Skip to content

[VL] Support partial fallback generic unmapped catalyst expressions #12486

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

Open
taiyang-li wants to merge 2 commits into
base: main
Choose a base branch
from
+47 −0
Open

[VL] Support partial fallback generic unmapped catalyst expressions #12486

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c4fedc2
[MINOR][VL] Fallback generic unmapped catalyst expressions via partia…
taiyang-li Jul 9, 2026
6b98127
[MINOR][VL] Shorten partial project comment
taiyang-li Jul 9, 2026
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
19 changes: 19 additions & 0 deletions backends-velox/src/main/scala/org/apache/gluten/execution/ColumnarPartialProjectExec.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -297,6 +297,22 @@ object ColumnarPartialProjectExec {
HiveUDFTransformer.isHiveUDF(h) && !VeloxHiveUDFTransformer.isSupportedHiveUDF(h)
}

/**
* Returns true for generic catalyst expressions that Velox cannot offload and should be pulled
* into ColumnarPartialProject, excluding ScalaUDF/HiveUDF and non-row-evaluable wrappers.
*/
private def isUnmappedGlutenExpr(expr: Expression): Boolean = expr match {
case _: ScalaUDF => false
case _ if HiveUDFTransformer.isHiveUDF(expr) => false
case _: Literal | _: AttributeReference | _: BoundReference => false
case _: NamedExpression => false
case _ =>
expr.deterministic &&
!expr.isInstanceOf[Unevaluable] &&
!expr.isInstanceOf[LambdaFunction] &&
!ExpressionMappings.expressionsMap.contains(expr.getClass)
}

private def isBlacklistExpression(e: Expression): Boolean = {
ExpressionMappings.blacklistExpressionMap.contains(e.getClass)
}
Expand All @@ -307,6 +323,7 @@ object ColumnarPartialProjectExec {
case _: ScalaUDF => true
case h if containsUnsupportedHiveUDF(h) => true
case e if isBlacklistExpression(e) => true
case e if isUnmappedGlutenExpr(e) => true
case p => p.children.exists(c => containsUDFOrBlacklistExpression(c))
}
}
Expand Down Expand Up @@ -339,6 +356,8 @@ object ColumnarPartialProjectExec {
replaceByAlias(h, replacedAlias)
case e if isBlacklistExpression(e) =>
replaceByAlias(e, replacedAlias)
case e if isUnmappedGlutenExpr(e) =>
replaceByAlias(e, replacedAlias)
case au @ Alias(_: ScalaUDF, _) =>
val replaceIndex = replacedAlias.indexWhere(r => r.exprId == au.exprId)
if (replaceIndex == -1) {
Expand Down
28 changes: 28 additions & 0 deletions backends-velox/src/test/scala/org/apache/gluten/expression/UDFPartialProjectSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,16 +20,31 @@ import org.apache.gluten.config.GlutenConfig
import org.apache.gluten.execution.{ColumnarPartialProjectExec, WholeStageTransformerSuite}

import org.apache.spark.SparkConf
import org.apache.spark.sql.catalyst.FunctionIdentifier
import org.apache.spark.sql.catalyst.expressions.{Expression, ExpressionInfo, UnaryExpression}
import org.apache.spark.sql.catalyst.expressions.codegen.CodegenFallback
import org.apache.spark.sql.catalyst.optimizer.{ConstantFolding, NullPropagation}
import org.apache.spark.sql.execution.ProjectExec
import org.apache.spark.sql.functions.udf
import org.apache.spark.sql.types.{BooleanType, DataType}

import java.io.File

case class MyStruct(a: Long, b: Array[Long])

case class MyStructWithNullValue(a: Option[Long], b: Array[Long])

case class DummyUnmappedExpr(child: Expression) extends UnaryExpression with CodegenFallback {
override def dataType: DataType = BooleanType

override protected def nullSafeEval(input: Any): Any = {
input.asInstanceOf[Long] % 2 == 0
}

override protected def withNewChildInternal(newChild: Expression): DummyUnmappedExpr =
copy(child = newChild)
}

class UDFPartialProjectSuite extends WholeStageTransformerSuite {
disableFallbackCheck
override protected val resourcePath: String = "/tpch-data-parquet"
Expand Down Expand Up @@ -64,9 +79,22 @@ class UDFPartialProjectSuite extends WholeStageTransformerSuite {
spark.udf.register("no_argument", noArgument)
val concat = udf((x: String) => x + "_concat")
spark.udf.register("concat_concat", concat)
spark.sessionState.functionRegistry.registerFunction(
FunctionIdentifier("dummy_unmapped"),
new ExpressionInfo(classOf[DummyUnmappedExpr].getName, "dummy_unmapped"),
(children: Seq[Expression]) => DummyUnmappedExpr(children.head)
)

}

test("fallback generic unmapped expression via partial project") {
runQueryAndCompare(
"SELECT sum(cast(dummy_unmapped(cast(l_orderkey as long)) as int)" +
" + hash(l_partkey)) from lineitem") {
checkGlutenPlan[ColumnarPartialProjectExec]
}
}

ignore("test plus_one") {
runQueryAndCompare("SELECT sum(plus_one(cast(l_orderkey as long))) from lineitem") {
checkGlutenPlan[ColumnarPartialProjectExec]
Expand Down
Loading