pytest-mg 0.0.9

A tiny plugin for pytest which runs MongoDB in Docker

pytest-mg

A pytest plugin that provides session-scoped MongoDB fixtures backed by Docker containers.

Features

• Zero configuration — Fixtures work out of the box, no setup required
• Automatic lifecycle management — Pulls images, allocates ports, starts containers, and cleans up after tests
• Fast — Data directory mounted to tmpfs for maximum speed
• Version-specific fixtures — Test against MongoDB 5, 6, 7, 8, or latest
• Replica set fixtures — Drop-in variants with --replSet enabled for change stream support
• Session-scoped — One container per test session, shared across all tests

Installation

pip install pytest-mg

Requirements

• Python 3.10+
• pytest 8.0+
• docker 7.0+
• Docker daemon running locally or accessible via DOCKER_HOST environment variable

Usage

Basic Example

The plugin automatically registers fixtures when installed. Import the Mongo type for type hints:

import pymongo
import pytest_mg


def test_pymongo(mongo: pytest_mg.Mongo):
    client = pymongo.MongoClient(host=mongo.host, port=mongo.port)

    db = client["test_db"]
    collection = db["test_collection"]

    collection.insert_one({"key": "value"})
    doc = collection.find_one({"key": "value"})

    assert doc["key"] == "value"

    client.close()

Available Fixtures

All fixtures are session-scoped and return a Mongo object with host and port attributes:

Standalone (default):

• mongo — Latest MongoDB version (mongo:latest)
• mongo_5 — MongoDB 5.x (mongo:5)
• mongo_6 — MongoDB 6.x (mongo:6)
• mongo_7 — MongoDB 7.x (mongo:7)
• mongo_8 — MongoDB 8.x (mongo:8)

Replica set (required for change streams and transactions):

• mongo_rs — Latest MongoDB version (mongo:latest)
• mongo_5_rs — MongoDB 5.x (mongo:5)
• mongo_6_rs — MongoDB 6.x (mongo:6)
• mongo_7_rs — MongoDB 7.x (mongo:7)
• mongo_8_rs — MongoDB 8.x (mongo:8)

Replica set fixtures require pymongo (pip install pymongo) and start MongoDB with --replSet rs0. When connecting, use directConnection=True to avoid topology-discovery issues from the Docker port mapping:

client = pymongo.MongoClient(
    f"mongodb://{mongo_6_rs.host}:{mongo_6_rs.port}/",
    directConnection=True,
)

Advanced Example: Configuring Environment Variables

Use session-scoped autouse fixtures to configure your application before tests run:

import os
import pytest
import pytest_mg


@pytest.fixture(scope="session", autouse=True)
def init_env(mongo_6: pytest_mg.Mongo) -> None:
    os.environ["MONGODB_CONNECTION_STRING"] = f"{mongo_6.host}:{mongo_6.port}"
    os.environ["MONGODB_DBNAME"] = "myapp"


def test_app():
    # Your application reads from environment variables,
    # no need to reference the fixture directly
    pass

Using with Motor (async)

import motor.motor_asyncio
import pytest
import pytest_mg


@pytest.mark.asyncio
async def test_motor(mongo: pytest_mg.Mongo):
    client = motor.motor_asyncio.AsyncIOMotorClient(host=mongo.host, port=mongo.port)

    db = client["test_db"]
    collection = db["test_collection"]

    await collection.insert_one({"key": "value"})
    doc = await collection.find_one({"key": "value"})

    assert doc["key"] == "value"

    client.close()

How It Works

The plugin uses Docker's Python API to:

1. Pull the specified MongoDB image (if not already cached)
2. Allocate an unused local port
3. Create a container with data directory mounted to tmpfs
4. Start the container and wait for MongoDB to be ready
5. Yield the Mongo object with connection details
6. Kill and remove the container after all tests complete