|Apache Arrow 5 Improves Asynchronous Scanner|
|Written by Kay Ewbank|
|Monday, 16 August 2021|
Apache Arrow 5 has been released, alongside Apache Arrow Rust 5. Both versions have a number of improvements, including a better asynchronous scanner for the Dataset layer. This is the first release where the Rust projects have moved to separate repositories outside the main Arrow monorepo.
The improvements to the Dataset layer start with the asynchronous scanner introduced in Arrow 4. This has been improved with truly asynchronous readers implemented for CSV, Parquet, and IPC file formats and file-level parallelism added.
The compute layer has lots of new scalar functions, including 30 new scalar arithmetic and math functions, a collection of scalar bitwise functions, 21 scalar string functions, 16 scalar temporal functions and a group of 'other' scalar functions such as case_when, coalesce, if_else, and make_struct.
The Flight support has been improved in Arrow's Go implementation, and now supports custom metadata and middleware.
Java improvements include Improved support for extension types using a complex storage type, e.g. struct, map or union.
Python support has been extended, with the ability to scan files asynchronously in Datasets. The developers say this should provide better performance in environments where I/O can be slow, such as with remote sources.
The developers working on the R support in Arrow say they've more than doubled the number of functions you can call on Arrow Datasets inside dplyr::filter(), mutate(), and arrange(), including many more string, datetime, and math functions.The support for the Arrow C interface has been deepened. This allows integration with other projects, such as DuckDB.
Apache Arrow 5 is available now.
or email your comment to: firstname.lastname@example.org
|Last Updated ( Monday, 16 August 2021 )|