[
https://issues.apache.org/jira/browse/SPARK-22042?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16169148#comment-16169148
]
Tejas Patil commented on SPARK-22042:
-------------------------------------
At the core, the problem is `ReorderJoinPredicates` expects the input tree to
be resolved (esp. when it comes to ensuring partitioning requirements).
Possible fixes:
- Option 1: Have `EnsureRequirements` be done before `ReorderJoinPredicates`.
Since `EnsureRequirements` would have inserted sort and shuffle nodes,
`ReorderJoinPredicates` would have to remove those out. This would definitely
produce sub-optimal plans as `EnsureRequirements` would have done some decision
making (eg. how many shuffle partitions to generate ?) and at that point
`ReorderJoinPredicates` cannot correct that easily.
- Option 2: Do `EnsureRequirements` over the children of input tree for
`ReorderJoinPredicates`. This will be simple change but would look weird given
that we dont call rules from other rules (not to my knowledge).
I will create a PR for option #2 but will be happy to discuss if there are
better solutions.
> ReorderJoinPredicates can break when child's partitioning is not decided
> ------------------------------------------------------------------------
>
> Key: SPARK-22042
> URL: https://issues.apache.org/jira/browse/SPARK-22042
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 2.1.0, 2.2.0
> Reporter: Tejas Patil
> Priority: Minor
>
> When `ReorderJoinPredicates` tries to get the `outputPartitioning` of its
> children, the children may not be properly constructed as the child-subtree
> has to still go through other planner rules.
> In this particular case, the child is `SortMergeJoinExec`. Since the required
> `Exchange` operators are not in place (because `EnsureRequirements` runs
> _after_ `ReorderJoinPredicates`), the join's children would not have
> partitioning defined. This breaks while creation the `PartitioningCollection`
> here :
> https://github.com/apache/spark/blob/94439997d57875838a8283c543f9b44705d3a503/sql/core/src/main/scala/org/apache/spark/sql/execution/joins/SortMergeJoinExec.scala#L69
> Small repro:
> {noformat}
> context.sql("SET spark.sql.autoBroadcastJoinThreshold=0")
> val df = (0 until 50).map(i => (i % 5, i % 13, i.toString)).toDF("i", "j",
> "k")
> df.write.format("parquet").saveAsTable("table1")
> df.write.format("parquet").saveAsTable("table2")
> df.write.format("parquet").bucketBy(8, "j", "k").saveAsTable("bucketed_table")
> sql("""
> SELECT *
> FROM (
> SELECT a.i, a.j, a.k
> FROM bucketed_table a
> JOIN table1 b
> ON a.i = b.i
> ) c
> JOIN table2
> ON c.i = table2.i
> """).explain
> {noformat}
> This fails with :
> {noformat}
> java.lang.IllegalArgumentException: requirement failed:
> PartitioningCollection requires all of its partitionings have the same
> numPartitions.
> at scala.Predef$.require(Predef.scala:224)
> at
> org.apache.spark.sql.catalyst.plans.physical.PartitioningCollection.<init>(partitioning.scala:324)
> at
> org.apache.spark.sql.execution.joins.SortMergeJoinExec.outputPartitioning(SortMergeJoinExec.scala:69)
> at
> org.apache.spark.sql.execution.ProjectExec.outputPartitioning(basicPhysicalOperators.scala:82)
> at
> org.apache.spark.sql.execution.joins.ReorderJoinPredicates$$anonfun$apply$1.applyOrElse(ReorderJoinPredicates.scala:91)
> at
> org.apache.spark.sql.execution.joins.ReorderJoinPredicates$$anonfun$apply$1.applyOrElse(ReorderJoinPredicates.scala:76)
> at
> org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$transformUp$1.apply(TreeNode.scala:289)
> at
> org.apache.spark.sql.catalyst.trees.TreeNode$$anonfun$transformUp$1.apply(TreeNode.scala:289)
> at
> org.apache.spark.sql.catalyst.trees.CurrentOrigin$.withOrigin(TreeNode.scala:70)
> at
> org.apache.spark.sql.catalyst.trees.TreeNode.transformUp(TreeNode.scala:288)
> at
> org.apache.spark.sql.execution.joins.ReorderJoinPredicates.apply(ReorderJoinPredicates.scala:76)
> at
> org.apache.spark.sql.execution.joins.ReorderJoinPredicates.apply(ReorderJoinPredicates.scala:34)
> at
> org.apache.spark.sql.execution.QueryExecution$$anonfun$prepareForExecution$1.apply(QueryExecution.scala:100)
> at
> org.apache.spark.sql.execution.QueryExecution$$anonfun$prepareForExecution$1.apply(QueryExecution.scala:100)
> at
> scala.collection.LinearSeqOptimized$class.foldLeft(LinearSeqOptimized.scala:124)
> at scala.collection.immutable.List.foldLeft(List.scala:84)
> at
> org.apache.spark.sql.execution.QueryExecution.prepareForExecution(QueryExecution.scala:100)
> at
> org.apache.spark.sql.execution.QueryExecution.executedPlan$lzycompute(QueryExecution.scala:90)
> at
> org.apache.spark.sql.execution.QueryExecution.executedPlan(QueryExecution.scala:90)
> at
> org.apache.spark.sql.execution.QueryExecution$$anonfun$simpleString$1.apply(QueryExecution.scala:201)
> at
> org.apache.spark.sql.execution.QueryExecution$$anonfun$simpleString$1.apply(QueryExecution.scala:201)
> at
> org.apache.spark.sql.execution.QueryExecution.stringOrError(QueryExecution.scala:114)
> at
> org.apache.spark.sql.execution.QueryExecution.simpleString(QueryExecution.scala:201)
> at
> org.apache.spark.sql.execution.command.ExplainCommand.run(commands.scala:147)
> at
> org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult$lzycompute(commands.scala:78)
> at
> org.apache.spark.sql.execution.command.ExecutedCommandExec.sideEffectResult(commands.scala:75)
> at
> org.apache.spark.sql.execution.command.ExecutedCommandExec.executeCollect(commands.scala:91)
> at org.apache.spark.sql.Dataset.explain(Dataset.scala:464)
> at org.apache.spark.sql.Dataset.explain(Dataset.scala:477)
> ... 60 elided
> {noformat}
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
