Github user yhuai commented on a diff in the pull request:
https://github.com/apache/spark/pull/4792#discussion_r25533656
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTableOperations.scala
---
@@ -143,19 +146,47 @@ private[sql] case class ParquetTableScan(
relation.partitioningAttributes
.map(a => Cast(Literal(partValues(a.name)),
a.dataType).eval(EmptyRow))
- new Iterator[Row] {
- def hasNext = iter.hasNext
- def next() = {
- val row = iter.next()._2.asInstanceOf[SpecificMutableRow]
-
- // Parquet will leave partitioning columns empty, so we fill
them in here.
- var i = 0
- while (i < requestedPartitionOrdinals.size) {
- row(requestedPartitionOrdinals(i)._2) =
- partitionRowValues(requestedPartitionOrdinals(i)._1)
- i += 1
+ if (primitiveRow) {
+ new Iterator[Row] {
+ def hasNext = iter.hasNext
+ def next() = {
+ // We are using CatalystPrimitiveRowConverter and it returns
a SpecificMutableRow.
+ val row = iter.next()._2.asInstanceOf[SpecificMutableRow]
+
+ // Parquet will leave partitioning columns empty, so we fill
them in here.
+ var i = 0
+ while (i < requestedPartitionOrdinals.size) {
+ row(requestedPartitionOrdinals(i)._2) =
+ partitionRowValues(requestedPartitionOrdinals(i)._1)
+ i += 1
+ }
+ row
+ }
+ }
+ } else {
+ // Create a mutable row since we need to fill in values from
partition columns.
+ val mutableRow = new GenericMutableRow(output.size)
--- End diff --
Seems it will be safer if we get `output.size` out side of
`mapPartitionsWithInputSplit`.
---
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]
