Testing Concepts & Exercises

A test double is any stand-in for a real component. Dummy: passed but never used. Stub: returns fixed canned values. Fake: a working but simplified implementation (e.g. an in-memory DB). Spy: records how it was called without changing behavior. Mock: a stub plus assertions on how it was called (e.g. sendEmail was called once).

Mock to isolate the unit under test from slow or unstable external dependencies (DB, third-party APIs, email), to simulate hard-to-reproduce errors like a 500, and to keep tests fast and deterministic. Don't mock when you specifically want to verify real integration, when a mock would diverge so far from reality it loses value, or when the real dependency is already local and fast.

Use page.route to intercept a URL pattern and fulfill it with a canned response: page.route('**/api/products', r => r.fulfill({ status: 200, body: JSON.stringify(data) })). You can return a 500 to test error UI, or add a delay before route.continue() to test loading states — all without touching the real backend.

Cover the full lifecycle and the auth/validation edges: create (POST -> 201), read it back (GET -> 200), update partially (PATCH -> 200, unchanged fields stay), delete (DELETE -> 204), then confirm it's gone (GET -> 404). Add negative cases: no token -> 401, wrong role -> 403, missing required field -> 422, duplicate -> 409. Each test should set up its own data so it stays independent.

TTL (time to live) is how long something stays valid: auth tokens/sessions (JWT expiry), cache entries (Redis, CDN, HTTP Cache-Control), and rate-limit windows (X-RateLimit-Reset). As an SDET you test that an expired token is rejected, a cache entry is a miss after its TTL, and that requests beyond the limit are throttled and reset correctly.

A fixed sleep (waitForTimeout(3000)) is either too short (flaky) or too long (slow), because real timing varies. Use condition-based waits instead: waitForSelector for an element, expect(...).toBeVisible(), waitForResponse for an API call, or a generic polling helper that re-checks a condition until a timeout. This makes tests both faster and more reliable.

Retry re-runs a failing operation up to N attempts, waiting longer between each try (delay * 2^attempt) so a struggling service can recover without being hammered. Only retry transient failures (network blips, 5xx, timeouts). Don't retry deterministic client errors like a 404 or 422 — a shouldRetry predicate lets you skip those, since retrying will never succeed.