Release 1.6.0

Added

• Add DataFusion datasource implementation in Python for pandas and DataFrame Interchange by @jonmmease in #438. This greatly speeds up default DataFusion data transformer by only serializing the columns that are used in the chart to arrow. Previously, the entire pandas DataFrame was serialized to arrow.

Updated

• Update sum([null]) aggregation to be null to match Vega 5.26.2 by @jonmmease in #441

Fixed

• Fix parent dataset resolution in facet lift planner stage by @jonmmease in #439

Full Changelog: v1.5.1...v1.6.0

Release 1.6.0-rc1

Added

• Add DataFusion datasource implementation in Python for pandas and DataFrame Interchange by @jonmmease in #438

Fixed

• Fix parent dataset resolution in facet lift planner stage by @jonmmease in #439

Full Changelog: v1.5.1...v1.6.0-rc1

Release 1.5.1

Fixed

• Fixes to support interval_selection_map_quakes example by @jonmmease in #435
• Support threshold aggregation example by @jonmmease in #436

Full Changelog: v1.5.0...v1.5.1

Release 1.5.0

Added

• Add ChartState for managing interactive chart state in Python by @jonmmease in #426
• ChartState Updates by @jonmmease in #428
• Add support for timeOffset expression function by @jonmmease in #427
• Wrap ChartState in pure Python class with typed API by @jonmmease in #429
• Add parquet support for DuckDB connection by @jonmmease in #430

Fixed

• Don't fuse non-inline datasets with multiple children by @jonmmease in #431

Full Changelog: v1.4.4...v1.5.0

Release 1.4.4

Added

• Experimental s3 and parquet support by @jonmmease in #417
• Add to_utc_timestamp support in more dialects by @jonmmease in #420

Full Changelog: v1.4.3...v1.4.4

Release 1.4.3

Updated

• Support aggregate transforms in facet group marks by @jonmmease in #411
• Support format function with empty format string, make sum of nulls evaluate to zero by @jonmmease in #414

Fixed

• SnowparkDataset: Cast arrow table with safe=False to allow precision truncation by @jonmmease in #410
• Fix cast timestamp to integer millisecond and add stack timestamp tests by @jonmmease in #413

Full Changelog: v1.4.2...v1.4.3

Release 1.4.2

Updated

• Update to DataFusion 32 by @jonmmease in #393
• Add fuse_datasets planner configuration for combining datasets during planning by @jonmmease in #407

Fixed

• Make patching fail if there are changes to any dataset by @jonmmease in #406

Full Changelog: v1.4.1...v1.4.2

Release 1.4.1

Updated

• Update DataFusion 28, Arrow to 43, sqlparser to 0.35, and other dep updates by @jonmmease in #388

Fixed

• Fix edge cases: Use TRY_CAST for type conversions, convert cols to datetime before stringification by @jonmmease in #378
• Fix parsing of single element scale range by @jonmmease in #384
• Allow scale.domain.fields to contain signals by @jonmmease in #385
• Handle domain data reference with list of fields by @jonmmease in #392
• Fix crash in bin transform with empty input table by @jonmmease in #395
• Fix patch_pre_transformed_spec when object is replaced by an array by @jonmmease in #398

Development

• Use pixi to manage development environment by @jonmmease in #375
• Unpin vl-convert-python now that 0.13.1 is out by @jonmmease in #383
• Free Python GIL on blocking operations to allow multi-threading runtime usage without deadlocks by @jonmmease in #387
• Rust 1.72.0 by @jonmmease in #389
• Remove panics from planning logic by @jonmmease in #399

Full Changelog: v1.4.0...v1.4.1

Release 1.4.0

Added

• Add q1 and q3 aggregations by @jonmmease in #340
• Add bitwise operators to expression language by @jonmmease in #359
• Support keeping select signals and datasets in pre-transformed specs by @jonmmease in #343
• Add pre_transform_spec support for inlining transformed data in arrow format by @jonmmease in #365
• Add Python Dataset interfaces by @jonmmease in #363
• Add lazy implementation of DataFrame Interchange Protocol to SqlDataset by @jonmmease in #373
• Add SnowparkDataset and date_to_utc_timestamp support across dialects by @jonmmease in #374

Updated

• Use fallback SQL connection when table registration is not supported by @jonmmease in #341
• Update to DataFusion 26, Arrow 40, sqlparser 0.34.0 by @jonmmease in #344
• Update to datafusion 27, arrow 42, pyO3 0.19, sqlparser-rs 0.35 by @jonmmease in #353
• Add dialect specialization for the row_number window function without ordering by @jonmmease in #346
• Default local_tz to get_local_tz() in vf.runtime functions by @jonmmease in #347
• Build abi3 wheels so we don't need to build separate wheels for each platform by @jonmmease in #349
• Add linux-arm64 builds by @jonmmease in #358
• SqlDataset.dialect doesn't have to be a classmethod by @jonmmease in #372

Fixed

• Remove some panics by @jonmmease in #339
• Filter out groupby columns that don't exist by @jonmmease in #362
• Flatten nested ternary statements to avoid stack overflow by @jonmmease in #366

Full Changelog: v1.3.0...v1.4.0

Release 1.4.0-rc5

Fixed

• Lower minimum pyarrow dependency back to v10 by @jonmmease in #368

Full Changelog: v1.4.0-rc4...v1.4.0-rc5