Skip to content

CHANGELOG.md

Latest commit

 

History

History
65 lines (63 loc) · 4.54 KB

CHANGELOG.md

File metadata and controls

65 lines (63 loc) · 4.54 KB

dataset ChangeLog

The changelog has only been started with version 0.3.12, previous changes must be reconstructed from revision history.

  • 2.0.0: Major modernization and type annotations
    • Type annotations: Full mypy --strict compliance across all modules
    • PEP 561: Added py.typed marker for downstream type checking
    • New types: Exported OutRow, RowFactory, QueryError for downstream use
    • RowFactory: The row_type parameter is now typed as Callable[[Iterable[tuple[str, Any]]], OutRow] instead of type
    • QueryError: New exception subclass of DatasetError for invalid filter operations
    • primary_type: Changed from Types to ColumnType (SQLAlchemy TypeEngine) — the actual accepted type
    • insert/insert_ignore/upsert: Return type changed from int | bool to Any (primary keys can be any type)
    • Removed banal dependency: Replaced ensure_list with typed ensure_strings utility
    • update_many: Fixed mutation of input rows — rows are now copied before modification
    • Dev tooling: Added mypy to dev dependencies, make lint now runs both ruff and mypy
    • Build system: Migrated from setuptools to modern pyproject.toml with Hatchling (PEP 621)
    • Linting: Replaced flake8 with ruff for faster, more comprehensive linting
    • CI/CD: Updated GitHub Actions to use modern action versions (checkout@v4, setup-python@v5)
    • SQLAlchemy 2.x: Full support for SQLAlchemy 2.0+ with backward compatibility to 1.4.0
    • Transaction handling: Fixed autobegin semantics and DDL lock contention for SQLAlchemy 2.x
    • Testing: Switched from nose to pytest, improved test fixtures and cleanup
    • Database support: Added lock timeout configurations for PostgreSQL and MySQL in CI
    • Python support: Now requires Python 3.10+, tested on 3.10-3.13
    • Documentation: Updated installation instructions, copyright year, and added comprehensive CLAUDE.md
    • Metadata: Changed development status from Alpha to Production/Stable
    • License: Renamed LICENSE.txt to LICENSE for standard convention
    • Dependencies: Updated SQLAlchemy constraint to allow versions up to 3.0.0
  • 1.6.2: Fix distinct() to respect _limit and _offset parameters (#424).
  • 1.6.1: Fix add_column method compatibility with Alembic 1.11+ (#423).
  • 1.6.0: Pin SQLAlchemy below 2.0.0 for compatibility.
  • 1.5.2: Consider primary key when checking for indexes (#382). Add missing arguments for query method (#391).
  • 1.5.1: Improve row conversion compatibility across SQLAlchemy 1.3 and 1.4.
  • 1.5.0: Add support for custom SQLite pragmas via on_connect_statements parameter. Switch from nose to pytest for testing.
  • 1.2.0: Add support for views, multiple comparison operators. Remove support for Python 2.
  • 1.1.0: Introduce types system to shortcut for SQLA types.
  • 1.0.0: Massive re-factor and code cleanup.
  • 0.6.0: Remove sqlite_datetime_fix for automatic int-casting of dates, make table['foo', 'bar'] an alias for table.distinct('foo', 'bar'), check validity of column and table names more thoroughly, rename reflectMetadata constructor argument to reflect_metadata, fix ResultIter to not leave queries open (so you can update in a loop).
  • 0.5.7: dataset Databases can now have customized row types. This allows, for example, information to be retrieved in attribute-accessible dict subclasses, such as stuf.
  • 0.5.4: Context manager for transactions, thanks to @victorkashirin.
  • 0.5.1: Fix a regression where empty queries would raise an exception.
  • 0.5: Improve overall code quality and testing, including Travis CI. An advanced getitem syntax which allowed for the specification of primary keys when getting a table was dropped. DDL is no longer run against a transaction, but the base connection.
  • 0.4: Python 3 support and switch to alembic for migrations.
  • 0.3.15: Fixes to update and insertion of data, thanks to @cli248 and @abhinav-upadhyay.
  • 0.3.14: dataset went viral somehow. Thanks to @gtsafas for refactorings, @alasdairnicol for fixing the Freezfile example in the documentation. @diegoguimaraes fixed the behaviour of insert to return the newly-created primary key ID. table.find_one() now returns a dict, not an SQLAlchemy ResultProxy. Slugs are now generated using the Python-Slugify package, removing slug code from dataset.
  • 0.3.13: Fixed logging, added support for transformations on result rows to support slug generation in output (#28).
  • 0.3.12: Makes table primary key's types and names configurable, fixing #19. Contributed by @dnatag.