[
https://issues.apache.org/jira/browse/SPARK-23325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16391768#comment-16391768
]
Ryan Blue commented on SPARK-23325:
-----------------------------------
By exposing an interface that uses UnsafeRow, don't we already have this
problem? The only difference is that UnsafeRow is harder to produce. We also
have a write interface in v2 that exposes InternalRow. I think now is the time
to start documenting these so we can officially support InternalRow instead of
effectively supporting InternalRow.
UnsafeRow would benefit from more documentation, too. To find out how to use
the write interfaces, I ended up using EXPLAIN CODEGEN on a bunch of different
queries and looking at the results, then inspecting the writers to find out the
in-memory representation.
As for the columnar format, I see that as a nice-to-have. The v2 API is based
on rows for a good reason, and we need to document and support that row format.
Unless we are going to change v2 to a columnar API, stabilizing and documenting
the columnar format doesn't help much.
What work needs to be done here to make InternalRow viable? If it is to
document the values used to internally represent different types, I can help
out with that. I already have matching representations documented in the
Iceberg spec anyway.
> DataSourceV2 readers should always produce InternalRow.
> -------------------------------------------------------
>
> Key: SPARK-23325
> URL: https://issues.apache.org/jira/browse/SPARK-23325
> Project: Spark
> Issue Type: Sub-task
> Components: SQL
> Affects Versions: 2.3.0
> Reporter: Ryan Blue
> Priority: Major
>
> DataSourceV2 row-oriented implementations are limited to producing either
> {{Row}} instances or {{UnsafeRow}} instances by implementing
> {{SupportsScanUnsafeRow}}. Instead, I think that implementations should
> always produce {{InternalRow}}.
> The problem with the choice between {{Row}} and {{UnsafeRow}} is that neither
> one is appropriate for implementers.
> File formats don't produce {{Row}} instances or the data values used by
> {{Row}}, like {{java.sql.Timestamp}} and {{java.sql.Date}}. An implementation
> that uses {{Row}} instances must produce data that is immediately translated
> from the representation that was just produced by Spark. In my experience, it
> made little sense to translate a timestamp in microseconds to a
> (milliseconds, nanoseconds) pair, create a {{Timestamp}} instance, and pass
> that instance to Spark for immediate translation back.
> On the other hand, {{UnsafeRow}} is very difficult to produce unless data is
> already held in memory. Even the Parquet support built into Spark
> deserializes to {{InternalRow}} and then uses {{UnsafeProjection}} to produce
> unsafe rows. When I went to build an implementation that deserializes Parquet
> or Avro directly to {{UnsafeRow}} (I tried both), I found that it couldn't be
> done without first deserializing into memory because the size of an array
> must be known before any values are written.
> I ended up deciding to deserialize to {{InternalRow}} and use
> {{UnsafeProjection}} to convert to unsafe. There are two problems with this:
> first, this is Scala and was difficult to call from Java (it required
> reflection), and second, this causes double projection in the physical plan
> (a copy for unsafe to unsafe) if there is a projection that wasn't fully
> pushed to the data source.
> I think the solution is to have a single interface for readers that expects
> {{InternalRow}}. Then, a projection should be added in the Spark plan to
> convert to unsafe and avoid projection in the plan and in the data source. If
> the data source already produces unsafe rows by deserializing directly, this
> still minimizes the number of copies because the unsafe projection will check
> whether the incoming data is already {{UnsafeRow}}.
> Using {{InternalRow}} would also match the interface on the write side.
--
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
