Files
aurum/backend/tests
ZProger 19f5cfa429 feat: import transactions from a CSV bank statement
Fourth of the four ideas picked from the Maybe/Sure feature audit, and the
last one queued up. A 3-step wizard (account -> column mapping -> preview)
at /transactions/import, parsed entirely client-side (new lib/csv.ts —
delimiter auto-detect, 5 date formats, thousands/decimal-separator-aware
amounts) and submitted through a new all-or-nothing POST /transactions/bulk
endpoint, mirroring backup restore's validate-everything-before-touching-
anything approach.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-08-25 02:44:30 +03:00
..

Backend tests

Real integration tests against a real Postgres database — a dedicated aurum_test database created fresh, migrated, and dropped again on every run. They never touch the real aurum database or its data: the app's lifespan startup (which seeds against the real DB) is never triggered, and every route's DB dependency is overridden to point at aurum_test instead.

Running

The stack must already be up (docker compose up -d). Test dependencies aren't baked into the production image, so install them once per container lifetime, then run pytest inside the running backend container — it's the only place with network access to Postgres:

docker compose exec backend pip install -r requirements-dev.txt
docker compose exec backend pytest -v

Re-run just pytest -v for subsequent runs; the pip install only needs repeating after a container restart/rebuild.

Adding a test

  • Use the client fixture (an httpx.AsyncClient wired to the app) to hit the API the same way the frontend does — prefer this over reaching into services/models directly, so tests keep verifying the actual contract.
  • account_id and categories fixtures give you the same default account/categories a real fresh install seeds.
  • Every table is truncated and reseeded before each test (see conftest.py::_clean_database), so tests don't need to worry about leftover state or guess at IDs from a previous test.