Workflows

Failed Test Workflow

When tests fail, failed test locations are automatically saved. Re-run only the failures:

check --only test     # Run tests, failures saved automatically
check --failed        # Re-run only the failed tests
check --failed --repeat 10  # Re-run with repeat

Failed tests are saved to .check/failed_tests.txt. Failures are extracted from both test partition outputs and any custom check output (e.g. modified_tests).

Auto-fix Workflow

Run checks first to store output, then apply fixes:

check --only format,credo  # Run checks and store output
check --fix                # Apply fixes from stored output

Fix commands are configurable in .check.json:

"fix": [
  {"run": "mix format"},
  {"run": "mix recode", "files": ".check/credo*.txt"}
]

Modified Tests Workflow

Run only tests that changed on your branch vs the base branch:

check --only modified_tests           # Run only changed test lines
check --only modified_test_modules    # Run whole changed test files
check --only modified_tests --repeat 5  # Repeat modified tests

modified_tests uses granular detection:

• Module-level setup changed → runs the whole file
• Setup inside describe changed → runs that describe block
• Describe line changed → runs that describe block
• Test body changed → runs only that specific test line

modified_test_modules is simpler - runs the entire file for any changed test module.

Coverage Workflow

Show coverage report (cached if cover/ data hasn't changed):

check --coverage       # Show coverage report (cached if cover/ unchanged)
check --no-coverage    # Run checks without coverage (overrides .check.json)

Coverage is also merged automatically after partitioned test runs. Configure in .check.json:

"coverage": {
  "mod": "native",
  "limit": 80,
  "html": false,
  "baseline_cmd": "git show origin/main:coverage.txt"
}

When baseline_cmd is set, shows delta vs baseline:

✓ Coverage: 85.5% | Report: cover/
  Coverage: +2.3% vs baseline (83.2%)

Watch Mode

Monitor test partition output files in real-time while tests run in another terminal:

check --watch