[SPARK-20281] Table-valued function range in SQL should use the same number of partitions as spark.range

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Incomplete
• Affects Version/s: 2.2.0
• Component/s: SQL
• Labels: bulk-closed
• Assignee: Unassigned
• Reporter: Jacek Laskowski
• Created: 10/Apr/17 12:11
• Updated: 21/May/19 04:16
• Resolved: 21/May/19 04:16

Description

Note the different number of partitions in range in SQL and as operator.

scala> spark.range(4).explain
== Physical Plan ==
*Range (0, 4, step=1, splits=Some(8)) // <-- note Some(8)

scala> sql("select * from range(4)").explain
== Physical Plan ==
*Range (0, 4, step=1, splits=None) // <-- note None

If I'm not mistaken, the change is to fix builtinFunctions in ResolveTableValuedFunctions (see https://github.com/apache/spark/blob/master/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/analysis/ResolveTableValuedFunctions.scala#L82-L93) to use sparkContext.defaultParallelism as SparkSession.range (see https://github.com/apache/spark/blob/master/sql/core/src/main/scala/org/apache/spark/sql/SparkSession.scala#L517).

Please confirm to work on a fix if and as needed.

Issue Links

• links to: [Github] Pull Request #17670 (maropu)