[FLINK-10566] Fix exponential planning time of large programs
The traversal of the DAG is not efficient enough at some places which can lead
to very long plan creation times.
This introduces caching for the traversal to avoid traversing nodes multiple
times. Caching is performed at two places:
- when registering Kryo types
- when determining the maximum parallelism
diff --git a/flink-core/src/main/java/org/apache/flink/api/common/Plan.java b/flink-core/src/main/java/org/apache/flink/api/common/Plan.java
index db10ce4..27bf3b4 100644
--- a/flink-core/src/main/java/org/apache/flink/api/common/Plan.java
+++ b/flink-core/src/main/java/org/apache/flink/api/common/Plan.java
@@ -23,6 +23,7 @@
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
+import java.util.HashSet;
import java.util.List;
import java.util.Map.Entry;
import java.util.Set;
@@ -379,10 +380,14 @@
private static final class MaxDopVisitor implements Visitor<Operator<?>> {
+ private final Set<Operator> visitedOperators = new HashSet<>();
private int maxDop = -1;
-
+
@Override
public boolean preVisit(Operator<?> visitable) {
+ if (!visitedOperators.add(visitable)) {
+ return false;
+ }
this.maxDop = Math.max(this.maxDop, visitable.getParallelism());
return true;
}
diff --git a/flink-java/src/main/java/org/apache/flink/api/java/ExecutionEnvironment.java b/flink-java/src/main/java/org/apache/flink/api/java/ExecutionEnvironment.java
index 5a5bddb..6e51bd5 100644
--- a/flink-java/src/main/java/org/apache/flink/api/java/ExecutionEnvironment.java
+++ b/flink-java/src/main/java/org/apache/flink/api/java/ExecutionEnvironment.java
@@ -73,6 +73,7 @@
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
+import java.util.Set;
import static org.apache.flink.util.Preconditions.checkNotNull;
@@ -963,12 +964,16 @@
if (!config.isAutoTypeRegistrationDisabled()) {
plan.accept(new Visitor<org.apache.flink.api.common.operators.Operator<?>>() {
- private final HashSet<Class<?>> deduplicator = new HashSet<>();
+ private final Set<Class<?>> registeredTypes = new HashSet<>();
+ private final Set<org.apache.flink.api.common.operators.Operator<?>> visitedOperators = new HashSet<>();
@Override
public boolean preVisit(org.apache.flink.api.common.operators.Operator<?> visitable) {
+ if (!visitedOperators.add(visitable)) {
+ return false;
+ }
OperatorInformation<?> opInfo = visitable.getOperatorInfo();
- Serializers.recursivelyRegisterType(opInfo.getOutputType(), config, deduplicator);
+ Serializers.recursivelyRegisterType(opInfo.getOutputType(), config, registeredTypes);
return true;
}
diff --git a/flink-tests/src/test/java/org/apache/flink/test/planning/LargePlanTest.java b/flink-tests/src/test/java/org/apache/flink/test/planning/LargePlanTest.java
new file mode 100644
index 0000000..6c30af8
--- /dev/null
+++ b/flink-tests/src/test/java/org/apache/flink/test/planning/LargePlanTest.java
@@ -0,0 +1,81 @@
+/*
+ * 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.flink.test.planning;
+
+import org.apache.flink.api.common.functions.RichMapFunction;
+import org.apache.flink.api.common.typeinfo.Types;
+import org.apache.flink.api.java.DataSet;
+import org.apache.flink.api.java.ExecutionEnvironment;
+import org.apache.flink.api.java.io.DiscardingOutputFormat;
+import org.apache.flink.api.java.tuple.Tuple2;
+import org.apache.flink.client.program.OptimizerPlanEnvironment;
+import org.apache.flink.client.program.PreviewPlanEnvironment;
+
+import org.junit.Test;
+
+/**
+ * Tests that large programs can be compiled to a Plan in reasonable amount of time.
+ */
+public class LargePlanTest {
+
+ @Test(expected = OptimizerPlanEnvironment.ProgramAbortException.class, timeout = 15_000)
+ public void testPlanningOfLargePlan() throws Exception {
+ runProgram(new PreviewPlanEnvironment(), 10, 50);
+ }
+
+ private static void runProgram(ExecutionEnvironment env, int depth, int width) throws Exception {
+ DataSet<String> input = env.fromElements("a", "b", "c");
+ DataSet<String> stats = null;
+
+ for (int i = 0; i < depth; i++) {
+ stats = analyze(input, stats, width / (i + 1) + 1);
+ }
+
+ stats.output(new DiscardingOutputFormat<>());
+ env.execute("depth " + depth + " width " + width);
+ }
+
+ private static DataSet<String> analyze(DataSet<String> input, DataSet<String> stats, int branches) {
+ for (int i = 0; i < branches; i++) {
+ final int ii = i;
+
+ if (stats != null) {
+ input = input.map(
+ new RichMapFunction<String, String>() {
+ @Override
+ public String map(String value) {
+ return value;
+ }
+ }).withBroadcastSet(stats.map(s -> "(" + s + ").map"), "stats");
+ }
+
+ DataSet<String> branch = input
+ .map(s -> new Tuple2<>(0, s + ii)).returns(Types.TUPLE(Types.STRING, Types.INT))
+ .groupBy(0)
+ .minBy(1)
+ .map(kv -> kv.f1).returns(Types.STRING);
+ if (stats == null) {
+ stats = branch;
+ } else {
+ stats = stats.union(branch);
+ }
+ }
+ return stats.map(s -> "(" + s + ").stats");
+ }
+}
