This package contains test fixtures and resources that bring up an isolated database and SQLAlchemy test fixtures, so that your Python unit tests will run without interfering with each other when you are using the SQLAlchemy ORM.
The database is initialised once at the start of a test process run, as is the session fixture. The session fixture ensures that any commits do not permanently commit, and rolls back the database to a clean state after each test completes.
Python 3.8 and beyond should work.
Install with pip:
pip install db-testtools
Example base test class:
class DBTestCase(testresources.ResourcedTestCase, testtools.TestCase):
"""Base class for all DB tests.
Brings up a temporary database and gives each test its own session.
"""
# These are resources that stay active for the entire
# duration of all the tests being run.
db_fixture = DatabaseResource(
ModelBase,
'myproject.models',
future=True,
)
resources = [('database', db_fixture)]
def setUp(self):
super().setUp()
self.session_fixture = SessionFixture(self.database, future=True)
self.useFixture(self.session_fixture)
# The session itself.
self.session = self.session_fixture.session
# The session factory.
self.Session = self.session_fixture.Session
This base test class will start a SQLite-based database by default and inject self.session as the SQLAlchemy session and self.Session as the SQLAlchemy session factory.
- If you need to use a different database, then you can either:
- pass the engine_fixture_name parameter to DatabaseResource
- set an environment variable TEST_ENGINE_FIXTURE
with the name of the engine fixture to use. Currently two are available:
- SqliteMemoryFixture
- PostgresContainerFixture
Currently the two drivers mentioned above are implemented. The SQLite fixture implements a simple in-memory database which is completely dropped and re-instated on every test.
The PostgresContainerFixture starts its own Postgres instance in a local Docker container. Therefore you must have Docker installed before using this fixture. The Postgres image used by default is 16.3-alpine, but this fixture is known to work all the way back to v11.
If you are already running inside Docker you will need to start the container with --network-"host" so that 127.0.0.1 routes to the started PG containers. You will need to do up to two extra things:
- Bind mount /var/run/docker.sock to the container so docker clients can create sibling containers on the host.
- If you cannot use host networking, supply the IP address of the host's network bridge (usually docker0, etc), so that the fixture knows where to find the PG server. The IP address is either supplied via the constructor to PostgresContainerFixture or you can set the DBTESTTOOLS_PG_IP_ADDR environment variable.
This code has been in use daily on a large project at Cisco for a few years now, and is very stable.
db-testtools is copyright (c) 2021-2024 Cisco Systems, Inc. and its affiliates All rights reserved.