Note on Python TIA (Test Impact Analysis)

python_test_impact_analysis.md

Question:

What could I try for doing Test Impact Analysis for Python projects?

Answer:

There are a few libraries out there you could try. Here are some entrypoints:

• python-tia - this is my preference
• "Potential Alternatives" page in pytest-tia's docs. You can use this as a starting point to comparing alternatives.

Note that TIA is rarely perfect (especially imperfect in dynamic languages), so use with caution.

One pattern to consider is to use a mixed approach. I like to use TIA to exclude some slow tests on branches until they have a merge-request. So master, tags, and merge-requests run full suite. I'm using GitLabCI for this (only: ['merge_requests']) but you can do it with some other tools too.

Lastly here are some reading materials.

• 2015, "Reducing Test Times by Only Running Impacted Tests - Python Edition" by Paul Hammant (archive.org snapshot)
• 2017, "The Rise of Test Impact Analysis" by Paul Hammant (archive.org snapshot)

copied from my answer here to help me remember it