Skip to content

fix: use inputRDD to get outputPartitions in CometScanExec #1162

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
andygrove merged 1 commit into from
Dec 11, 2024
Merged

fix: use inputRDD to get outputPartitions in CometScanExec #1162

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion spark/src/main/scala/org/apache/spark/sql/comet/CometScanExec.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,7 +132,7 @@ case class CometScanExec(
lazy val bucketedScan: Boolean = wrapped.bucketedScan

override lazy val (outputPartitioning, outputOrdering): (Partitioning, Seq[SortOrder]) =
(wrapped.outputPartitioning, wrapped.outputOrdering)
(UnknownPartitioning(wrapped.inputRDD.getNumPartitions), wrapped.outputOrdering)
Copy link
Member

@viirya viirya Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What the output partitioning is when it fails the test? I assume that the outputPartitioning should have correct partition number but not?

Copy link
Member

@andygrove andygrove Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We were seeing zero partitions. Here is some debug output from testing:

firstNonBroadcastPlan = Some((CometScan parquet [_1#6,_2#7] Batched: true, DataFilters: [isnotnull(_1#6)], Format: CometParquet, Location: InMemoryFileIndex(1 paths)[file:/tmp/spark-a21845bf-72dc-49d1-94f0-f785bb6f2f18], PartitionFilters: [], PushedFilters: [IsNotNull(_1)], ReadSchema: struct<_1:int,_2:int>
,0))

firstNonBroadcastPlanNumPartitions = Some(0)

Copy link
Contributor Author

@parthchandra parthchandra Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Correct. wrapped.outputPartitioning was zero

Copy link
Member

@viirya viirya Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It is somehow weird. It means the outputPartitioning info is not correct in original ScanExec and not matched the output RDD's partition number. 🤔

Copy link
Member

@viirya viirya Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Btw, I think we should keep original partition instead a hard-coded UnknownPartitioning?

Copy link
Contributor Author

@parthchandra parthchandra Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I did this based on what we do in CometNativeScanExec as well as the original output partition. The original wrapped.outputPartitioning for CometScanExec inherits from FileSourceScanLike and this returns a hardcoded UnknownPartitioning(0) for non-bucketed scan.

Copy link
Member

@viirya viirya Dec 11, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Okay


@transient
private lazy val pushedDownFilters = getPushedDownFilters(relation, dataFilters)
Expand Down
Loading