holdenk commented on a change in pull request #24070: [SPARK-23961][PYTHON] Fix
error when toLocalIterator goes out of scope
URL: https://github.com/apache/spark/pull/24070#discussion_r275460453
##########
File path: core/src/main/scala/org/apache/spark/api/python/PythonRDD.scala
##########
@@ -168,7 +168,32 @@ private[spark] object PythonRDD extends Logging {
}
def toLocalIteratorAndServe[T](rdd: RDD[T]): Array[Any] = {
- serveIterator(rdd.toLocalIterator, s"serve toLocalIterator")
+ val (port, secret) = SocketAuthServer.setupOneConnectionServer(
+ authHelper, "serve toLocalIterator") { s =>
+ val out = new DataOutputStream(s.getOutputStream)
+ val in = new DataInputStream(s.getInputStream)
+ Utils.tryWithSafeFinally {
+
+ // Collects a partition on each iteration
+ val collectPartitionIter = rdd.partitions.indices.iterator.map { i =>
+ rdd.sparkContext.runJob(rdd, (iter: Iterator[Any]) => iter.toArray,
Seq(i)).head
Review comment:
That's a good point @BryanCutler, we could implement the lookahead as a
seperate PR/JIRA and allow it to be turned-on/off. I'd suggest this PR is more
about fixing the behaviour of toLocalIterator memory wise than the out-of-scope
issue (although the out-of-scope issue is maybe more visible in the logs).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to 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]
