CHANGELOG

[1.1.5] - 2026-09-26

🚀 Performance

• Prepare field-specific decoders and the update validation route at registration. In a local paired conversion benchmark on 20,000 rows with scalar, datetime, and list fields, median time fell from 0.194 to 0.166 seconds.
• Cached type inspection while decoding rows. In a local 10,000-row benchmark, median query time fell from 0.098 to 0.068 seconds (sync) and from 0.108 to 0.074 seconds (async).
• Bound LIMIT and OFFSET as SQL parameters. In a local 1,000-query pagination benchmark, median time fell from 0.0341 to 0.0315 seconds.
• Removed unused JSON conversion in __in filters. Building 500 filters of 1,000 integers fell from 0.138 to 0.091 seconds locally.

🐞 Fixed

• Updated documentation dependencies to address known vulnerabilities and limited the optional cryptography dependency to a fixed release on Python 3.10+.
• Protected in-flight SQLite calls after async cancellation and rolled back open transactions; sync transactions also roll back on interruptions.
• update() now validates changed values, with affected-row validation for models that have custom validators or configuration.
• register() now raises SchemaMismatchError when an existing table's columns, types, primary key, nullability, or FTS searchable columns differ from the model.
• Adding full-text search to an existing table now indexes existing rows immediately.

📚 Docs

• Corrected the error and exception guides, backup example, projection advice, batched async reads, and continuation-based pagination.

⚠️ Compatibility

• On Python 3.9, the encryption, all, and dev extras no longer install cryptography. Core Python 3.9 support remains.
• Invalid partial updates now raise Pydantic ValidationError. Incompatible existing tables raise SchemaMismatchError on registration; migrate them explicitly.

---

[1.1.4] - 2026-06-22

🐞 Fixed

• Validated all generated SQLite identifiers and rejected unknown composite-index fields before constructing schema SQL.
• Hardened string filters so unsupported raw SQL fragments now raise QueryValidationError instead of being passed through to SQL.
• Aligned async query validation with the sync implementation, including unknown-field checks and model-aware filter value serialization.
• Fixed async filtering for date, datetime, list, and dict values.
• Fixed __in handling so normal lists/tuples/sets work consistently and empty collections return no rows.
• Fixed offset-only pagination by emitting SQLite's required unlimited LIMIT clause.
• Tightened limit / offset validation to reject negative values, floats, booleans, and strings.
• Added a clear TypeError when insert_many() receives mixed model types.
• Serialized reads and connection shutdown with the existing locks so close() cannot race active queries.
• Protected vacuum() and async execute_raw() with the existing connection locks.

🧪 Tests

• Added regression tests for unsafe string filters, async field validation, serialized filter values, __in, offset-only pagination, strict pagination types, and mixed-model bulk inserts.

📚 Docs

• Added an APSW full-access implementation plan.

⚠️ Compatibility

• String filters now support only simple comparisons such as "age > 10" or "name = 'Alice'". Use keyword filters such as age__gte=10 for advanced filtering.
• Python 3.9 remains supported for the core package. Some optional and development dependencies cannot provide their latest security-fixed wheels on Python 3.9; use Python 3.10+ for the speed, encryption, and dev extras when processing untrusted input.

---

[1.1.1] - 2026-05-16

🚀 Added

• Explicit Transactions: Added atomic() context manager to NyanSQLite and async with atomic() to NyanSQLiteAIO for manual transaction control.
• Nested Transactions: Added support for nested atomic() blocks.

🔄 Changed

• Thread Safety: Improved thread safety by switching to threading.RLock in NyanSQLite.
• Async Safety: Implemented re-entrant async lock in NyanSQLiteAIO to prevent deadlocks when using atomic().

---

[1.1.0] - 2026-05-16

🚀 Added

• Asynchronous Support: Full support for asyncio via NyanSQLiteAIO class.
• Improved Performance: Optimized read operations by minimizing thread context switching and processing rows efficiently in asyncio.to_thread.
• Documentation Updates: Added English and Japanese documentation for asynchronous usage.

🔄 Changed

• Internal optimization for query, select, and search methods in NyanSQLiteAIO.
• Optimized read operations in synchronous NyanSQLite class by minimizing lock duration.

---

[1.0.1] - 2026-05-15

🐞 Fixed

• Minor bug fixes and performance improvements.

---

[1.0.0] - 2026-05-15

🚀 Added

• Pydantic v2 support: Models can be used directly as database schemas.
• Django-like Query Syntax: Support for intuitive filtering such as __gte, __in, __like, etc.
• FTS5 Full-Text Search: Fast full-text search capabilities using SQLite's FTS5 extension.
• Automatic Index Management: B-tree indexes are automatically created using Indexed[T] and UniqueIndexed[T] annotations.
• Composite Indexes: Support for CompositeIndex via Pydantic's Field extra metadata.
• Transparent Type Handling: Automatically handles complex types like dict and list by serializing them to JSON.
• WAL Mode Support: Write-Ahead Logging is enabled by default for better performance and concurrency.
• Context Manager Support: NyanSQLite can be used as a context manager for automatic connection closing.

🔄 Changed

• Initial public release of NyanSQLite.

---