IvanVergiliev commented on a change in pull request #23766: [SPARK-26859][SQL]
Fix field writer index bug in non-vectorized ORC deserializer
URL: https://github.com/apache/spark/pull/23766#discussion_r258092063
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/orc/OrcDeserializer.scala
##########
@@ -37,28 +37,34 @@ class OrcDeserializer(
private val resultRow = new
SpecificInternalRow(requiredSchema.map(_.dataType))
+ // `fieldWriters(index)` is
+ // - null if the respective source column is missing, since the output value
+ // is always null in this case
+ // - a function that updates target column `index` otherwise.
private val fieldWriters: Array[WritableComparable[_] => Unit] = {
requiredSchema.zipWithIndex
- // The value of missing columns are always null, do not need writers.
- .filterNot { case (_, index) => requestedColIds(index) == -1 }
.map { case (f, index) =>
- val writer = newWriter(f.dataType, new RowUpdater(resultRow))
- (value: WritableComparable[_]) => writer(index, value)
+ if (requestedColIds(index) == -1) {
+ null
+ } else {
+ val writer = newWriter(f.dataType, new RowUpdater(resultRow))
+ (value: WritableComparable[_]) => writer(index, value)
+ }
}.toArray
}
- private val validColIds = requestedColIds.filterNot(_ == -1)
-
def deserialize(orcStruct: OrcStruct): InternalRow = {
- var i = 0
- while (i < validColIds.length) {
- val value = orcStruct.getFieldValue(validColIds(i))
- if (value == null) {
- resultRow.setNullAt(i)
- } else {
- fieldWriters(i)(value)
+ var targetColumnIndex = 0
+ while (targetColumnIndex < fieldWriters.length) {
+ if (fieldWriters(targetColumnIndex) != null) {
+ var value = orcStruct.getFieldValue(requestedColIds(targetColumnIndex))
Review comment:
No idea how that happened.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
