Testing

The suite is pytest + pytest-django. Settings: EMAW_BACKEND.settings.test (development minus external services: locmem cache and email, eager Celery, no Redis needed).

One-time: a local PostgreSQL the suite may own

Tests create and drop databases, which the shared dev role cannot do. Run a throwaway container once:

docker run -d --name emaw_test_pg `
  -e POSTGRES_USER=emaw_test -e POSTGRES_PASSWORD=emaw_test -e POSTGRES_DB=emaw_test `
  -p 127.0.0.1:5434:5432 postgres:16-alpine

Start it again later with docker start emaw_test_pg.

Run

The TEST_DB_* variables replace the matching DB_* for the test process only. From Git Bash:

TEST_DB_HOST=127.0.0.1 TEST_DB_PORT=5434 TEST_DB_USER=emaw_test TEST_DB_PASSWORD=emaw_test TEST_DB_NAME=emaw_test \
  .venv/Scripts/python.exe -m pytest

From PowerShell:

$env:TEST_DB_HOST="127.0.0.1"; $env:TEST_DB_PORT="5434"; $env:TEST_DB_USER="emaw_test"; $env:TEST_DB_PASSWORD="emaw_test"; $env:TEST_DB_NAME="emaw_test"
.venv\Scripts\python.exe -m pytest

Useful variants:

python -m pytest crm -k mailable        # one app, one keyword
python -m pytest --create-db            # after schema changes (--reuse-db is on by default)
python -m pytest -x -q                  # stop at first failure

Layout

Item	Where
Config	pytest.ini: testpaths = crm campaigns EMAW events resources store donations payments shipping deposco notes bibles ai, addopts = --reuse-db -q, python_files = test_*.py
Shared fixtures	root conftest.py: plain_user, superuser, crm_staff_user, contact, api_client, auth_client (issues a real SimpleJWT bearer token), and an autouse fixture that patches Stripe card/subscription/invoice lookups
Factories	<app>/tests/factories.py (factory-boy); HTTP mocking with responses
Tests	<app>/tests/test_*.py

Every ticket ships tests for its own code in its app's tests/ folder.

Coverage today

App	Test files
crm	31
store	15
campaigns	12
EMAW	10
events	9
deposco	6
donations	5
resources	4
shipping	3
payments	2
notes, bibles, ai	0 (in testpaths, nothing written yet)

The three empty apps are on the known gaps list.

Writing a test

import pytest
from crm.tests.factories import ContactFactory

@pytest.mark.django_db
def test_contact_list_requires_capability(auth_client, plain_user):
    client = auth_client(plain_user)
    response = client.get("/rest/crm/contacts/")
    assert response.status_code == 403

@pytest.mark.django_db
def test_staff_sees_contacts(auth_client, crm_staff_user):
    ContactFactory(first_name="Ada")
    client = auth_client(crm_staff_user)
    response = client.get("/rest/crm/contacts/")
    assert response.status_code == 200
    assert response.json()["count"] == 1

Rules of thumb:

• Hit endpoints through auth_client, not view functions, so permissions and throttles are exercised.
• Mock Stripe, Deposco, ACS, Graph, OpenAI and Azure Maps at the HTTP boundary (responses) or at the client module; never let a test reach the network.
• Celery is eager in test settings, so tasks run inline; assert on their effects.

Root scratch scripts

verify_redis_down.py, verify_redis_opt.py and verify_seed_dupes.py in the repo root use settings/local_verify.py and a committed verify_db.sqlite3. They are manual checks, not part of the suite.