← Back to index page
- Configurable DuckDB Batch Sizes: Added support for configuring DuckDB batch
sizes in Rift materialization jobs.
- Configurable Parquet Read-Ahead: Added support for configuring the number of
Parquet fragments read ahead when reading FileConfig sources.
- Zero-Copy Conversion from PyArrow to Pandas: Allows for zero-copy data
transformation when converting from PyArrow to Pandas for improved
performance.
- Improved Calculation expression parsing error messages.
- Added support for Realtime Feature Views with and without Transformation
functions to belong to the same FeatureService.