Skip to content

testing.md

Latest commit

 

History

History
98 lines (78 loc) · 4.66 KB

testing.md

File metadata and controls

98 lines (78 loc) · 4.66 KB
Raw

Testing

Three layers, each with its own isolated setup:

bun run test              # unit tests (vitest) — pure domain logic, no I/O
bun run test:integration  # integration tests against a real, isolated Postgres DB
bun run test:e2e          # Playwright browser tests covering the user journeys

Run bun run typecheck for a strict TypeScript pass (the project has no any-shaped escape hatches in application code) and bun run lint for ESLint.

Dead-code checks

bun run knip reports unused files, dependencies, and exports. It exits non-zero when it finds anything, which is how it would gate CI — today it is a local check, so run it after adding or deleting modules, dependencies, or exports.

Two things in knip.json are worth knowing before trusting its output:

  • Both workspaces are declared (. and website). Without that, Knip analyses only the root project, never sees Vite's index.html as an entry point, and reports every website/src/** file as unused.
  • sharp is in ignoreDependencies. No module imports it; Next loads it at runtime for image optimization on a self-hosted deploy. Removing it because Knip called it unused would silently degrade image handling in production.

The e2e and integration suites are declared as entry points too, so their helpers don't read as dead code.

Unit tests

lib/**/__tests__ (config: vitest.config.mts). Pure logic only — no database, no network. Covered modules:

  • lib/domain/: authorization predicates (including the observer role), position/reorder math, filter logic, invitation expiry rules, WIP-limit checks, due-date bucketing, mention parsing, notification fan-out rules, search-query normalization, attachment validation/filename sanitizing, SSE event shaping, activity formatting, analytics tallying, calendar month-grid math, subtask rollup, dependency cycle detection, Gantt bar/range layout, and health-status aggregation.
  • lib/__tests__/markdown.test.ts: markdown-lite XSS safety.
  • lib/auth/__tests__/password.test.ts: scrypt password hashing.
  • lib/templates/__tests__/boards.test.ts: board-template definitions.
  • lib/import/__tests__/csv.test.ts: CSV import parsing.

Integration tests

test/integration/db.test.ts (config: vitest.integration.config.mts) runs against stackboard_test, a separate database from your dev DB. It drops and recreates the schema before each run and wipes tables between tests, so it is safe to run repeatedly but must never point at a database with real data — the suite refuses to start unless DATABASE_URL contains stackboard_test. Create it once with:

docker exec -it <postgres-container> psql -U stackboard -d stackboard -c "CREATE DATABASE stackboard_test;"

End-to-end tests

test/e2e/*.spec.ts (Playwright). By default the suite builds the app and boots it on port 3100 (playwright.config.ts) against your real dev database, reseeded via global-setup.ts — set E2E_BASE_URL to point it at an already-running instance instead. It then drives a real browser through:

  1. owner creates a board, invites a teammate, adds a card with assignee/due date/checklist (journey-create-board)
  2. teammate filters to their cards, completes a checklist item, comments, and drags a card to Done — real pointer events, not a simulated event (journey-teammate-workflow)
  3. owner archives a card and an empty column, reviews the activity trail, restores the card into a chosen column (journey-archive-restore)
  4. a card created by one member appears on another member's open board with no interaction — real-time SSE sync (journey-realtime)
  5. owner drag-reorders columns in Settings (journey-column-reorder)
  6. a board round-trips through CSV export and import (journey-csv-import-export)
  7. a non-member is denied access to the board via a direct URL, and the board is usable at a 390×844 mobile viewport (permission-and-mobile)
  8. the health endpoint reports a live database and job queue, and leaks no connection details (health-endpoint)

Continuous integration

.github/workflows/ci.yml runs on every push to main and every pull request, in three parallel jobs:

Job What it runs
unit lint, typecheck, test, build
integration test:integration against a postgres:16 service container (stackboard_test DB)
e2e db:migrate then test:e2e (Chromium); uploads the Playwright report on failure