mirror of
https://github.com/yusufipk/OpenFrame.git
synced 2026-09-11 17:46:06 +00:00
The repo had no automated tests. Every change was verified by hand. Adds four layers, 2023 tests in total, runnable with one command: - 1191 unit tests over the pure logic in lib/, including the full computeProjectAccess permission matrix and the billing gate - 167 component and hook tests in jsdom, covering the hooks that hold real logic rather than presentational wrappers - 647 API integration tests against a real Postgres, with only auth() mocked, including a data-driven sweep asserting that none of the 60 route modules answers 2xx to an unauthenticated caller - 18 Playwright specs driving a real browser against a real build Infrastructure: vitest.config.ts with three projects, a disposable Postgres and MinIO in docker-compose.test.yml, factories and helpers under tests/, scripts/test.sh as the single entry point, a pre-push hook running bun run verify, and CI split into check, test and e2e jobs. The test database is built with prisma db push plus a replay of the hand-written SQL, because prisma migrate deploy cannot build this schema from empty: the migration history has no captured baseline. This mirrors what scripts/docker-db-bootstrap.ts already does in production, and tests/setup/db-global.ts carries a drift guard so a new migration fails the run until someone reviews it. Production code is unchanged apart from one pure-function extraction out of use-video-player.ts, which was too large to test in jsdom. Several tests pin behaviour that looks wrong, each marked KNOWN BUG in place. TESTING.md section 12 records where the plan turned out to be wrong, and AGENTS.md now states which layer a change needs a test in.
110 lines
3.9 KiB
TypeScript
110 lines
3.9 KiB
TypeScript
// Test-database lifecycle helpers.
|
|
//
|
|
// Importing this module imports `@/lib/db`, which reads DATABASE_URL at module
|
|
// load. Anything that imports this must have loaded `tests/helpers/env.ts`
|
|
// first; `tests/setup/api.ts` does exactly that.
|
|
|
|
import { db } from '@/lib/db';
|
|
|
|
// Prisma owns this table and emptying it would make `prisma db push` believe the
|
|
// database has never been set up.
|
|
const PRESERVED_TABLES = new Set(['_prisma_migrations']);
|
|
|
|
let cachedTableNames: string[] | null = null;
|
|
let cachedResetStatement: string | null = null;
|
|
|
|
/**
|
|
* Every base table in the `public` schema, read from `information_schema` so
|
|
* the list can never drift out of sync with `prisma/schema.prisma`. A model
|
|
* added tomorrow is emptied tomorrow, with no edit here.
|
|
*/
|
|
export async function listResettableTables(): Promise<string[]> {
|
|
if (cachedTableNames) return cachedTableNames;
|
|
|
|
const rows = await db.$queryRaw<Array<{ table_name: string }>>`
|
|
SELECT table_name
|
|
FROM information_schema.tables
|
|
WHERE table_schema = 'public'
|
|
AND table_type = 'BASE TABLE'
|
|
ORDER BY table_name
|
|
`;
|
|
|
|
const names = rows.map((row) => row.table_name).filter((name) => !PRESERVED_TABLES.has(name));
|
|
|
|
if (names.length === 0) {
|
|
throw new Error(
|
|
'resetDb() found no tables in the public schema. The test database was ' +
|
|
'probably never migrated. Check that tests/setup/db-global.ts ran ' +
|
|
'`prisma db push` against DATABASE_URL.'
|
|
);
|
|
}
|
|
|
|
cachedTableNames = names;
|
|
return names;
|
|
}
|
|
|
|
/**
|
|
* Builds the single statement that empties the database.
|
|
*
|
|
* Two decisions worth explaining, because the obvious implementation is both
|
|
* slower and wrong:
|
|
*
|
|
* - DELETE, not TRUNCATE. `TRUNCATE` rewrites the relation file for every
|
|
* table and every index, which measured at ~36ms per call against this
|
|
* container even with fsync off and the data directory on tmpfs. Across a
|
|
* suite that resets after every test that is most of the runtime. The
|
|
* equivalent DELETE costs ~3ms.
|
|
*
|
|
* - One statement, via data-modifying CTEs, rather than one DELETE per table.
|
|
* Prisma's foreign keys are NOT DEFERRABLE, so a sequence of separate
|
|
* DELETEs has to run in child-before-parent order or it trips a constraint.
|
|
* Inside a single statement the FK triggers all fire after the whole
|
|
* statement has run, by which point every row is already gone, so no
|
|
* ordering is needed and a newly added table cannot break the order.
|
|
*
|
|
* The trailing `setval` calls stand in for TRUNCATE's `RESTART IDENTITY`, so a
|
|
* test can still rely on `rate_limits.id` starting from 1.
|
|
*/
|
|
async function buildResetStatement(): Promise<string> {
|
|
if (cachedResetStatement) return cachedResetStatement;
|
|
|
|
const tables = await listResettableTables();
|
|
const sequences = await db.$queryRaw<Array<{ sequence_name: string }>>`
|
|
SELECT sequence_name
|
|
FROM information_schema.sequences
|
|
WHERE sequence_schema = 'public'
|
|
ORDER BY sequence_name
|
|
`;
|
|
|
|
const deletes = tables
|
|
.map((table, index) => `"d${index}" AS (DELETE FROM "public"."${table}")`)
|
|
.join(', ');
|
|
|
|
const projection =
|
|
sequences.length > 0
|
|
? sequences.map((row) => `setval('"public"."${row.sequence_name}"', 1, false)`).join(', ')
|
|
: '1';
|
|
|
|
cachedResetStatement = `WITH ${deletes} SELECT ${projection}`;
|
|
return cachedResetStatement;
|
|
}
|
|
|
|
/**
|
|
* Empties the test database. Registered as `afterEach` in tests/setup/api.ts,
|
|
* so every test starts from zero rows and no test may depend on another test's
|
|
* data or on file execution order.
|
|
*/
|
|
export async function resetDb(): Promise<void> {
|
|
await db.$executeRawUnsafe(await buildResetStatement());
|
|
}
|
|
|
|
/** Row count for a table, for assertions like "nothing was written". */
|
|
export async function countRows(table: string): Promise<number> {
|
|
const rows = await db.$queryRawUnsafe<Array<{ count: bigint }>>(
|
|
`SELECT COUNT(*)::bigint AS count FROM "public"."${table}"`
|
|
);
|
|
return Number(rows[0]?.count ?? 0);
|
|
}
|
|
|
|
export { db };
|