Github user MLnick commented on a diff in the pull request:
https://github.com/apache/spark/pull/11919#discussion_r60725882
--- Diff: mllib/src/main/scala/org/apache/spark/ml/recommendation/ALS.scala
---
@@ -1306,4 +1306,33 @@ object ALS extends DefaultParamsReadable[ALS] with
Logging {
* satisfies this requirement, we simply use a type alias here.
*/
private[recommendation] type ALSPartitioner =
org.apache.spark.HashPartitioner
+
+ /**
+ * Private function to checkpoint the RDD and clean up its all of its
parents' shuffles eagerly.
+ */
+ private[spark] def checkpointAndCleanParents[T](rdd: RDD[T], blocking:
Boolean = false): Unit = {
+ val sc = rdd.sparkContext
+ // If there is no reference tracking we skip clean up.
+ if (sc.cleaner.isEmpty) {
+ return rdd.checkpoint()
--- End diff --
If we bail out early here, then we don't do a `count`. So in the explicit
case, I believe the checkpoint won't work correctly at
[L677](https://github.com/apache/spark/pull/11919/files#diff-be65dd1d6adc53138156641b610fcadaR677).
Perhaps we can refactor like:
```scala
val cleaner = rdd.sparkContext.cleaner
val deps = rdd.dependencies
rdd.checkpoint()
rdd.count()
cleaner.foreach { c =>
deps.foreach(_ => cleanEagerly(_, c))
}
```
---
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]
