Apache Arrow 16.1.0 Release
Published
14 May 2024
By
The Apache Arrow PMC (pmc)
The Apache Arrow team is pleased to announce the 16.1.0 release. This is a minor release that includes 34 resolved issues from 16 distinct contributors. See the Install Page to learn how to get the libraries for your platform.
The release notes below are not exhaustive and only expose selected highlights of the release. Other bugfixes and improvements have been made: we refer you to the complete changelog.
C++ notes
The scratch space required by some Scalar subclasses is now immutable after
initialization (GH-40069). This fixes thread-safety bugs when this scratch
space was lazily initialized, but introduces an API incompatibility because
writing to the value member of some concrete Scalar subclasses is not
allowed anymore. Affected classes include BaseBinaryScalar, BaseListScalar,
SparseUnionScalar, DenseUnionScalar and RunEndEncodedScalar.
The bit_width and byte_width methods on ExtensionType now return the
corresponding value for the underlying storage type (GH-41353).
Parquet
A regression that prevented reading BYTE_STREAM_SPLIT columns with null values was fixed (GH-41562).
C# notes
- Recompute a sliced array’s null count on demand when it is unknown (GH-41136)
- Support writing sliced arrays in the Arrow IPC format (GH-40517, GH-41225, GH-41231)
- Bug fixes for union array behaviour (GH-41137, GH-41140)
Go Notes
- Enable support for reading date64 from CSV (GH-41594)
- Update MarshalJSON() for Float32 and Float64 to be able to handle NaN, +Inf and -Inf values (GH-40563)