zhztheplayer commented on code in PR #7592: URL: https://github.com/apache/incubator-gluten/pull/7592#discussion_r1815934071
########## gluten-arrow/src/main/scala/org/apache/gluten/expression/InterpretedColumnarProjection.scala: ########## @@ -0,0 +1,108 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.gluten.expression + +import org.apache.gluten.vectorized.ArrowColumnarRow + +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions.{Attribute, Expression} +import org.apache.spark.sql.catalyst.expressions.BindReferences.bindReferences +import org.apache.spark.sql.catalyst.expressions.aggregate.NoOp +import org.apache.spark.sql.internal.SQLConf +import org.apache.spark.sql.types.{BinaryType, DataType, StringType} +import org.apache.spark.unsafe.types.UTF8String + +/** + * A [[ArrowProjection]] that is calculated by calling `eval` on each of the specified expressions. + * + * @param expressions + * a sequence of expressions that determine the value of each column of the output row. + */ +class InterpretedColumnarProjection(expressions: Seq[Expression]) extends ArrowProjection { Review Comment: `InterpretedColumnarProjection` vs `InterpretedArrowProjection` ? ########## shims/spark32/src/main/scala/org/apache/spark/sql/catalyst/expressions/ExpressionsEvaluatory.scala: ########## @@ -0,0 +1,50 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.spark.sql.catalyst.expressions + +import org.apache.spark.sql.internal.SQLConf + +// A helper class to evaluate expressions. +trait ExpressionsEvaluator { Review Comment: The file is named `ExpressionsEvaluatory`, i guess this is a typo? ########## gluten-arrow/src/main/scala/org/apache/gluten/expression/ArrowProjection.scala: ########## @@ -0,0 +1,67 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.gluten.expression + +import org.apache.gluten.vectorized.ArrowColumnarRow + +import org.apache.spark.sql.catalyst.InternalRow +import org.apache.spark.sql.catalyst.expressions.{Attribute, BoundReference, Expression, ExpressionsEvaluator} +import org.apache.spark.sql.catalyst.expressions.BindReferences.bindReferences +import org.apache.spark.sql.types.{DataType, StructType} + +// Not thread safe. +abstract class ArrowProjection extends (InternalRow => ArrowColumnarRow) with ExpressionsEvaluator { + def currentValue: ArrowColumnarRow + + /** Uses the given row to store the output of the projection. */ + def target(row: ArrowColumnarRow): ArrowProjection +} + +/** The factory object for `ArrowProjection`. */ +object ArrowProjection { + + /** + * Returns an ArrowProjection for given StructType. + * + * CAUTION: the returned projection object is *not* thread-safe. + */ + def create(schema: StructType): ArrowProjection = create(schema.fields.map(_.dataType)) + + /** + * Returns an ArrowProjection for given Array of DataTypes. + * + * CAUTION: the returned projection object is *not* thread-safe. + */ + def create(fields: Array[DataType]): ArrowProjection = { + create(fields.zipWithIndex.map(x => BoundReference(x._2, x._1, nullable = true))) + } + + /** Returns an ArrowProjection for given sequence of bound Expressions. */ + def create(exprs: Seq[Expression]): ArrowProjection = { + InterpretedColumnarProjection.createProjection(exprs) + } + + def create(expr: Expression): ArrowProjection = create(Seq(expr)) + + /** + * Returns an ArrowProjection for given sequence of Expressions, which will be bound to + * `inputSchema`. + */ + def create(exprs: Seq[Expression], inputSchema: Seq[Attribute]): ArrowProjection = { + create(bindReferences(exprs, inputSchema)) + } Review Comment: Didn't see all the APIs are used so far. Shall we have chance to use them in future? ########## backends-velox/src/main/scala/org/apache/gluten/execution/ColumnarPartialProjectExec.scala: ########## @@ -183,9 +181,8 @@ case class ColumnarPartialProjectExec(original: ProjectExec, child: SparkPlan)( } else { val start = System.currentTimeMillis() val childData = ColumnarBatches.select(batch, projectIndexInChild.toArray) Review Comment: Is there possibility that `projectIndexInChild` includes duplicated indexes? If yes can we have a test case for that? Thanks. ########## backends-velox/src/main/scala/org/apache/gluten/execution/ColumnarPartialProjectExec.scala: ########## @@ -214,60 +211,12 @@ case class ColumnarPartialProjectExec(original: ProjectExec, child: SparkPlan)( } } - // scalastyle:off line.size.limit - // String type cannot use MutableProjection - // Otherwise will throw java.lang.UnsupportedOperationException: Datatype not supported StringType - // at org.apache.spark.sql.execution.vectorized.MutableColumnarRow.update(MutableColumnarRow.java:224) - // at org.apache.spark.sql.catalyst.expressions.GeneratedClass$SpecificMutableProjection.apply(Unknown Source) - // scalastyle:on line.size.limit - private def canUseMutableProjection(): Boolean = { - replacedAliasUdf.forall( - r => - r.dataType match { - case StringType | BinaryType => false - case _ => true - }) - } - - /** - * add c2r and r2c for unsupported expression child data c2r get Iterator[InternalRow], then call - * Spark project, then r2c - */ - private def getProjectedBatch( - childData: ColumnarBatch, - c2r: SQLMetric, - r2c: SQLMetric): Iterator[ColumnarBatch] = { - // select part of child output and child data - val proj = UnsafeProjection.create(replacedAliasUdf, projectAttributes.toSeq) - val numOutputRows = new SQLMetric("numOutputRows") - val numInputBatches = new SQLMetric("numInputBatches") - val rows = VeloxColumnarToRowExec - .toRowIterator( - Iterator.single[ColumnarBatch](childData), - projectAttributes.toSeq, - numOutputRows, - numInputBatches, - c2r) - .map(proj) - - val schema = - SparkShimLoader.getSparkShims.structFromAttributes(replacedAliasUdf.map(_.toAttribute)) - RowToVeloxColumnarExec.toColumnarBatchIterator( - rows, - schema, - numOutputRows, - numInputBatches, - r2c, - childData.numRows()) - // TODO: should check the size <= 1, but now it has bug, will change iterator to empty - } - private def getProjectedBatchArrow( childData: ColumnarBatch, c2a: SQLMetric, a2c: SQLMetric): Iterator[ColumnarBatch] = { // select part of child output and child data - val proj = MutableProjection.create(replacedAliasUdf, projectAttributes.toSeq) + val proj = ArrowProjection.create(replacedAliasUdf, projectAttributes.toSeq) val numRows = childData.numRows() val start = System.currentTimeMillis() val arrowBatch = if (childData.numCols() == 0 || ColumnarBatches.isHeavyBatch(childData)) { Review Comment: Guessing dong a `checkOffloaded` check is enough, as the input of this operator is guaranteed to be in Velox format -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
