|Apache Arrow Reaches 1.0|
|Written by Kay Ewbank|
|Monday, 03 August 2020|
Apache Arrow 1.0 has been released, with a better metadata version and support with dictionary indices along with improved C++ libraries.
The metadata version is now at version V5, which does mean an incompatible change in the buffer layout of Union types. All other types keep the same layout as in V4. The improved dictionary indices can now be made on unsigned integers rather than only signed integers.
The Flight element of Arrow has also been updated. Flight provides stream management and is intended to overcome the problem that Apache Arrow's primary medium is in-memory data, but not all systems can be co-located. Flight now offers DoExchange, a fully bidirectional data endpoint, in addition to DoGet and DoPut, in C++, Java, and Python. The C++ and Python libraries for Flight now expose more options from gRPC, including the address of the client (on the server) and the ability to set low-level gRPC client options. Flight also supports mutual TLS authentication and the ability for a client to control the size of a data message on the wire. C++ support has also been improved, with better support for static linking with Arrow.
The compute kernel layer has been extensively reworked, and now offers a generic function lookup, dispatch and execution mechanism. Around 30 new array compute functions have also been added.
or email your comment to: firstname.lastname@example.org