[SPARK-20451] Filter out nested mapType datatypes from sort order in randomSplit #17751
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
gatorsmile
reviewed
Apr 24, 2017
| // ordering deterministic. Note that MapTypes cannot be sorted and are explicitly pruned out | ||
| // from the sort order. | ||
| val sortOrder = logicalPlan.output | ||
| .filterNot(_.dataType.existsRecursively(dt => dt.isInstanceOf[MapType])) |
There was a problem hiding this comment.
How about calling RowOrdering.isOrderable?
There was a problem hiding this comment.
UDT with underlying MapType is also not sortable.
gatorsmile
reviewed
Apr 24, 2017
| val plan = if (sortOrder.nonEmpty) { | ||
| Sort(sortOrder, global = false, logicalPlan) | ||
| } else { | ||
| // SPARK-12662: If sort order is empty, we materialize the dataset to guarantee determinism |
There was a problem hiding this comment.
We actually discussed materialization in https://issues.apache.org/jira/browse/SPARK-12662 so that ticket should provide direct context.
|
Test build #76118 has finished for PR 17751 at commit
|
|
LGTM pending Jenkins |
|
Test build #76123 has finished for PR 17751 at commit
|
asfgit
pushed a commit
that referenced
this pull request
Apr 25, 2017
…randomSplit ## What changes were proposed in this pull request? In `randomSplit`, It is possible that the underlying dataset doesn't guarantee the ordering of rows in its constituent partitions each time a split is materialized which could result in overlapping splits. To prevent this, as part of SPARK-12662, we explicitly sort each input partition to make the ordering deterministic. Given that `MapTypes` cannot be sorted this patch explicitly prunes them out from the sort order. Additionally, if the resulting sort order is empty, this patch then materializes the dataset to guarantee determinism. ## How was this patch tested? Extended `randomSplit on reordered partitions` in `DataFrameStatSuite` to also test for dataframes with mapTypes nested mapTypes. Author: Sameer Agarwal <[email protected]> Closes #17751 from sameeragarwal/randomsplit2. (cherry picked from commit 31345fd) Signed-off-by: Wenchen Fan <[email protected]>
asfgit
pushed a commit
that referenced
this pull request
Apr 25, 2017
…randomSplit ## What changes were proposed in this pull request? In `randomSplit`, It is possible that the underlying dataset doesn't guarantee the ordering of rows in its constituent partitions each time a split is materialized which could result in overlapping splits. To prevent this, as part of SPARK-12662, we explicitly sort each input partition to make the ordering deterministic. Given that `MapTypes` cannot be sorted this patch explicitly prunes them out from the sort order. Additionally, if the resulting sort order is empty, this patch then materializes the dataset to guarantee determinism. ## How was this patch tested? Extended `randomSplit on reordered partitions` in `DataFrameStatSuite` to also test for dataframes with mapTypes nested mapTypes. Author: Sameer Agarwal <[email protected]> Closes #17751 from sameeragarwal/randomsplit2. (cherry picked from commit 31345fd) Signed-off-by: Wenchen Fan <[email protected]>
|
thanks, merging to master/2.2/2.1/2.0! |
asfgit
pushed a commit
that referenced
this pull request
Apr 25, 2017
…randomSplit ## What changes were proposed in this pull request? In `randomSplit`, It is possible that the underlying dataset doesn't guarantee the ordering of rows in its constituent partitions each time a split is materialized which could result in overlapping splits. To prevent this, as part of SPARK-12662, we explicitly sort each input partition to make the ordering deterministic. Given that `MapTypes` cannot be sorted this patch explicitly prunes them out from the sort order. Additionally, if the resulting sort order is empty, this patch then materializes the dataset to guarantee determinism. ## How was this patch tested? Extended `randomSplit on reordered partitions` in `DataFrameStatSuite` to also test for dataframes with mapTypes nested mapTypes. Author: Sameer Agarwal <[email protected]> Closes #17751 from sameeragarwal/randomsplit2. (cherry picked from commit 31345fd) Signed-off-by: Wenchen Fan <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changes were proposed in this pull request?
In
randomSplit, It is possible that the underlying dataset doesn't guarantee the ordering of rows in its constituent partitions each time a split is materialized which could result in overlappingsplits.
To prevent this, as part of SPARK-12662, we explicitly sort each input partition to make the ordering deterministic. Given that
MapTypescannot be sorted this patch explicitly prunes them out from the sort order. Additionally, if the resulting sort order is empty, this patch then materializes the dataset to guarantee determinism.How was this patch tested?
Extended
randomSplit on reordered partitionsinDataFrameStatSuiteto also test for dataframes with mapTypes nested mapTypes.