Daft 0.0.17 Release Notes#

Highlights from 0.0.17:

  • Improved local development experience with caching of intermediate steps

  • .cast is now available for casting between types - users can now coerce Python types such as PY[int] to Daft-native primitives such as INTEGER

  • Daft now includes a .explain operation on dataframes, which explains how a dataframe will be executed by pretty-printing the execution graph

New Features#

Caching of intermediate steps#

The Dataframe now caches intermediate steps whenever a user executes the dataframe. Subsequent operations on an executed dataframe will perform an in-memory scan of the materialized data instead of re-executing the entire plan.

See: #297

Cast Expression#

Users can now cast between types, such as INTEGER to FLOAT or PY[int] to INTEGER.

See: #293

Explain Dataframe Operation#

Users can now view the tree of operations that Daft will execute for their dataframe, allowing for optimizations and debugging.

See: #301

Enhancements#

  • Add HTTP URL handling for DataFrame.from_parquet #292

Bug Fixes#

  • Fix quantile calculations in sort #300

Closed Issues#

  • Fix loading Parquet file from https URLs #288

  • Cast operator for Expressions #180