Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/6726#discussion_r32074216
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/sources/commands.scala ---
@@ -218,24 +218,20 @@ private[sql] case class InsertIntoHadoopFsRelation(
val partitionProj = newProjection(codegenEnabled, partitionOutput,
output)
val dataProj = newProjection(codegenEnabled, dataOutput, output)
- if (needsConversion) {
- val converter =
CatalystTypeConverters.createToScalaConverter(dataSchema)
- while (iterator.hasNext) {
- val row = iterator.next()
- val partitionPart = partitionProj(row)
- val dataPart = dataProj(row)
- val convertedDataPart = converter(dataPart).asInstanceOf[Row]
-
writerContainer.outputWriterForRow(partitionPart).write(convertedDataPart)
- }
+ val partitionSchema = StructType.fromAttributes(partitionOutput)
+ val partConverter =
CatalystTypeConverters.createToScalaConverter(partitionSchema)
--- End diff --
Actually, is there ever a case that we actually need to convert the result
of the partitioning projection? The only case that I can think of would be a
situation where the projection is a no-op that returns the original row and we
apply it to a Scala row that hasn't undergone conversion. If that's the case,
though, why did the old code only convert the projection in the
`!needsConversion` branch?
---
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]
