scarlin-cloudera commented on code in PR #4378: URL: https://github.com/apache/hive/pull/4378#discussion_r1222060404
########## ql/src/java/org/apache/hadoop/hive/ql/optimizer/calcite/rules/HiveOptimizeInlineArrayTableFunctionRule.java: ########## @@ -0,0 +1,128 @@ +/* + * 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.hadoop.hive.ql.optimizer.calcite.rules; + +import org.apache.calcite.plan.RelOptCluster; +import org.apache.calcite.plan.RelOptRule; +import org.apache.calcite.plan.RelOptRuleCall; +import org.apache.calcite.rel.RelNode; +import org.apache.calcite.rel.type.RelDataType; +import org.apache.calcite.rel.type.RelDataTypeField; +import org.apache.calcite.rex.RexCall; +import org.apache.calcite.rex.RexInputRef; +import org.apache.calcite.rex.RexNode; +import org.apache.calcite.tools.RelBuilderFactory; +import org.apache.hadoop.hive.ql.optimizer.calcite.HiveRelFactories; +import org.apache.hadoop.hive.ql.optimizer.calcite.reloperators.HiveTableFunctionScan; + +import com.google.common.base.Preconditions; +import com.google.common.collect.Lists; + +import java.util.ArrayList; +import java.util.List; + +/** + * This rule optimizes the inline udtf in a HiveTableFunctionScan when it + * has an array of structures. The RelNode for a HiveTableFunctionScan places + * the input references as the first elements in the return type followed by + * the udtf return value which represents the items in the generated table. Take the + * case where the base (input) table has col1, and the inline function is represented by: + * inline(array( struct1(col2, col3), struct2(col2, col3), struct3(col2, col3), etc...)), + * ...and the return value for the table scan node is (col1, col2, col3). In this case, + * the same col1 value is joined with the structures within the inline array for the + * col2 and col3 values. + * + * The optimization is to put the "col1" value within the inline array, resulting in + * in the new structure: + * inline(array(struct1(col1, col2, col3), struct2(col1, col2, col3), ...) + * By doing this, we avoid creating a lateral view join operator and a lateral view forward + * operator at runtime. + */ +public class HiveOptimizeInlineArrayTableFunctionRule extends RelOptRule { + + public static final HiveOptimizeInlineArrayTableFunctionRule INSTANCE = + new HiveOptimizeInlineArrayTableFunctionRule(HiveRelFactories.HIVE_BUILDER); + + public HiveOptimizeInlineArrayTableFunctionRule(RelBuilderFactory relBuilderFactory) { + super(operand(HiveTableFunctionScan.class, any()), relBuilderFactory, null); + } + + @Override + public boolean matches(RelOptRuleCall call) { + final HiveTableFunctionScan tableFunctionScanRel = call.rel(0); + + Preconditions.checkState(tableFunctionScanRel.getCall() instanceof RexCall); + RexCall udtfCall = (RexCall) tableFunctionScanRel.getCall(); + if (!udtfCall.getOperator().getName().equalsIgnoreCase("inline")) { + return false; + } + + Preconditions.checkState(!udtfCall.getOperands().isEmpty()); + RexNode operand = udtfCall.getOperands().get(0); + if (!(operand instanceof RexCall)) { + return false; + } + RexCall firstOperand = (RexCall) operand; + if (!firstOperand.getOperator().getName().equalsIgnoreCase("array")) { + return false; + } + Preconditions.checkState(!firstOperand.getOperands().isEmpty()); + int numStructParams = firstOperand.getOperands().get(0).getType().getFieldCount(); + + if (tableFunctionScanRel.getRowType().getFieldCount() == numStructParams) { + return false; + } + + return true; + } + + public void onMatch(RelOptRuleCall call) { + final HiveTableFunctionScan tfs = call.rel(0); + RelNode inputRel = tfs.getInput(0); + RexCall inlineCall = (RexCall) tfs.getCall(); + RexCall arrayCall = (RexCall) inlineCall.getOperands().get(0); + RelOptCluster cluster = tfs.getCluster(); + + List<RexNode> inputRefs = Lists.transform(inputRel.getRowType().getFieldList(), + input -> new RexInputRef(input.getIndex(), input.getType())); + List<RexNode> newStructExprs = new ArrayList<>(); + + for (RexNode currentStructOperand : arrayCall.getOperands()) { + List<RexNode> allOperands = new ArrayList<>(inputRefs); + RexCall structCall = (RexCall) currentStructOperand; + allOperands.addAll(structCall.getOperands()); + newStructExprs.add(cluster.getRexBuilder().makeCall(structCall.op, allOperands)); + } + + List<RelDataType> returnTypes = new ArrayList<>( + Lists.transform(inputRel.getRowType().getFieldList(), RelDataTypeField::getType)); + RexCall firstStructCall = (RexCall) arrayCall.getOperands().get(0); + returnTypes.addAll(Lists.transform(firstStructCall.getOperands(), RexNode::getType)); + + List<RexNode> newArrayCall = + Lists.newArrayList(cluster.getRexBuilder().makeCall(arrayCall.op, newStructExprs)); Review Comment: Done -- 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]
