Phpunit Tia Laravel Package

Product Decisions This Supports

• CI/CD investment sizing: Test Impact Analysis (TIA) changes the economics of the pipeline. Every release in this series makes the "affected test files" story more legible (#25 summarizes affected files on every run; #26 indexes tests by source for testsLinkedTo()), which lets a PM argue for TIA adoption with measurable evidence — wall-clock savings, tests skipped, confidence intervals — rather than vibes.
• Build vs. buy on test selection: The #24 "explicit edges and integration API" plus #27 "expose every known test file on the Edges contract" turn this from a black-box speed hack into a programmable dependency graph. That is the deciding input for build-vs-buy: if you need custom edge definitions (non-PHPUnit suites, generated code, config-driven coupling), buy and extend the API; if your graph is trivial, don't.
• Parallelization roadmap: #28 fixes graph recording under ParaTest. A PM can now sequence "enable TIA" and "shard the suite" as independent roadmap items instead of a forced bundle — sharding no longer blocks or corrupts the impact graph.
• Trust and safety posture: #22 (configured group selection counts as a full run) and #21 (never replay without a recorded baseline) are conservative-correctness defaults. These are the two failure modes that make teams abandon TIA after one escaped defect, so they are releasable risk-reduction work, not polish.
• Rollout gating: Because a missing baseline now forces a full run, the PM can approve a phased rollout (observe-only → gate on PRs → gate on main) without demanding a bespoke fallback harness from engineering.

When to Consider This Package

Adopt when:

• Your PHPUnit suite has grown past the point where developers wait on it, and the pain is feedback latency, not flakiness or isolation.
• You run ParaTest or plan to — v0.5.0 makes graph recording viable in parallel CI, which is where most of the ROI lives.
• You want test selection as a contract other tooling can consume (Edges contract, testsLinkedTo()), e.g. dashboards, PR annotations, or a custom CI reporter.
• You are on a Laravel/PHP codebase with conventional test-to-source conventions — impact analysis is cheap when the mapping is discoverable and expensive when it isn't.
• You need defensible correctness behavior: full-run fallbacks on ambiguous configuration and no baseline-less replays are the guardrails that make this shippable to a risk-averse org.

Look elsewhere when:

• Your bottleneck is test flakiness, ordering, or shared state — TIA reduces which tests run, it does not fix unstable ones, and selective execution can mask order-dependence.
• You depend on dynamic or reflective source→test coupling that doesn't show up in the graph; false negatives are silent and expensive.
• You have no reliable baseline artifact storage across CI runs — v0.5.0 will correctly refuse to replay, so you get full runs and none of the benefit until that's solved.
• You are pre-1.0-averse. This is 0.5.x; the integration API (#24) in particular is young and may shift.

How to Pitch It (Stakeholders)

To executives: This is a CI time-and-cost lever. A PHPUnit test impact analysis tool that only runs tests affected by a change, with conservative fallbacks — if it can't prove a shortcut is safe (no baseline, ambiguous group selection), it runs everything. v0.5.0 is the release that makes it usable in real parallel CI and reportable to leadership: every run now reports the affected test files, which gives us the before/after evidence to justify keeping it.

To engineering: v0.5.0 lands the integration surface we'd want before committing — an explicit edges contract, testsLinkedTo() backed by a source index, a per-run summary of affected files, and a ParaTest fix. The two behavioral changes worth knowing are safety defaults, not restrictions: configured group selection is treated as a full run, and the package refuses to replay without a recorded baseline. Both trade a little speed for eliminating the silent-skip failure mode. Still 0.x, so treat Edges as an interface we build against but may need to adjust.

Ask: Timebox a 2–4 week pilot — observe-only mode on CI, baseline artifact storage sorted, and one dashboard consuming the affected-files summary. Success criteria: measured reduction in executed tests on PRs, zero escaped defects attributable to skipped tests, and a stable baseline across ParaTest shards.