state-hub/workplans/STATE-WP-0087-bounded-repository-collection-projection.md

id	type	title	domain	repo	status	owner	topic_slug	created	updated	origin	origin_ref	quality_dor	quality_dor_at	quality_dor_by	quality_dor_note	quality_dod	quality_dod_at	quality_dod_by	quality_dod_note	reviewed_at	reviewed_against_commit
STATE-WP-0087	workplan	Bound the repository collection projection	infotech	state-hub	finished	codex	infotech	2026-09-04	2026-09-05	intake:01a06e0c-3feb-7920-a6f2-961b03dc170d	STATE-WP-0081	DoR-Ok	2026-09-04	codex	The live 131-repository collection exceeds 15 seconds and the implementation reveals per-repository identity queries. Scope, owner, dependency-free patch, and query-bound regression signal are explicit.	DoD-Ok	2026-09-05	codex	Bulk identity projection and domain-loading bounds are covered by 102 passing repository/router regressions. Production-shaped PostgreSQL timing and SQL counts are recorded in docs/evidence/repository-collection-projection-20260905.md.	2026-09-05	4cef58c

Bound the repository collection projection

Goal

Make GET /repos/ complete with a database-query count that does not grow with the number of repositories, while preserving its existing list response and alias metadata. This promotes residual intake 01a06e0c-3feb-7920-a6f2-961b03dc170d from STATE-WP-0081.

Scope includes the repository collection query, identity/alias projection, and focused regression tests. Pagination and changes to the public response shape are out of scope because removing the N+1 projection makes the existing fleet read bounded without a collection-contract migration. Dependencies: none.

Bulk-project repository identities

id: STATE-WP-0087-T01
status: done
priority: high
quality_dor: DoR-Ok
quality_dor_at: "2026-09-04"
quality_dor_by: codex

Load canonical and alias slug records once for the collection, construct the existing identity response from that bulk result, and suppress unrelated eager relationships. Preserve compatibility for repositories created before the slug registry backfill. Done when focused API tests prove alias fidelity and a constant repository-slug query count, and the full repository router suite is green.

Acceptance

• Collection identity projection does not issue queries per repository.
• Canonical, alias, and pre-backfill compatibility responses are preserved.
• Focused and repository router tests pass.
• Live or production-shaped timing evidence is recorded.

Completion — 2026-09-05

Reviewed and completed the pending collection patch. The collection performs one repository/domain SELECT and one registry SELECT; a domain filter adds one scalar domain-ID SELECT. Suppressing eager domain relationships also prevents unrelated topics, goals, repositories, and rename history from being loaded. The response shape, aliases, pre-backfill behavior, conflict responses, and classification filters are preserved.

Validation: uv run pytest -q tests/test_repository_collection_projection.py tests/test_routers_core.py tests/test_repository_alias_routing.py tests/test_repository_rename_api.py tests/test_repository_rename_persistence.py tests/test_sbom_nexus_compat.py — 102 passed in 129.14s. git diff --check also passed. Production-shaped timing evidence: docs/evidence/repository-collection-projection-20260905.md (4, 131, and 501 repositories; constant two/three SELECTs without/with a domain filter).

No implementation residuals. This closes the scope promoted from intake 01a06e0c-3feb-7920-a6f2-961b03dc170d. Deployment is outside this source-change workplan; the timing evidence uses the local test database, not the running central service. The separate retirement and HA workplans retain their gates.