Merge pull request #168 from justrach/release/1.0.30

release: merge v1.0.30 into main

release: v1.0.29

release: v1.0.28

Merge pull request #120 from justrach/release-1.0.27-pr

Fix release process and bump to v1.0.27

release: bump TurboAPI to v1.0.24 (#112)

chore(release): bump version to 1.0.23

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Release v1.0.22

Release v1.0.21

Release v1.0.20

v1.0.17: raw pg.zig queries, LRU cache, pool rotation, pipelining (#61)

* feat: raw pg.zig query path from Python + bool param fix

Two changes:
1. Updated pg.zig fork (bool param encoding fix for []const u8 values)
   - Fixes #59: pg_type query with boolean params no longer crashes
2. Added _db_query_raw C API for direct pg.zig queries without HTTP
   - Fixes #60: TurboPG can now query Postgres directly via Zig
   - db.query_native() bypasses the HTTP server entirely
   - Enables fair driver-to-driver benchmarks against asyncpg

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* refactor: cache _db_query_raw availability at init, not per-call

Move hasattr check from _query_native() to _connect(). Stores result
in self._native_raw bool. Zero overhead on the query hot path.

The HTTP path (db_get/db_query decorators) never touches turbopg/client.py
- it goes server.zig -> handleDbRoute -> db.zig directly. So the raw
query path has zero impact on HTTP performance.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* chore: update pg.zig to include LRU cache, pool rotation, pipelining

Points to perf/zero-copy-and-lru-cache branch with:
- Bounded LRU stmt cache (256 max, evicts oldest)
- Pool connection rotation (max_queries, max_lifetime)
- Query pipelining (batch multiple queries in one round trip)
- Bool param encoding fix

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* v1.0.17: raw pg.zig query path, LRU stmt cache, pool rotation, pipelining

- feat: _db_query_raw C API for direct pg.zig queries without HTTP (#60)
- fix: boolean param encoding in pg.zig fork (#59)
- perf: bounded LRU stmt cache (256 max, evicts oldest)
- perf: pool connection rotation (max_queries, max_lifetime)
- feat: query pipelining (batch queries in one round trip)
- refactor: cache _native_raw at init, zero per-call overhead

Benchmarks: 1.25x faster than asyncpg on concurrent queries,
151k+ req/s on HTTP path (no regression).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>