DB Tsai created SPARK-24371:
-------------------------------
Summary: Added isinSet in DataFrame API for Scala and Java.
Key: SPARK-24371
URL: https://issues.apache.org/jira/browse/SPARK-24371
Project: Spark
Issue Type: New Feature
Components: SQL
Affects Versions: 2.3.0
Reporter: DB Tsai
Assignee: DB Tsai
Fix For: 2.4.0
Implemented *{{isinSet}}* in DataFrame API for both Scala and Java, so users
can do
{code}
val profileDF = Seq(
Some(1), Some(2), Some(3), Some(4),
Some(5), Some(6), Some(7), None
).toDF("profileID")
val validUsers: Set[Any] = Set(6, 7.toShort, 8L, "3")
val result = profileDF.withColumn("isValid", $"profileID".isinSet(validUsers))
result.show(10)
"""
+----------+------+
|profileID|isValid|
+----------+------+
|1|false|
|2|false|
|3|true|
|4|false|
|5|false|
|6|true|
|7|true|
|null|null|
+----------+------+
""".stripMargin
{code}
Two new rules in the logical plan optimizers are added.
# When there is only one element in the *{{Set}}*, the physical plan will be
optimized to *{{EqualTo}}*, so predicate pushdown can be used.
{code}
profileDF.filter( $"profileID".isinSet(Set(6))).explain(true)
"""
|== Physical Plan ==|
|*(1) Project [profileID#0|#0]|
|+- *(1) Filter (isnotnull(profileID#0) && (profileID#0 = 6))|
|+- *(1) FileScan parquet [profileID#0|#0] Batched: true, Format: Parquet,|
|PartitionFilters: [],|
|PushedFilters: [IsNotNull(profileID), EqualTo(profileID,6)],|
|ReadSchema: struct<profileID:int>
""".stripMargin
{code}
# When the *{{Set}}* is empty, and the input is nullable, the logical plan will
be simplified to
{code}
profileDF.filter( $"profileID".isinSet(Set())).explain(true)
"""
|== Optimized Logical Plan ==|
|Filter if (isnull(profileID#0)) null else false|
|+- Relation[profileID#0|#0] parquet
""".stripMargin
{code}
TODO:
# For multiple conditions with numbers less than certain thresholds, we should
still allow predicate pushdown.
# Optimize the `In` using tableswitch or lookupswitch when the numbers of the
categories are low, and they are `Int`, `Long`.
# The default immutable hash trees set is slow for query, and we should do
benchmark for using different set implementation for faster query.
# `filter(if (condition) null else false)` can be optimized to false.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
