lu-python 0.0.5

Lightweight recorder/fixture helper (formerly mcr)

lu — a vcr-style record/replay stubbing library for Python methods

lu (录) — Chinese for "recording".

lu provides a small context manager, record(), which can patch module-qualified functions or instance methods to record return values (or exceptions) to compressed pickle files and generate a manifest.json describing the recordings.

On the first run, the patched methods execute normally and their outputs are recorded. Subsequent runs replay the recorded return values (or re-raise recorded exceptions) instead of executing the original code — similar to how HTTP cassette libraries like vcrpy work, but for arbitrary Python methods.

Quickstart

Install from PyPI:

pip install lu-python

Basic usage:

import lu

with lu.record(
    target={
        'module1.Foo.expensive_method1': ['arg1', 'kwarg1'],
        'package2.module2.expensive_method2': ['arg2'],
    },
    recordings_dir='tests/fixtures/recordings/',
):
    from module1 import Foo
    foo = Foo()
    foo.expensive_method('bar', kwarg1=3)

    from package2 import module2
    module2.expensive_method2('baz')

See tests/test_lu.py for more examples.

Using with pytest

Add an autouse fixture (for example in tests/conftest.py) to enable recording/replay for the test session:

import pytest
import lu

@pytest.fixture(scope='session', autouse=True)
def stub_calls():
    with lu.record(
        target={
            # 'module.Class.method': ['arg_name', 'kwarg_name'],
            # add targets you want recorded here
        },
        recordings_dir='tests/fixtures/recordings/',
    ):
        yield

How lu differs from other tools

I wrote lu because I couldn't find a simple record/replay tool for arbitrary Python methods. A few alternatives exist but they don't match this use case:

• testcontainers: runs real services in Docker, which requires setup and infrastructure and does not provide automatic request/response recording.
• keploy: captures and replays network calls at a lower level; powerful but heavier-weight.
• vcrpy / pytest-recordings: focused on HTTP clients (requests/urllib3) only.
• unittest.mock: flexible but manual — lu provides an automated record-and-replay workflow on top of method patching.

lu works at the method level, so you can record calls to client libraries (e.g., database drivers, HTTP clients) or your own functions.

Limitations

• Prefer using lu for read-only calls. While you can record methods with side effects (for example, database writes), lu does not maintain external state. Replaying a recorded write will not update a database or other external system.
• No transaction or rollback support. lu records and replays return values and exceptions only.

Contributing

Contributions and bug reports are welcome. Please open an issue or submit a pull request with tests that demonstrate the change or fix.

License

This project is provided under the terms of the repository license. See the LICENSE file for details.