Github user srowen commented on a diff in the pull request:
https://github.com/apache/spark/pull/21638#discussion_r214685953
--- Diff:
core/src/main/scala/org/apache/spark/input/PortableDataStream.scala ---
@@ -47,7 +47,7 @@ private[spark] abstract class StreamFileInputFormat[T]
def setMinPartitions(sc: SparkContext, context: JobContext,
minPartitions: Int) {
val defaultMaxSplitBytes =
sc.getConf.get(config.FILES_MAX_PARTITION_BYTES)
val openCostInBytes = sc.getConf.get(config.FILES_OPEN_COST_IN_BYTES)
- val defaultParallelism = sc.defaultParallelism
+ val defaultParallelism = Math.max(sc.defaultParallelism, minPartitions)
--- End diff --
I think it's hard to test, technically, because `setMinPartitions` is only
a hint. In the case of `binaryFiles` we know it will put a hard limit on the
number of partitions, but it isn't true of other implementations. We can still
make a simple test for all of these, it just may be asserting behavior that
could change in the future in Hadoop, though I strongly doubt it would.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
