Github user BryanCutler commented on the issue:
https://github.com/apache/spark/pull/15821
This has been updated after integrating changes made with @icexelloss and
@wesm. There has been good progress made and it would be great if others could
take a look and review/test this out.
The current state of `toPandas()` with Arrow has support for Datasets with
primitive, string, and timestamp data types. Complex types such as Structs,
Array, and Mapped are not yet supported but are a wip. There is a suite of
tests in Scala to test Dataset -> ArrowRecordBatch conversion and a collection
on JSON files that serve to validate the converted data is correct. Also,
added PySpark tests to verify Pandas frame is correct. It is compiled with the
current arrow master 0.1.1-SNAPSHOT at commit
https://github.com/apache/arrow/commit/7d3e2a3ab90324625b738e464a020758379f457a
The performance so far shows a significant increase and I will follow up
with a script to run and details of the results seen. Please ping me with any
questions on setting up the build of Arrow or running the benchmarks. It would
be great if this could be considered for Spark 2.2 as Arrow 0.2 will be
released soon and be able to support the functionality used here.
@holdenk @davies @rxin, I would love to hear your thoughts on this so far.
Thanks!
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
