feat(python): add Python/pytest support

[gnufede]

What

Adds Python/pytest support to ddtest, making it possible to use Test Impact Analysis and test skipping for Python projects using pytest.

New platform (python) and framework (pytest) that integrate with the existing ddtest architecture:

• Platform detection: reads runtime tags (runtime.name, runtime.version, os.platform, etc.) from the Python interpreter via an embedded script, and verifies that ddtrace >= 4.10.3 is installed
• Test discovery: resolves pytest config (pytest.ini, pyproject.toml, tox.ini, setup.cfg) to find testpaths and python_files, builds a glob pattern, and passes explicit file paths to python -m pytest (pytest has no --pattern flag like RSpec does)
• Test running: injects --ddtrace into PYTEST_ADDOPTS so the ddtrace pytest plugin loads automatically without requiring changes to the project's pytest config
• Manifest: sets both TEST_OPTIMIZATION_MANIFEST_FILE and DD_TEST_OPTIMIZATION_MANIFEST_FILE in worker env — dd-trace-py reads the latter

Why

Python/pytest is one of the most widely used test stacks. Supporting it in ddtest allows Python teams to benefit from Test Impact Analysis (skipping tests unaffected by a code change), reducing CI time.

E2E testing

Requirements: Python with ddtrace >= 4.10.3 and pytest installed, a Datadog API key with Test Optimization enabled.

Create a small fixture project:

mkdir -p /tmp/pytest-fixture/tests && cd /tmp/pytest-fixture
git init

cat > tests/test_math.py <<'PY'
def test_add(): assert 1 + 1 == 2
def test_subtract(): assert 5 - 3 == 2
PY

cat > pytest.ini <<'INI'
[pytest]
testpaths = tests
INI

git add . && git commit -m "initial fixture"

Run plan then run:

export DD_API_KEY=<your-key>
export DD_CIVISIBILITY_AGENTLESS_ENABLED=true
export DD_ENV=<your-env>
export DD_SERVICE=<your-service>
export DD_TEST_OPTIMIZATION_RUNNER_PLATFORM=python
export DD_TEST_OPTIMIZATION_RUNNER_FRAMEWORK=pytest

ddtest plan   # discovers tests, fetches skippable list from Datadog
ddtest run    # runs pytest in parallel workers with --ddtrace injected

Expected plan output includes Platform: python / pytest, correct runtime tags (CPython, runtime.version), and discovered tests listed per file. After a few runs with coverage data collected, an unrelated commit (e.g. editing a README) should show Skippable tests for this run: N with the full test suites skipped.

[datadog-official]

 

🎉 All green!

🧪 All tests passed
❄️ No new flaky tests detected

🎯 Code Coverage (details)
• Patch Coverage: 79.44%
• Overall Coverage: 87.13% (-0.36%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 7031c71 | Docs | Datadog PR Page | Give us feedback!}