ayushi-agarwal commented on code in PR #8484: URL: https://github.com/apache/incubator-gluten/pull/8484#discussion_r2044261986
########## tools/qualification-tool/src/main/java/org/apache/gluten/qt/support/NodeSupportVisitor.java: ########## @@ -0,0 +1,225 @@ +/* + * 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.qt.support; + +import org.apache.gluten.qt.graph.SparkPlanGraphNodeInternal; + +import com.google.common.collect.ImmutableList; +import com.google.common.collect.ImmutableSet; + +import java.math.BigDecimal; +import java.math.BigInteger; +import java.util.Arrays; +import java.util.List; +import java.util.Optional; +import java.util.regex.Matcher; +import java.util.regex.Pattern; +import java.util.stream.Collectors; + +import static org.apache.gluten.qt.support.NotSupportedCategory.NODE_NOT_SUPPORTED; + +/** + * {@link GraphVisitor} implementation that inspects each node in the {@code ExecutionDescription} + * to determine whether it is fully supported by Gluten. Specifically, this visitor classifies each + * node as either {@link Supported} or {@link NotSupported} based on: + * + * <ul> + * <li>The operator name (e.g., {@code BroadcastHashJoin}, {@code Scan}) + * <li>Data formats (e.g., Parquet, ORC) and their unsupported types + * <li>Resource constraints (e.g., broadcast data size) + * </ul> + * + * The aggregated results are then stored in {@code resultSupportMap} and incorporated into a new + * {@link ExecutionDescription} instance. + */ +public class NodeSupportVisitor extends GraphVisitor { + private static final Pattern MEM_PATTERN = + Pattern.compile("^\\d+\\.?\\d*\\s*(EiB|PiB|TiB|GiB|MiB|KiB|B)\\b"); + private static final Pattern SIZE_PATTERN = Pattern.compile("([\\d.]+)\\s*([a-zA-Z]+)"); + private static final Pattern FORMAT_PATTERN = Pattern.compile("Format: ([^,]+)"); + private static final Pattern SCHEMA_PATTERN = Pattern.compile("ReadSchema: struct<(.*)"); + private static final ImmutableSet<String> FULLY_SUPPORTED_OPERATORS = + ImmutableSet.of( + "AdaptiveSparkPlan", + "TakeOrderedAndProject", + "Project", + "Filter", + "Window", + "Coalesce", + "CartesianProduct", + "HashAggregate", + "SortAggregate", + "Sort", + "ObjectHashAggregate", + "Expand", + "Union", + "ShuffleExchange", + "ShuffledHashJoin", + "SortMergeJoin", + "BroadcastExchange", + "GlobalLimit", + "LocalLimit", + "Generate", + "EvalPython", + "AQEShuffleRead", + "ReusedExchange", + "Subquery", + "SubqueryBroadcast", + "ShuffleQueryStage", + "Exchange", + "BroadcastQueryStage", + "ColumnarToRow", + "RowToColumnar", + "ReusedSubquery", + "ColumnarExchange", + "ColumnarBroadcastExchange", + "VeloxColumnarToRowExec", + "RowToVeloxColumnar", + "ColumnarSubqueryBroadcast", + "ColumnarUnion", + "InMemoryTableScan", + "BroadcastNestedLoopJoin", + "CollectLimit"); + + public NodeSupportVisitor(ExecutionDescription executionDescription) { + super(executionDescription); + } + + @Override + protected void visitor(long nodeId) { + resultSupportMap.put(nodeId, isSupported(nodeId)); + } + + private String schemaConditions(String schema, List<String> conditions) { + return conditions.stream() + .map(cond -> ":" + cond) + .filter(schema::contains) + .collect(Collectors.joining(" and ")) + .replace(":", ""); + } + + private Optional<String> extractSize(String input) { + Matcher matcher = MEM_PATTERN.matcher(input); + + if (matcher.find()) { + return Optional.of(matcher.group()); + } else { + return Optional.empty(); + } + } + + private Optional<BigInteger> stringToBytes(String sizeStr) { + Matcher matcher = SIZE_PATTERN.matcher(sizeStr); + if (matcher.matches()) { + BigDecimal size = new BigDecimal(matcher.group(1)); + String unit = matcher.group(2); + switch (unit) { + case "EiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 60)).toBigInteger()); + case "PiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 50)).toBigInteger()); + case "TiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 40)).toBigInteger()); + case "GiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 30)).toBigInteger()); + case "MiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 20)).toBigInteger()); + case "KiB": + return Optional.of(size.multiply(BigDecimal.valueOf(1L << 10)).toBigInteger()); + case "B": + return Optional.of(size.toBigInteger()); + default: + return Optional.empty(); + } + } + return Optional.empty(); + } + + private static final ImmutableList<String> PARQUET_UNSUPPORTED_TYPES = + ImmutableList.of("struct", "map", "list", "tinyint"); + + private static final ImmutableList<String> ORC_UNSUPPORTED_TYPES = + ImmutableList.of("struct", "map", "list", "tinyint", "timestamp", "varchar", "char", "union"); + + private boolean checkTypeSupport(String dataType, List<String> unsupportedList) { + return unsupportedList.stream().map(s -> ":" + s).noneMatch(dataType::contains); + } + + private GlutenSupport isSupported(long nodeId) { + SparkPlanGraphNodeInternal node = executionDescription.getNode(nodeId); + String name = node.getName(); + String anonymizedName = node.getAnonymizedName(); + String desc = node.getDesc(); + + if (FULLY_SUPPORTED_OPERATORS.contains(name)) { + return new Supported(); + } else if (name.equals("BroadcastHashJoin")) { + boolean bhjSupported = + executionDescription.getChildren(nodeId).stream() + .map(executionDescription::getNode) + .filter(childNode -> childNode.getName().equals("BroadcastExchange")) + .flatMap( + childNode -> + childNode.getMetrics().stream() + .filter(metric -> metric.getName().contains("data size"))) + .flatMap( + metric -> + Optional.ofNullable( + executionDescription.getMetrics(metric.getAccumulatorId())) + .stream()) + .flatMap( + metricStr -> + Arrays.stream(metricStr.split("\n", -1)) + .reduce((first, second) -> second) + .stream()) + .flatMap(metricStr -> extractSize(metricStr).stream()) + .flatMap(metricStr -> stringToBytes(metricStr).stream()) + .anyMatch(size -> size.compareTo(BigInteger.valueOf(100L << 20)) <= 0); Review Comment: Could you help understand why there is special case for BroadcastHashJoin? -- 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]
