Skip to content

[SPARK-39784][SQL] Put Literal values on the right side of the data source filter after translating Catalyst Expression to data source filter #37197

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

Closed
huaxingao wants to merge 8 commits into from
Closed

[SPARK-39784][SQL] Put Literal values on the right side of the data source filter after translating Catalyst Expression to data source filter #37197

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
f742b20
Literal values should be on the right side of the data source filter
huaxingao Jul 14, 2022
fa4d57a
normalize V2 attr cmp lit in V2ExpressionBuilder
huaxingao Jul 21, 2022
a708ad6
fix test failure
huaxingao Jul 21, 2022
47e30c4
remove extra space
huaxingao Jul 22, 2022
1e94c53
address comments
huaxingao Jul 22, 2022
324dd87
fix test failure
huaxingao Jul 22, 2022
99b6c45
4 space indentation for breaking a line
huaxingao Jul 22, 2022
a4e8170
address comments
huaxingao Jul 22, 2022
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
24 changes: 22 additions & 2 deletions ...e/src/main/scala/org/apache/spark/sql/execution/datasources/v2/DataSourceV2Strategy.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,7 @@ import org.apache.spark.sql.catalyst.plans.logical._
import org.apache.spark.sql.catalyst.util.{toPrettySQL, ResolveDefaultColumns, V2ExpressionBuilder}
import org.apache.spark.sql.connector.catalog.{Identifier, StagingTableCatalog, SupportsDelete, SupportsNamespaces, SupportsPartitionManagement, SupportsWrite, Table, TableCapability, TableCatalog, TruncatableTable}
import org.apache.spark.sql.connector.catalog.index.SupportsIndex
import org.apache.spark.sql.connector.expressions.FieldReference
import org.apache.spark.sql.connector.expressions.{FieldReference, LiteralValue}
import org.apache.spark.sql.connector.expressions.filter.{And => V2And, Not => V2Not, Or => V2Or, Predicate}
import org.apache.spark.sql.connector.read.LocalScan
import org.apache.spark.sql.connector.read.streaming.{ContinuousStream, MicroBatchStream}
Expand Down Expand Up @@ -502,11 +502,31 @@ private[sql] object DataSourceV2Strategy {

private def translateLeafNodeFilterV2(predicate: Expression): Option[Predicate] = {
predicate match {
case PushablePredicate(expr) => Some(expr)
case PushablePredicate(expr) =>
if (expr.children().length == 2) {
expr.children()(0) match {
Copy link

@zinking zinking Jul 21, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

could this be written into something like expr.children() match { case Literal :: xs =>

case LiteralValue(_, _) =>
Some(new Predicate(flipComparisonFilterName(expr.name()),
Copy link
Member

@sunchao sunchao Jul 21, 2022

Choose a reason for hiding this comment

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

I wonder if we should be more strict and only flip the filter iff the operator is in the set { >, <, >=, <=}

Array(expr.children()(1), expr.children()(0))))
case _ => Some(expr)
}
} else {
Some(expr)
}
case _ => None
}
}

private def flipComparisonFilterName(filterName: String): String = {
filterName match {
case ">" => "<"
case "<" => ">"
case ">=" => "<="
case "<=" => ">="
case _ => filterName
Copy link

@zinking zinking Jul 21, 2022

Choose a reason for hiding this comment

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

is = handled as well ?

Copy link
Member

@dongjoon-hyun dongjoon-hyun Jul 21, 2022 •
edited
Loading

Choose a reason for hiding this comment

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

We don't need to flip the operator, =, do we? @zinking
The operands are flipped already via Array(expr.children()(1), expr.children()(0)).

}
}

/**
* Tries to translate a Catalyst [[Expression]] into data source [[Filter]].
*
Expand Down
67 changes: 65 additions & 2 deletions .../test/scala/org/apache/spark/sql/execution/datasources/v2/DataSourceV2StrategySuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,14 +18,77 @@
package org.apache.spark.sql.execution.datasources.v2

import org.apache.spark.sql.catalyst.dsl.expressions._
import org.apache.spark.sql.catalyst.expressions.Expression
import org.apache.spark.sql.catalyst.expressions._
import org.apache.spark.sql.catalyst.plans.PlanTest
import org.apache.spark.sql.connector.expressions.{FieldReference, LiteralValue}
import org.apache.spark.sql.connector.expressions.filter.Predicate
import org.apache.spark.sql.test.SharedSparkSession
import org.apache.spark.sql.types.BooleanType
import org.apache.spark.sql.types.{BooleanType, IntegerType, StringType, StructField, StructType}

class DataSourceV2StrategySuite extends PlanTest with SharedSparkSession {
val attrInts = Seq(
$"cint".int,
$"c.int".int,
GetStructField($"a".struct(StructType(
StructField("cstr", StringType, nullable = true) ::
StructField("cint", IntegerType, nullable = true) :: Nil)), 1, None),
GetStructField($"a".struct(StructType(
StructField("c.int", IntegerType, nullable = true) ::
StructField("cstr", StringType, nullable = true) :: Nil)), 0, None),
GetStructField($"a.b".struct(StructType(
StructField("cstr1", StringType, nullable = true) ::
StructField("cstr2", StringType, nullable = true) ::
StructField("cint", IntegerType, nullable = true) :: Nil)), 2, None),
GetStructField($"a.b".struct(StructType(
StructField("c.int", IntegerType, nullable = true) :: Nil)), 0, None),
GetStructField(GetStructField($"a".struct(StructType(
StructField("cstr1", StringType, nullable = true) ::
StructField("b", StructType(StructField("cint", IntegerType, nullable = true) ::
StructField("cstr2", StringType, nullable = true) :: Nil)) :: Nil)), 1, None), 0, None)
).zip(Seq(
"cint",
"`c.int`", // single level field that contains `dot` in name
"a.cint", // two level nested field
"a.`c.int`", // two level nested field, and nested level contains `dot`
"`a.b`.cint", // two level nested field, and top level contains `dot`
"`a.b`.`c.int`", // two level nested field, and both levels contain `dot`
"a.b.cint" // three level nested field
))

test("translate simple expression") { attrInts
.foreach { case (attrInt, intColName) =>
testTranslateFilter(EqualTo(attrInt, 1),
Some(new Predicate("=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(EqualTo(1, attrInt),
Some(new Predicate("=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))

testTranslateFilter(EqualNullSafe(attrInt, 1),
Some(new Predicate("<=>", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(EqualNullSafe(1, attrInt),
Some(new Predicate("<=>", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))

testTranslateFilter(GreaterThan(attrInt, 1),
Some(new Predicate(">", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(GreaterThan(1, attrInt),
Some(new Predicate("<", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))

testTranslateFilter(LessThan(attrInt, 1),
Some(new Predicate("<", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(LessThan(1, attrInt),
Some(new Predicate(">", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))

testTranslateFilter(GreaterThanOrEqual(attrInt, 1),
Some(new Predicate(">=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(GreaterThanOrEqual(1, attrInt),
Some(new Predicate("<=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))

testTranslateFilter(LessThanOrEqual(attrInt, 1),
Some(new Predicate("<=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
testTranslateFilter(LessThanOrEqual(1, attrInt),
Some(new Predicate(">=", Array(FieldReference(intColName), LiteralValue(1, IntegerType)))))
}
}

test("SPARK-36644: Push down boolean column filter") {
testTranslateFilter($"col".boolean,
Some(new Predicate("=", Array(FieldReference("col"), LiteralValue(true, BooleanType)))))
Expand Down