rs-domain-data
/rs-domain-data
Pomaga w pracy z bazami, pamięcią podręczną i magazynami analitycznymi w Rust: sqlx, transakcje, migracje, zapytania i wzorce cache.
- Autor
- bassem-elsodany ↗
- Repozytorium
- rs-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 11.11.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 9
- Skrypty
- 0
- Komentarze HTML
- 2
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 11.11.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
- SKILL.md zawiera komentarze HTML (2). Nie widać ich w podglądzie Markdown, a model je czyta.
Co robi i kiedy
Pomaga w pracy z bazami, pamięcią podręczną i magazynami analitycznymi w Rust: sqlx, transakcje, migracje, zapytania i wzorce cache.
Oryginalny opis (EN)
Use when working with databases, caches or analytics stores from Rust — choosing sqlx/diesel/sea-orm, connection pools, transactions, migrations, query performance, N+1, pagination, caching patterns (cache-aside, write-through, TTL, invalidation, stampede), Redis, ClickHouse, time-series and append-only event stores. Triggers: database, SQL, Postgres, MySQL, SQLite, sqlx, diesel, sea-orm, ORM, connection pool, deadpool, bb8, transaction, isolation level, deadlock, migration, schema change, expand contract, index, N+1, pagination, keyset, OFFSET, query performance, EXPLAIN, statement timeout, c
Jak działa
- Core Question
- Thinking Prompt
- Choosing the access layer (relational)
- Rules that prevent most incidents
- Caching in one table (details guides/caching.md)
- Trace
- Rules (5)
- Related Skills
Podgląd SKILL.md
--- name: rs-domain-data description: "Use when working with databases, caches or analytics stores from Rust — choosing sqlx/diesel/sea-orm, connection pools, transactions, migrations, query performance, N+1, pagination, caching patterns (cache-aside, write-through, TTL, invalidation, stampede), Redis, ClickHouse, time-series and append-only event stores. Triggers: database, SQL, Postgres, MySQL, SQLite, sqlx, diesel, sea-orm, ORM, connection pool, deadpool, bb8, transaction, isolation level, deadlock, migration, schema change, expand contract, index, N+1, pagination, keyset, OFFSET, query performance, EXPLAIN, statement timeout, cache, caching, Redis, Valkey, moka, TTL, cache-aside, write-through, invalidation, stampede, thundering herd, ClickHouse, analytics, OLAP, batch insert, time series, event store, outbox" user-invocable: false --- # Data — databases, caches and analytics stores > **Layer 3: Domain** · why a data-heavy service is designed the way it is > > Depth: `guides/database.md` · `guides/caching.md` · `guides/analytics-stores.md`. Always combine with **rs-ports-adapters** (every store sits behind a port) and **rs-reuse-first** (official driver, not hand-written wire protocol). ## Core Question **What must be true about this data at rest and in motion — consistent, durable, fast, or cheap — and which store gives exactly that and no more?** ## Thinking Prompt 1. **System of record?** One store owns each fact. Caches, search indexes and analytics copies are *derived* and must be rebuildable. 2. **Consistency need:** read-your-writes? serialisable? "eventually, within N seconds"? Name it — it decides cache and replication design. 3. **Access pattern:** point lookups, range scans, aggregations over millions of rows, append-only events? Pick the store for the pattern (row store, cache, columnar). 4. **Failure behaviour:** what does the caller see when the store is slow, down, or returns a conflict? (timeouts, classified errors, retry budget — rs-domain-error) 5. **Change:** how will the schema evolve while the old and new versions run side by side? (expand/contract) 6. **Replaceability:** port + memory fake + contract tests (rs-ports-adapters) — the product may change, the contract won't. ## Choosing the access layer (relational) | Need | Choice | Notes | |---|---|---| | async, plain SQL, compile-time-checked queries, migrations built in | **`sqlx`** (`query!`, `query_as!`, `migrate!`, `#[sqlx::test]`) | checked macros need `DATABASE_URL` or `cargo sqlx prepare` (offline data committed) | | compile-time schema DSL, mature, sync (or `diesel-async`) | `diesel` (+ `diesel_migrations`) | strong type safety; steeper learning curve | | ORM-style models and relations, async | `sea-orm` (+ `sea-orm-migration`) | generated entities; keep them in the adapter, not the domain | | bare driver | `tokio-postgres`, `rusqlite` | when you need control; pool with `deadpool-postgres`/`bb8` | | non-relational | `mongodb` (official), `redis`, `clickhouse` (official), `object_store` | registry: rs-reuse-first `registry/standards.md` | Whatever you pick: **entities in the adapter, domain types in the domain** (rs-ports-adapters `port-row-types`). ## Rules that prevent most incidents | Rule | Detail | |---|---| | **Timeouts everywhere** | pool acquire timeout, connect timeout, statement timeout (`statement_timeout`), overall request deadline; a missing timeout is an outage amplifier | | **Pool sized deliberately** | `max_connections` across *all* instances ≤ the DB's limit (leave headroom); small pools + queueing beat huge pools; behind PgBouncer in transaction mode avoid session state and check prepared-statement support | | **Transactions are short** and never span network calls or user think-time; no `.await` on unrelated I/O inside one; retry only the *whole* transaction on serialization failure (`40001`) / deadlock (`40P01`) with backoff | | **No N+1** | load children with one batched query (`WHERE id = ANY($1)` in Postgres / `IN (…)`) or a join; dataloader for GraphQL; test with query-count assertions | | **Keyset pagination** for large/unbounded lists (`WHERE (created_at, id) < ($1,$2) ORDER BY … LIMIT n`); `OFFSET` degrades linearly | | **Parameterised queries only** | never `format!` values into SQL (injection); identifiers from an allow-list | | **Idempotent writes** | `INSERT … ON CONFLICT DO NOTHING/UPDATE` with a natural or client-supplied key; the basis of safe retries (rs-domain-distributed) | | **Money and time** | `NUMERIC`/`rust_decimal` (never float); `timestamptz` stored in UTC | | **Migrations: expand → migrate → contract** | additive change first, backfill, switch code, drop later; never edit an applied migration; indexes with `CREATE INDEX CONCURRENTLY` (outside a transaction); set `lock_timeout` for DDL | | **Test against the real engine** | `#[sqlx::test]` or testcontainers; remove containers afterwards (rs-cleanup) | ## Caching in one table (details `guides/caching.md`) | Pattern | Use when | Watch | |---|---|---| | **cache-aside** (read: cache → miss → load → fill) | default for read-heavy data | stampede (single-flight), stale reads after writes | | **write-through** | reads must see writes at once | write latency, partial failure ordering | | **write-behind** | very write-heavy, loss tolerable | data loss on crash — rarely acceptable | | **TTL + jitter** | bounded staleness | synchronized expiry → herd; add random jitter | | **explicit invalidation** (events, versioned keys) | strict freshness | missed invalidation = stale forever → TTL as a safety net | | **negative caching** | many lookups for non-existent keys | short TTL | | **in-process (`moka`)** vs **shared (Redis)** | per-instance hot set vs cross-instance | in-process invalidation across instances needs events; Redis adds a network hop and a failure mode | **Never cache authorization decisions without a version/epoch in the key** (rs-domain-policy) and never use a cache as the source of truth unless it is configured and operated as a database. ## Trace | Direction | Go to | |---|---| | ↑ Up | rs-domain-fintech (money, ledgers), rs-domain-cloud-native (connections per pod, secrets), rs-domain-distributed (idempotency, outbox, consistency) | | ↓ Down | rs-ports-adapters (port/adapter, contract tests), rs-error-handling + rs-domain-error (classification, retry), rs-concurrency (never hold a lock across `.await`), rs-lifecycle (pools, shutdown) | | → Delivery | rs-testing (real DB in tests), rs-observability (query latency, pool wait metrics), rs-performance (profiling queries before Rust), rs-cleanup (containers/volumes) | <!-- rules-index:start --> ## Rules (5) Depth rules under `rules/` — open the one the question needs, not all of them. **cache-** - [`cache-stampede-and-jitter`](rules/cache-stampede-and-jitter.md) — Single-flight on miss, jitter on TTL, bound the size **data-** - [`data-keyset-pagination`](rules/data-keyset-pagination.md) — Page large result sets with a keyset (seek) cursor, not OFFSET - [`data-migrations-expand-contract`](rules/data-migrations-expand-contract.md) — Change schemas in compatible steps so old and new code can run together - [`data-short-transactions`](rules/data-short-transactions.md) — Keep transactions short, on one store, with no foreign I/O inside - [`data-timeouts-everywhere`](rules/data-timeouts-everywhere.md) — Every store call has a deadline: connect, pool acquire, statement, and overall request <!-- rules-index:end --> --- ## Related Skills | When | See | |---|---| | Making the store replaceable | rs-ports-adapters | | Picking the driver | rs-reuse-first | | Money and decimals | rs-domain-fintech | | Retry and recovery | rs-domain-error |
Pliki w paczce (9)
- SKILL.md 8 KB
- guides/analytics-stores.md 3 KB
- guides/caching.md 5 KB
- guides/database.md 4 KB
- rules/cache-stampede-and-jitter.md 1 KB
- rules/data-keyset-pagination.md 1 KB
- rules/data-migrations-expand-contract.md 1 KB
- rules/data-short-transactions.md 1 KB
- rules/data-timeouts-everywhere.md 1 KB
Wersje
W bazie nie ma innych skilli o nazwie rs-domain-data.
Migawka bazy z 13.11.2026 · wersja paczki 39cdde363ffa (skrót SHA-256 całej paczki)
Pobrania w ostatnich 30 dniach: 0 · ranking
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100rust-database
Projektuje i testuje dostęp do baz w Rust: SQLx, Diesel, SeaORM, migracje, transakcje, pule połączeń i prawdziwe bazy testowe. - 100sqlx-code-review
Przegląda kod Rust z sqlx pod kątem sprawdzania zapytań w czasie kompilacji, zarządzania pulą połączeń, migracji i użycia PostgreSQL, w tym tryb offline, mapowanie typów i transakcje. - 100caching-strategies
Projektuje wielowarstwowe architektury cache dla aplikacji webowych: wzorce cache-aside i write-through, TTL, unieważnianie, Redis, CDN, nagłówki Cache-Control i ochrona przed stampede. - 90rc-sqlx
Pisze i sprawdza dostęp do PostgreSQL w Rust przez SQLx 0.8+: pule, transakcje, migracje i testy bazy. - 100update-sqlx
Opisuje bezpieczną aktualizację offline cache zapytań SQLx, którą trzeba wykonać po każdej zmianie zapytań SQL. - 90caching-strategies
Podpowiada wzorce cache: cache-aside, write-through, zapobieganie stampede, nagłówki CDN, poziomy L1, L2 i L3 oraz unieważnianie.