The mozanalysis
Python library is a library to standardize experiment analysis
at Mozilla for the purpose of producing decision reports templates that are
edited by data scientists.
Online documentation is available at https://mozilla.github.io/mozanalysis/
- To install this package from pypi run:
pip install mozanalysis
Linting and Formatting are done with Ruff.
First one must ensure that all the test dependencies are installed. To do so, navigate to the root of the repo and run
pip install -e ".[testing]"
Then run pytest
on the commandline
Install tox into your global Python environment and run tox
.
You can pass flags to tox to limit the different environments you test in
or the tests you run. Options after --
or positional arguments are forwarded to pytest.
For example, you can run:
-
tox -e lint
to lint -
tox -e py310 -- -k utils
to only run tests with "utils" somewhere in the name, on Python 3.10 -
tox tests/test_utils.py
to run tests in a specific file
To test/debug this package locally, you can run exactly the job that CircleCI runs for continuous integration by installing the CircleCI local CLI and invoking:
circleci build --job py310
See .circleci/config.yml for the other configured job names (for running tests on different python versions).
Releasing mozanalysis happens by tagging a CalVer based Git tag with the following pattern:
YYYY.M.MINOR
where YYYY is the four-digit year number, M is a single-digit month number and MINOR is a single-digit zero-based counter which does NOT relate to the day of the release. Valid versions numbers are:
2017.10.0
2018.1.0
2018.12.12
Once the (signed) Git tag has been pushed to the main GitHub repository using git push origin --tags, Circle CI will automatically build and push a release to PyPI after the tests have passed.