ci: Only run-tests if tests/demo/examples/core is modified

[angela-ko]

Relevant issue or PR

#205

Description of changes

Add filter to only run-tests if tests demo examples tesseract-core is modified

Testing done

This will be the test (if run-tests doesn't trigger)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 76.11%. Comparing base (0b0aedc) to head (18af8a6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #210   +/-   ##
=======================================
  Coverage   76.11%   76.11%           
=======================================
  Files          28       28           
  Lines        3082     3082           
  Branches      480      480           
=======================================
  Hits         2346     2346           
  Misses        523      523           
  Partials      213      213           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jpbrodrick89]

Love this Angela! Will use something similar in one of my repos! What about .githuh/workflows/run_tests.yaml too?

[jpbrodrick89]

Also, should probably consider adding production.uv.lock (and maybe requirements.txt or pyproject.toml) so that they run when dependencies are updated?

[dionhaefner]

Hmm isn't this pretty much all paths now? 😅

I suggest a slightly different approach:

• Always run base tests and doc build.
• Always run all tests on pushes to main.
• Run pip install actions only if requirements.txt has changed.
• Run E2E tests and examples only if the following paths have changed:

      - "tesseract_core/**"
      - "tests/endtoend_tests/**"
      - "demo/**"
      - "examples/**"
      - "production.uv.lock"

(accepting the rare false-negative but that should be OK and immediately caught by the push to main test)

This makes it so we don't need to run the expensive E2E tests if changes are isolated to docs, .github, pyproject.toml, base tests, or README.md / other top-level files.

[angela-ko]

I guess the concern I thought Jonathan might have (hence why I added) is that if the packages change, the tests may fail due to changes in the packages? But I supposed that would be relatively rare and could be caught in the regular benchmarking?

I think currently, docs changes + workflow files other than e2e test file changes are excluded, but nothing else. Added base tests to exclusion list is a good catch as well.

I don't quite understand the production.uv.lock actually -- i thought that's just generated based on the pyproject.toml / requirements? So would adding that be effectively the same as having pyproject.toml in the list?

[dionhaefner]

I guess the concern I thought Jonathan might have (hence why I added) is that if the packages change, the tests may fail due to changes in the packages? But I supposed that would be relatively rare and could be caught in the regular benchmarking?

Yes, that's what I meant with my comment regarding false-negatives. It's not inconceivable that only an E2E test breaks due to a package update, but unlikely (and it would be caught immediately after pushing to main).

I don't quite understand the production.uv.lock actually -- i thought that's just generated based on the pyproject.toml / requirements? So would adding that be effectively the same as having pyproject.toml in the list?

The lockfile contains specific versions, not abstract dependencies. pyproject.toml only changes if we add / remove deps, while the lock changes whenever there's an upstream release.

[angela-ko]

@dionhaefner I did leave run_tests.yml in there since the tests should be run if the test configs are modified?

[angela-ko]

Need to make some changes for end2end vs base*

[angela-ko]

This works for the whole workflow run_tests, but if we want to prescribe path for specific jobs (i.e.. end-to-end vs base) is that possible to achieve in a similar way?

I saw that there's some hacky way about it using if conditions in the jobs and matching paths? but I think it's best to just split it into 2 workflow files for e2e vs base...

[angela-ko]

Current version only runs tests-base but not e2e.. weird bc I added the .yml to the paths file so it should run, but good to see the base only part is working..

[angela-ko]

https://github.com/pasteurlabs/tesseract-core/actions/runs/15568981560/job/43840017745?pr=219

Did a test branch here and checked that e2e tests are not fully running when only doc is changed vs in current branch it is fullyg running e2e tests

[dionhaefner]

Pre-approved, this looks really clean now! Let's do the final cosmetics then merge.

Thanks @angela-ko

[angela-ko]

Need to do some changes to make it skip entire jobs (if possible) instead of steps -- but will do after dinner..

[angela-ko]

https://github.com/pasteurlabs/tesseract-core/actions/runs/15573704318/job/43854596186?pr=219

Ok reran the tests and now it's fully showing up as skipped (did not know if conditions could be put on the full job)

vs tests are running here.

Also looks a LOT cleaner now!