[GitHub] viirya commented on a change in pull request #23873: [SPARK-26975][SQL] Support nested-column pruning over limit/sample/repartition

[GitBox <gi...@apache.org>]

viirya commented on a change in pull request #23873: [SPARK-26975][SQL] Support nested-column pruning over limit/sample/repartition
URL: https://github.com/apache/spark/pull/23873#discussion_r260354889
 
 

 ##########
 File path: sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/NestedColumnPruning.scala
 ##########
 @@ -0,0 +1,193 @@
+/*
+ * 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.optimizer
+
+import scala.collection.mutable
+
+import org.apache.spark.sql.catalyst.expressions._
+import org.apache.spark.sql.catalyst.plans.QueryPlan
+import org.apache.spark.sql.catalyst.plans.logical._
+import org.apache.spark.sql.types.StructType
+
+/**
+ * This aims to handle nested column pruning pattern inside `ColumnPruning` optimizer rule.
+ * If a project or its child references to nested fields, and not all the fields
+ * in a nested attribute are used, we can create a new [[CreateNamedStruct]] containing only
+ * subset of the fields as a new project in the child.
+ */
+object NestedColumnPruning {
+
+  type ExprMap = Map[Expression, Expression]
+  type RequiredFieldMap = Map[Attribute, Set[GetStructField]]
+
+  def unapply(plan: LogicalPlan): Option[(Map[ExprId, Alias], ExprMap)] = plan match {
+    case p @ Project(_, child) if canProjectPushThrough(child) =>
+      getPrunedNestedAttrsAndProjects(getRequiredFieldMap(p, child))
+    case _ => None
+  }
+
+  /**
+   * Add projects to prune unused nested columns.
+   */
+  def prune(
+      plan: LogicalPlan,
+      nestedAttrs: Map[ExprId, Alias],
+      projectsOnNestedAttrs: ExprMap): LogicalPlan = plan match {
+    case Project(projectList, g @ GlobalLimit(_, grandChild: LocalLimit)) =>
+      Project(
+        getNewProjectList(projectList, projectsOnNestedAttrs),
+        g.copy(child = addPrunedProjectToChildren(grandChild, nestedAttrs)))
+    case Project(projectList, child) =>
+      Project(
+        getNewProjectList(projectList, projectsOnNestedAttrs),
+        addPrunedProjectToChildren(child, nestedAttrs))
+    case _ => plan
+  }
+
+  /**
+   * Get a map from a referenced attribute to its set of required fields.
+   */
+  private def getRequiredFieldMap(plans: LogicalPlan*): RequiredFieldMap = {
+    val rootReferenceAttrSet = plans.map(getRootReferences).reduce(_ ++ _)
+    val nestedFieldReferences = plans.map(getNestedFieldReferences).reduce(_ ++ _)
+
+    nestedFieldReferences
+      .filter(!_.references.subsetOf(rootReferenceAttrSet))
+      .groupBy(_.references)
+      .filter(_._1.size == 1)
+      .map(x => (x._1.head, x._2.toSet))
+  }
+
+  /**
+   * Return a replaced project list.
+   */
+  private def getNewProjectList(projectList: Seq[NamedExpression], projectsOnNestedAttrs: ExprMap) =
+    projectList.map(_.transform {
+      case e if projectsOnNestedAttrs.contains(e) => projectsOnNestedAttrs(e)
+    }.asInstanceOf[NamedExpression])
+
+  /**
+   * Return a plan with new childen pruned with `Project`.
+   */
+  private def addPrunedProjectToChildren(plan: LogicalPlan, nestedAttrs: Map[ExprId, Alias]) =
+    plan.withNewChildren(plan.children.map { child =>
+      Project(child.output.map(a => nestedAttrs.getOrElse(a.exprId, a)), child)
+    })
+
+  /**
+   * Returns true for those operators that project can be pushed through.
+   */
+  private def canProjectPushThrough(plan: LogicalPlan) = plan match {
+    case _: GlobalLimit => true
+    case _: LocalLimit => true
+    case _: Repartition => true
+    case _: Sample => true
+    case _ => false
+  }
+
+  /**
+   * Similar to [[QueryPlan.references]], but this only returns all attributes
+   * that are explicitly referenced on the root levels in a [[LogicalPlan]].
+   */
+  private def getRootReferences(plan: LogicalPlan) = {
+    def helper(e: Expression): AttributeSet = e match {
+      case attr: AttributeReference => AttributeSet(attr)
+      case _: GetStructField => AttributeSet.empty
+      case es if es.children.nonEmpty => AttributeSet(es.children.flatMap(helper))
+      case _ => AttributeSet.empty
+    }
+    AttributeSet.fromAttributeSets(plan.expressions.map(helper))
+  }
+
+  /**
+   * Returns all the nested fields that are explicitly referenced as [[Expression]]
+   * in a [[LogicalPlan]]. Currently, we only support having [[GetStructField]] in the chain
+   * of the expressions. If the chain contains GetArrayStructFields, GetMapValue, or
+   * GetArrayItem, the nested field substitution will not be performed.
+   */
+  private def getNestedFieldReferences(plan: LogicalPlan): Seq[GetStructField] = {
+    def helper(e: Expression): Seq[GetStructField] = e match {
+      case f @ GetStructField(child, _, _) if isGetStructFieldOrAttr(child) => Seq(f)
+      case es if es.children.nonEmpty => es.children.flatMap(e => helper(e))
+      case _ => Seq.empty
+    }
+
+    def isGetStructFieldOrAttr(e: Expression): Boolean = e match {
+      case GetStructField(child, _, _) => isGetStructFieldOrAttr(child)
+      case _: AttributeReference => true
+      case _ => false
+    }
+
+    plan.expressions.flatMap(helper)
+  }
+
+  /**
+   * Return `CreateNamedStruct` iff the given expression is a struct type consisting of
+   * some of the given nested fields. Otherwise, this returns None.
+   */
+  private def getReducedCreateNamedStruct(expression: Expression, nestedFields: Set[GetStructField])
+    : Option[CreateNamedStruct] = {
+    expression.dataType match {
+      case st: StructType =>
+        val pairs = st.zipWithIndex.flatMap { case (structField, ordinal) =>
+          val newField = GetStructField(expression, ordinal, Some(structField.name))
+          if (nestedFields.map(_.canonicalized).contains(newField.canonicalized)) {
+            Seq(Literal(structField.name), newField)
+          } else {
+            None
+          }
+        }
+        if (pairs.nonEmpty && pairs.length / 2 < st.length) {
+          Some(CreateNamedStruct(pairs))
+        } else {
+          None
+        }
+      case _ => None
 
 Review comment:
   I think all attributes in `requiredFieldMap` are struct type, isn't?

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
users@infra.apache.org


With regards,
Apache Git Services

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org