[
https://issues.apache.org/jira/browse/SPARK-26858?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16772266#comment-16772266
]
Bryan Cutler commented on SPARK-26858:
--------------------------------------
[~hyukjin.kwon] actually {{pyarrow.Table.from_batches}} does require a schema,
but it's a little tricky. In python, to create a {{RecordBatch}} object, a
schema is always required. Then it attaches to the {{RecordBatch}} object, so
you can always get the schema from a batch and do things like
{{Table.from_batches}} with just a list of batches. Look at the full api
[here|https://arrow.apache.org/docs/python/generated/pyarrow.Table.html#pyarrow.Table.from_batches]
and you can see the schema is optional and it just pulls from the first batch.
Serialized record batches are a bit different and do not contain the schema,
which is why the stream protocol goes {{Schema, RecordBatch, RecordBatch, ..}}.
So you can't just send serialized Arrow batches and then build a Table.
Hopefully that makes sense :P
Back on your sequence diagram at step (9), does R write the data frame row by
row bytes over the socket? I'm wondering how it gets serialized to see where
Arrow might best help out.
> Vectorized gapplyCollect, Arrow optimization in native R function execution
> ---------------------------------------------------------------------------
>
> Key: SPARK-26858
> URL: https://issues.apache.org/jira/browse/SPARK-26858
> Project: Spark
> Issue Type: Sub-task
> Components: SparkR, SQL
> Affects Versions: 3.0.0
> Reporter: Hyukjin Kwon
> Assignee: Hyukjin Kwon
> Priority: Major
>
> Unlike gapply, gapplyCollect requires additional ser/de steps because it can
> omit the schema, and Spark SQL doesn't know the return type before actually
> execution happens.
> In original code path, it's done via using binary schema. Once gapply is done
> (SPARK-26761). we can mimic this approach in vectorized gapply to support
> gapplyCollect.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
