# Checkpoint 335 — Idempotent mapped CSV bank imports without bank transaction IDs

**Date:** 2026-10-27  
**Story:** BE-080 — Reconcile bank payments and receipts (**PARTIAL**)

## Change

Mapped CSV/Excel rows without a bank-provided transaction ID now receive a deterministic SHA-256 import ID derived from the normalized date, amount, currency, description, reference, narration, partner and counterparty account. An occurrence counter distinguishes repeated identical rows in the same file. Re-importing the same transactions therefore skips existing rows while preserving legitimate equal transactions from the first import. Explicit bank IDs continue to be used unchanged.

## Verification

- Parser regression confirms fallback IDs are stable across parses and that repeated equal rows get distinct IDs.
- Native Accounting KSA import regression imports three CSV rows (including two identical transactions) and verifies all three persist with unique IDs; importing the same file again is rejected as already imported; a later file containing old rows and one new transaction imports only the new row.
- The full `TestCsvParser` class plus the native mapped-CSV re-import regression passed **12 tests, 0 failures/errors** on a disposable clone of `namaa_ksa`. The clone was dropped; the shared database was unchanged.
- Evidence: [isolated QA](KSA_BANK_CSV_IMPORT_QA.json), [parser](../../addons/accounting_ksa/features/bank_import/parsers/csv_file.py), and [native import tests](../../addons/accounting_ksa/features/bank_import/tests/test_statement_import.py).

## Remaining acceptance

BE-080 remains PARTIAL. Fallback fingerprints cannot distinguish a legitimate changed transaction from an amended bank export when the bank supplies no stable ID and all mapped transaction fields match. Bank-specific maps/samples, banking-role UAT, broader reconciliation policy variants and full A1–A4 acceptance remain open.
