Affected-Test Selection

The planner maps a Git change set to a safe unit and Playwright test plan.

Selection Pipeline

1. Calculate the protected merge base.
2. Collect committed, staged, unstaged, renamed, deleted, and configured untracked changes.
3. Find owning Rush projects and reverse dependents.
4. Ask CodeGraph for statically affected tests.
5. Apply explicit mappings and scenario coverage edges.
6. Add configured core and smoke safety tests.
7. Evaluate fail-open conditions and limits.
8. Write a canonical, fingerprinted plan.

Selection Reasons

Each test records why it was included:

• changed-test
• codegraph-reachable
• covers
• mapping
• core-safety
• smoke-safety
• Project ownership and dependent-project expansion

Use explain for one changed file:

npx playwright-codegraph explain packages/cart/src/quantity.ts

Pull Request Execution

git fetch origin main
npx playwright-codegraph graph sync
npx playwright-codegraph scenarios
npx playwright-codegraph plan --base origin/main
npx playwright-codegraph run

Authoritative Full Execution

Protected branches, scheduled runs, and releases can create an explicit full plan:

npx playwright-codegraph plan --base HEAD --full
npx playwright-codegraph run

The plan records manual-full-run, inventories every test, and uses each runner’s configured fullArgs.

Shadow Mode

Shadow mode compares selective execution with an authoritative full run:

npx playwright-codegraph shadow --base origin/main

Use shadow records before making selective execution a required pull request gate. Replay retained records to calculate fallback and missed-failure counts:

npx playwright-codegraph replay --records test-selection-history

Safety Guarantee

CodeGraph guarantees deterministic selection for identical inputs and explicit full-suite fallback when evidence is incomplete. It does not claim that applications, browsers, networks, or tests are inherently deterministic.