Files
accounted/app/api/v1/companies/route.ts
T
Jakob WennbergandClaude Opus 4.8 fc2b4d1e23 fix(api): declare the real { data, meta } envelope for v1 list endpoints (#802)
The OpenAPI success schemas for v1 list endpoints declared a bare
{ <name>: [...] } object that no handler returns, so the published spec
advertised a shape the API never emits (#781, item 2). response.success is
doc-only (feeds zodToJsonSchema for /openapi.json; not validated at runtime),
so this is a documentation fix with no behaviour change.

Add listEnvelope() ({ data: [...], meta }) and dataEnvelope() ({ data, meta })
plus a shared ResponseMetaSchema. Ten endpoints that return paginated() now use
listEnvelope; the three that deliberately wrap their array under a named key via
ok() (accounts, fiscal-periods, webhooks — a shape their route tests lock in)
use dataEnvelope. Also corrects the accounts/fiscal-periods examples, which
showed an unwrapped data: [...] that contradicted their handlers.

Refs #781.

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-26 15:30:48 +02:00

197 lines
6.4 KiB
TypeScript

/**
* GET /api/v1/companies — list companies the calling API key can access.
*
* The API key is bound to a user; that user may be a member of multiple
* companies (consultant-style). This endpoint returns every company the user
* has a non-archived membership for, in stable created_at order.
*
* Used by 3rd-party integrations to discover which company IDs to scope
* subsequent calls to.
*/
import { z } from 'zod'
import { paginated } from '@/lib/api/v1/response'
import {
encodeDefaultCursor,
parsePaginationParams,
decodeDefaultCursor,
} from '@/lib/api/v1/pagination'
import { registerEndpoint, listEnvelope } from '@/lib/api/v1/registry'
import { withApiV1 } from '@/lib/api/v1/with-api-v1'
import { v1ErrorResponse } from '@/lib/api/v1/errors'
const Company = z.object({
id: z.string().uuid(),
name: z.string(),
org_number: z.string().nullable(),
entity_type: z.string(),
role: z.enum(['owner', 'admin', 'member', 'viewer']),
created_at: z.string(),
})
const CompaniesListResponse = listEnvelope(Company)
registerEndpoint({
operation: 'companies.list',
method: 'GET',
path: '/api/v1/companies',
summary: 'List companies the API key can access.',
description:
'Returns every non-archived company the API key user is a member of, together with their role. ' +
'Use the returned `id` as `{companyId}` in subsequent endpoints.',
useWhen:
'You need to discover which company IDs an API key has access to before calling company-scoped endpoints.',
doNotUseFor:
'Fetching a single company you already know the id of — use GET /api/v1/companies/{companyId} for that.',
pitfalls: [
'Multi-company keys (e.g. consultants) will see >1 result. Always pass the correct companyId in subsequent paths.',
'Archived companies are excluded; if a company disappears the user has been removed from it or it was archived.',
],
example: {
response: {
data: [
{
id: '8fd5b1f4-…',
name: 'Acme AB',
org_number: '556677-8899',
entity_type: 'aktiebolag',
role: 'owner',
created_at: '2025-01-04T08:00:00Z',
},
],
meta: { request_id: 'req_…', api_version: '2026-05-12', next_cursor: null },
},
},
scope: 'companies:read',
risk: 'low',
idempotent: true,
reversible: false,
dryRunSupported: false,
response: { success: CompaniesListResponse },
})
export const GET = withApiV1('companies.list', async (request, ctx) => {
const url = new URL(request.url)
const { limit, cursor } = parsePaginationParams(url)
const decoded = decodeDefaultCursor(cursor)
// Authorization boundary: the query below filters by `user_id = ctx.userId`
// BEFORE the cursor's keyset is applied, so a tampered cursor can only
// reorder rows the caller is already entitled to see. Cursors are not
// signed; that's intentional. See PR #450 review for the trade-off.
//
// Keyset pagination uses (joined_at ASC, id ASC) on `company_members`. Two
// memberships sharing a `joined_at` (bulk-imported, concurrent registrations)
// are disambiguated by the `company_members.id` tiebreaker so rows on a page
// boundary are never skipped or duplicated.
// We over-fetch by one to determine whether a next page exists.
let query = ctx.supabase
.from('company_members')
.select(
`
id,
role,
joined_at,
companies:company_id (
id,
name,
org_number,
entity_type,
archived_at,
created_at
)
`,
)
.eq('user_id', ctx.userId)
.is('companies.archived_at', null)
.order('joined_at', { ascending: true })
.order('id', { ascending: true })
.limit(limit + 1)
if (decoded) {
// Compound keyset: joined_at > cursor.ts OR (joined_at = cursor.ts AND id > cursor.id).
// `.or()` takes a comma-separated PostgREST filter string.
query = query.or(
`joined_at.gt.${decoded.ts},and(joined_at.eq.${decoded.ts},id.gt.${decoded.id})`,
)
}
const { data, error } = await query
if (error) {
return v1ErrorResponse(error, ctx.log, { requestId: ctx.requestId })
}
type CompanyRow = {
id: string
name: string
org_number: string | null
entity_type: string
archived_at: string | null
created_at: string
}
type Row = {
// `company_members.id` — the membership row's own UUID, used as the
// cursor's secondary key. Not exposed in the response.
id: string
role: 'owner' | 'admin' | 'member' | 'viewer'
joined_at: string
// PostgREST returns the joined company as either an object (one-to-one FK
// resolution) or an array. Accept both — Supabase's auto-typing chooses
// the array shape, but actual responses for a single-row FK are objects.
companies: CompanyRow | CompanyRow[] | null
}
const rows = ((data ?? []) as unknown) as Row[]
const trimmed = rows.slice(0, limit)
const hasMore = rows.length > limit
const pickCompany = (r: Row): CompanyRow | null => {
if (!r.companies) return null
return Array.isArray(r.companies) ? (r.companies[0] ?? null) : r.companies
}
// Defense in depth: PostgREST's `.is('companies.archived_at', null)` is
// expected to filter out archived companies before the row reaches us.
// If a `company_members` row arrives without an associated company object,
// the join filter behaved differently than expected — drop the row AND
// surface it as a warn so we notice silent data-integrity regressions.
let droppedNulls = 0
const companies = trimmed
.map((r) => {
const c = pickCompany(r)
if (!c) {
droppedNulls += 1
return null
}
return {
id: c.id,
name: c.name,
org_number: c.org_number,
entity_type: c.entity_type,
role: r.role,
created_at: c.created_at,
}
})
.filter((x): x is NonNullable<typeof x> => x !== null)
if (droppedNulls > 0) {
ctx.log.warn('companies.list: dropped rows with null company join', { droppedNulls })
}
// Cursor encodes the LAST row's `(joined_at, company_members.id)` — always
// present, no null-guard needed. Independent of whether the joined company
// dropped out of the response shape.
const last = trimmed[trimmed.length - 1]
const nextCursor = hasMore && last
? encodeDefaultCursor({ id: last.id, created_at: last.joined_at })
: null
return paginated(companies, {
requestId: ctx.requestId,
nextCursor: nextCursor ?? undefined,
})
})