"
assert end == "2025-01-31"
def test_from_to(self):
text = "From 01/01/2025 to 01/31/2025"
start, end = _extract_statement_period(text)
assert start == "2025-01-01"
assert end == "2025-01-31"
def test_single_date_both_fields(self):
# When only one date appears near the label, return it for both.
text = "Statement Date: 01/31/2025"
start, end = _extract_statement_period(text)
assert start == "2025-01-31"
assert end == "2025-01-31"
def test_no_label_no_match(self):
text = "Some random text with 01/01/2025 in it"
start, end = _extract_statement_period(text)
# No "Period" / "From" / "Statement Date" label
assert (start, end) == (None, None)
class TestExtractStatementMetadata:
def test_full_header(self):
pages = [Page(
page_no=1, width=600, height=800,
text=(
"ACME BANK\n"
"Customer: John Doe\n"
"Account Number: ****5678\n"
"Statement Period: 01/01/2025 - 01/31/2025\n"
"Beginning balance: $1,000.00\n"
),
words=[],
)]
meta = extract_statement_metadata(pages)
assert meta["account_number"] == "****5678"
assert meta["period_start"] == "2025-01-01"
assert meta["period_end"] == "2025-01-31"
def test_no_pages(self):
meta = extract_statement_metadata([])
assert meta == {
"account_number": None,
"period_start": None,
"period_end": None,
}
def test_fallback_to_page_two(self):
# Page 1 has only account; period is on page 2.
p1 = Page(
page_no=1, width=600, height=800,
text="Account Number: ****1234\nBalance summary",
words=[],
)
p2 = Page(
page_no=2, width=600, height=800,
text="Statement Period: 02/01/2025 - 02/28/2025",
words=[],
)
meta = extract_statement_metadata([p1, p2])
assert meta["account_number"] == "****1234"
assert meta["period_start"] == "2025-02-01"
assert meta["period_end"] == "2025-02-28"
class TestInferYearForShortDate:
def test_us_short_with_period_end(self):
assert _infer_year_for_short_date("01/13", "2025-01-31") == "2025-01-13"
def test_short_dash(self):
assert _infer_year_for_short_date("01-13", "2025-01-31") == "2025-01-13"
def test_month_name(self):
assert _infer_year_for_short_date("Jan 13", "2025-01-31") == "2025-01-13"
def test_no_period_end(self):
assert _infer_year_for_short_date("01/13", None) is None
def test_unparseable(self):
assert _infer_year_for_short_date("xx/yy", "2025-01-31") is None

View File

@@ -15,6 +15,45 @@ from __future__ import annotations
import pytest
def _build_statement_pdf_with_header() -> bytes:
"""Statement with realistic header (account + period) plus
transactions. Exercises the metadata-extraction path end-to-end."""
from fpdf import FPDF
pdf = FPDF(orientation="P", unit="pt", format="letter")
pdf.add_page()
pdf.set_font("Helvetica", size=12)
pdf.set_xy(40, 50)
pdf.cell(0, 14, "ACME BANK STATEMENT", new_x="LMARGIN", new_y="NEXT")
pdf.set_xy(40, 70)
pdf.cell(0, 14, "Account Number: ****5678", new_x="LMARGIN", new_y="NEXT")
pdf.set_xy(40, 85)
pdf.cell(0, 14, "Statement Period: 01/01/2025 - 01/31/2025",
new_x="LMARGIN", new_y="NEXT")
# Header row
pdf.set_xy(40, 130)
pdf.cell(120, 14, "Date")
pdf.set_xy(160, 130)
pdf.cell(200, 14, "Description")
pdf.set_xy(360, 130)
pdf.cell(80, 14, "Amount")
# Transactions with SHORT dates — year is implied by period.
rows = [
("01/13", "Coffee Shop", "(4.50)"),
("01/16", "Refund Vendor", "$12.00"),
]
y = 160
for date, desc, amt in rows:
pdf.set_xy(40, y)
pdf.cell(120, 14, date)
pdf.set_xy(160, y)
pdf.cell(200, 14, desc)
pdf.set_xy(360, y)
pdf.cell(80, 14, amt)
y += 20
return bytes(pdf.output())
def _build_tiny_statement_pdf() -> bytes:
"""One-page PDF: header line + three transaction rows + a
closing-balance footer. The scanner should pick up exactly the
@@ -97,13 +136,34 @@ class TestScanPdfForTransactions:
f"{[r.get('raw') for r in rows]}"
)
def test_parses_dates_to_iso(self, pdf_bytes):
def test_dates_formatted_yyyymmdd_by_default(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(pdf_bytes)
# Default output format is %Y%m%d
assert [r["date"] for r in rows] == [
"20260115", "20260116", "20260117",
]
def test_output_date_format_override(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(
pdf_bytes, output_date_format="%Y-%m-%d",
)
assert [r["date"] for r in rows] == [
"2026-01-15", "2026-01-16", "2026-01-17",
]
def test_metadata_fields_present_on_every_row(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(pdf_bytes)
# The fixture PDF has no statement-period or account
# header, so the metadata fields exist but are empty
# strings — the contract is: ALWAYS present on every row.
for r in rows:
assert "account_number" in r
assert "statement_period_start" in r
assert "statement_period_end" in r
def test_parses_amounts_with_signs(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(pdf_bytes)
@@ -144,6 +204,42 @@ class TestScanPdfForTransactions:
# ---------------------------------------------------------------------------
class TestStatementHeaderEndToEnd:
"""A real PDF with a real header — exercise the full pipeline:
metadata extraction + year inference for short dates + format
application. This is the failure mode most likely to break on
the user's actual Chase statements."""
@pytest.fixture
def pdf_bytes(self) -> bytes:
return _build_statement_pdf_with_header()
def test_metadata_extracted_and_stamped(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(pdf_bytes)
assert rows, "expected at least one transaction"
for r in rows:
assert r["account_number"] == "****5678"
assert r["statement_period_start"] == "20250101"
assert r["statement_period_end"] == "20250131"
def test_short_dates_get_year_from_period(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(pdf_bytes)
# Short ``01/13`` + period ending in 2025 → 20250113
assert rows[0]["date"] == "20250113"
assert rows[1]["date"] == "20250116"
def test_iso_format_round_trip(self, pdf_bytes):
from src.pdf_extract import scan_pdf_for_transactions
rows, _ = scan_pdf_for_transactions(
pdf_bytes, output_date_format="%Y-%m-%d",
)
assert rows[0]["date"] == "2025-01-13"
assert rows[0]["statement_period_start"] == "2025-01-01"
assert rows[0]["statement_period_end"] == "2025-01-31"
class TestMultiDateRow:
"""Some statements (Chase, BofA) show both a transaction date
and a posting date per row. The scanner uses the first date