https://github.com/kevlened/pytest-parallel

A pytest plugin for parallel and concurrent testing
https://github.com/kevlened/pytest-parallel

concurrent parallel pytest

Last synced: 2 days ago
JSON representation

A pytest plugin for parallel and concurrent testing

• Host: GitHub
• URL: https://github.com/kevlened/pytest-parallel
• Owner: kevlened
• License: mit
• Archived: true
• Created: 2018-05-05T02:29:51.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2024-05-29T15:45:11.000Z (5 months ago)
• Last Synced: 2024-10-29T01:12:14.285Z (11 days ago)
• Topics: concurrent, parallel, pytest
• Language: Python
• Homepage: https://github.com/browsertron/pytest-parallel/issues/104#issuecomment-1293941066
• Size: 153 KB
• Stars: 313
• Watchers: 10
• Forks: 60
• Open Issues: 64
• Metadata Files:
  • Readme: README.md
  • Contributing: CONTRIBUTING.md
  • License: LICENSE

Awesome Lists containing this project

• my-awesome-starred - kevlened/pytest-parallel - A pytest plugin for parallel and concurrent testing (Python)

README

        
## Maintainers needed

[The project is currently unmaintained](https://github.com/browsertron/pytest-parallel/issues/104#issuecomment-1293941066)

# pytest-parallel

a pytest plugin for parallel and concurrent testing

## What?

This plugin makes it possible to run tests quickly using multiprocessing (parallelism) and multithreading (concurrency).

## Why?

`pytest-xdist` is great to run tests that:

  1. aren't threadsafe

  2. perform poorly when multithreaded

  3. need state isolation

`pytest-parallel` is better for some use cases (like Selenium tests) that:

  1. can be threadsafe

  2. can use non-blocking IO for http requests to make it performant

  3. manage little or no state in the Python environment

Put simply, `pytest-xdist` does parallelism while `pytest-parallel` does parallelism and concurrency.

## Requirements

* Python3 version [3.6+]

* Unix or Mac for `--workers`

* Unix, Mac, or Windows for `--tests-per-worker`

## Installation

`pip install pytest-parallel`

## Options

* `workers` (optional) - max workers (aka processes) to start. Can be a **positive integer or `auto`** which uses one worker per core. **Defaults to 1**.

* `tests-per-worker` (optional) - max concurrent tests per worker. Can be a **positive integer or `auto`** which evenly divides tests among the workers up to 50 concurrent tests. **Defaults to 1**.

## Examples

```bash

# runs 2 workers with 1 test per worker at a time

pytest --workers 2

# runs 4 workers (assuming a quad-core machine) with 1 test per worker

pytest --workers auto

# runs 1 worker with 4 tests at a time

pytest --tests-per-worker 4

# runs 1 worker with up to 50 tests at a time

pytest --tests-per-worker auto

# runs 2 workers with up to 50 tests per worker

pytest --workers 2 --tests-per-worker auto

```

## Notice

Beginning with Python 3.8, forking behavior is forced on macOS at the expense of safety.

    Changed in version 3.8: On macOS, the spawn start method is now the default. The fork start method should be considered unsafe as it can lead to crashes of the subprocess. See bpo-33725.

[Source](https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods)

## License

MIT