leontrolski / pgtestdbpy

pgtestdbpy

Python clone of pgtestdb.

pip install pgtestdbpy

In summary, it's a couple of helper functions that allow you to quickly clone Postgres databases that you've applied migrations to. In a small number of milliseconds, each test (including tests running in parallel) gets a fresh db with empty tables, all the sequences reset, etc.

In developing this on my mac, for reasons I don't quite understand, running with Postgres in docker (via colima) was substantially quicker that running Postgres natively. So I agree with Peter's advice, just copy this file and docker compose up -d db.

The library consists of two context managers (that can be used in conjunction or independently depending on test setup):

• pgtestdbpy.template(config, migrator):
  • Creates a new user and database for the migrator.
  • Runs migrations.
  • Marks the database as a TEMPLATE DATABASE so that it can be cheaply cloned.
  • Yields.
  • Drops the template database and the user.
• pgtestdbpy.clone(config, migrator):
  • Does a CREATE DATABASE WITH TEMPLATE (from a template database made above) giving it a unique random name.
  • Yields a Postgres url for it.
  • Drops the database.

Example conftest.py usage below, in theory (I haven't tested this) it should be easy to run tests in parallel using the conn fixture - each with a separate database instance - and pytest-xdist or equivalent.

from typing import Iterator
import pgtestdbpy
import psycopg
import pytest

def migrate(url: str) -> None:
    with psycopg.connect(url) as conn:
        conn.execute("CREATE TABLE foo (a INT)")

migrator = pgtestdbpy.Migrator(migrate)
config = pgtestdbpy.Config()

@pytest.fixture(scope="session")
def db() -> Iterator[None]:
    with pgtestdbpy.template(config, migrator):
        yield

@pytest.fixture()
def conn(db) -> Iterator[pgtestdbpy.PsycoConn]:
    with pgtestdbpy.clone(config, migrator) as url:
        with psycopg.connect(url) as _conn:
            yield _conn