rc-sqlx

SQLx 0.8 + PostgreSQL

Async SQL for Rust without a heavy ORM. Target: sqlx 0.8.x, PostgreSQL 16 (compatible with 14+). Complements rc-sql (general SQL design) and rc-axum (HTTP layer).

Core workflow

1. Schema first — migrations under version control; no ad-hoc prod DDL from app code without a migration path.
2. Pool — one PgPool in app state; tune max_connections to VPS/DB limits.
3. Queries — always bind parameters ($1, $2); never format! user input into SQL.
4. Mapping — query_as / FromRow or compile-time query_as! when DATABASE_URL is available at build.
5. Transactions — multi-step writes use pool.begin() → commit/rollback.
6. Security — least-privilege DB role; secrets only in env. Read references/security.md.
7. Test — sqlx::test or isolated DB. Read references/testing.md.

Reference guide

Topic	File	Load when
Pool, query_as, transactions, macros	references/guide.md	Implementing queries
Injection, roles, secrets	references/security.md	Reviewing data access
Migrations (sqlx-cli / embed)	references/migrations.md	Schema changes
Integration tests	references/testing.md	DB tests

Must do

• Use bound parameters for all dynamic values (Postgres $n).
• Prefer query_as::<_, T> or query_as! into serde/FromRow structs.
• Set pool size consciously (small VPS: start with 5–10).
• Run migrations in deploy pipeline before app start (or controlled startup migrate).
• Map sqlx::Error to domain/HTTP errors without leaking SQL internals to clients.

Must not do

• Share one long-lived transaction across HTTP requests.
• Use the superuser role for the app in production.
• Commit migrations that are not reversible or documented when data loss is possible.
• Ignore offline mode needs for CI (SQLX_OFFLINE=true + .sqlx data) if using macros without a live DB.

Output template

1. Migration SQL (up, and down if required)
2. Rust types (FromRow / domain models)
3. Query functions (pool/tx receiver)
4. Error mapping
5. Tests + cargo test / sqlx prepare notes