supabase / auth / 21566798362

69%

master: 69%

Pull #2361

github

perf(models): optimize user loading from 3 queries to 1 using JSON aggregation

Reduces FindUserWithRefreshToken from 3 separate queries (user + identities + factors)
to a single query using json_agg subqueries. This optimization impacts the
/token endpoint (~45% of total traffic) and /user endpoint by eliminating 2 database
round-trips per call. Since both of those endpoints call FindUserWithRefreshToken twice
it removes 4 database roundtrips per request.

Performance impact:
- Query execution: 399µs → 209µs (47.6% faster)
- Memory allocations: 18.4KB → 7.1KB (61% reduction)
- Allocation count: 299 → 117 allocs (61% reduction)
- /token throughput: +20.3% (55.18 vs 45.87 req/s)
- /token latency: -16.8% (181ms vs 218ms)

In local testing with the database in docker, the optimization allows 20% higher
throughput on the POST /token endpoint.

Replaces Pop ORM .Eager() pattern with explicit SQL column enumeration and
coalesce(json_agg()) for related entities. No changes to User struct or API.

55 of 59 new or added lines in 1 file covered. (93.22%)

14948 of 21692 relevant lines covered (68.91%)

81.04 hits per line