How to Run Only Affected Tests in CI

Most PRs touch a fraction of the codebase but rerun the whole suite. Running only affected tests scales CI to the size of the diff.

Affected-test selection maps a code change to the tests that exercise it and runs only those on PRs. The full suite still runs on the main branch as a safety net.

1. Scope by changed files

Compute the diff against the base branch and pass the changed paths to your test runner.

.github/workflows/ci.yml

- run: |
    CHANGED=$(git diff --name-only origin/main...HEAD)
    npx jest --findRelatedTests $CHANGED --ci

2. Use monorepo affected detection

Nx and Turborepo compute an affected graph from the diff and run tasks only for impacted projects.

.github/workflows/ci.yml

- run: npx nx affected -t test --base=origin/main

3. Map tests to code with coverage

Tools like pytest-testmon and Jest related-tests track which tests cover which lines, so the selection is based on real coverage, not file-name guessing.

4. Always run the full suite on main

Affected selection is for fast PR feedback. Keep an unconditional full run on the default branch so nothing slips through a missed dependency edge.

Key takeaways

Map the diff to the tests that exercise it and run only those on PRs.
Monorepo tools give an affected graph for free.
Keep a full suite run on main so missed edges are still caught.

Related guides

How to Use Test Result Caching in CICache passing test results in CI so unchanged tests are not rerun. Hash test inputs, store the verdict, and r… How to Set Up Nx and Turborepo Remote Caching in CIEnable remote caching for Nx and Turborepo in CI so unchanged tasks are restored instead of rerun. Configure… How to Optimize CI for a MonorepoMake monorepo CI fast: affected-project detection, sparse checkout, per-package caching, and a remote cache s… How to Speed Up Jest in CIMake Jest fast in CI: tune workers, shard the suite, enable the Jest cache, and run only tests affected by th…