/**
* 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.plan;
import java.io.Serializable;
import java.util.ArrayList;
import java.util.Collection;
import java.util.Collections;
import java.util.HashMap;
import java.util.HashSet;
import java.util.LinkedHashSet;
import java.util.LinkedList;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import java.util.concurrent.atomic.AtomicInteger;
import org.apache.commons.lang3.tuple.ImmutablePair;
import org.apache.commons.lang3.tuple.Pair;
import org.apache.hadoop.hive.ql.plan.Explain.Vectorization;
import org.apache.hadoop.hive.ql.plan.Explain.Level;
import com.google.common.base.Preconditions;
/**
* This class encapsulates all the work objects that can be executed
* in a single Spark job. Currently it's basically a tree with MapWork at the
* roots and and ReduceWork at all other nodes.
*/
@SuppressWarnings("serial")
@Explain(displayName = "Spark", explainLevels = { Level.USER, Level.DEFAULT, Level.EXTENDED },
vectorization = Vectorization.SUMMARY_PATH)
public class SparkWork extends AbstractOperatorDesc {
private static final AtomicInteger counter = new AtomicInteger(1);
private final String name;
private final Set<BaseWork> roots = new LinkedHashSet<BaseWork>();
private final Set<BaseWork> leaves = new LinkedHashSet<>();
protected final Map<BaseWork, List<BaseWork>> workGraph =
new HashMap<BaseWork, List<BaseWork>>();
protected final Map<BaseWork, List<BaseWork>> invertedWorkGraph =
new HashMap<BaseWork, List<BaseWork>>();
protected final Map<Pair<BaseWork, BaseWork>, SparkEdgeProperty> edgeProperties =
new HashMap<Pair<BaseWork, BaseWork>, SparkEdgeProperty>();
private Map<String, List<String>> requiredCounterPrefix;
private Map<BaseWork, BaseWork> cloneToWork;
public SparkWork(String name) {
this.name = name + ":" + counter.getAndIncrement();
cloneToWork = new HashMap<BaseWork, BaseWork>();
}
@Explain(displayName = "DagName")
public String getName() {
return name;
}
/**
* @return a map of "vertex name" to BaseWork
*/
@Explain(displayName = "Vertices", explainLevels = { Explain.Level.USER, Explain.Level.DEFAULT, Explain.Level.EXTENDED },
vectorization = Vectorization.SUMMARY_PATH)
public Map<String, BaseWork> getWorkMap() {
Map<String, BaseWork> result = new LinkedHashMap<String, BaseWork>();
for (BaseWork w: getAllWork()) {
result.put(w.getName(), w);
}
return result;
}
/**
* @return a topologically sorted list of BaseWork
*/
public List<BaseWork> getAllWork() {
List<BaseWork> result = new LinkedList<BaseWork>();
Set<BaseWork> seen = new HashSet<BaseWork>();
for (BaseWork leaf: leaves) {
// make sure all leaves are visited at least once
visit(leaf, seen, result);
}
return result;
}
public Collection<BaseWork> getAllWorkUnsorted() {
return workGraph.keySet();
}
private void visit(BaseWork child, Set<BaseWork> seen, List<BaseWork> result) {
if (seen.contains(child)) {
// don't visit multiple times
return;
}
seen.add(child);
for (BaseWork parent: getParents(child)) {
if (!seen.contains(parent)) {
visit(parent, seen, result);
}
}
result.add(child);
}
/**
* Add all nodes in the collection without any connections.
*/
public void addAll(Collection<BaseWork> c) {
for (BaseWork w: c) {
this.add(w);
}
}
/**
* Add all nodes in the collection without any connections.
*/
public void addAll(BaseWork[] bws) {
for (BaseWork w: bws) {
this.add(w);
}
}
/**
* Whether the specified BaseWork is a vertex in this graph
* @param w the BaseWork to check
* @return whether specified BaseWork is in this graph
*/
public boolean contains(BaseWork w) {
return workGraph.containsKey(w);
}
/**
* add creates a new node in the graph without any connections
*/
public void add(BaseWork w) {
if (workGraph.containsKey(w)) {
return;
}
workGraph.put(w, new LinkedList<BaseWork>());
invertedWorkGraph.put(w, new LinkedList<BaseWork>());
roots.add(w);
leaves.add(w);
}
/**
* disconnect removes an edge between a and b. Both a and
* b have to be in the graph. If there is no matching edge
* no change happens.
*/
public void disconnect(BaseWork a, BaseWork b) {
workGraph.get(a).remove(b);
invertedWorkGraph.get(b).remove(a);
if (getParents(b).isEmpty()) {
roots.add(b);
}
if (getChildren(a).isEmpty()) {
leaves.add(a);
}
edgeProperties.remove(new ImmutablePair<BaseWork, BaseWork>(a, b));
}
/**
* getRoots returns all nodes that do not have a parent.
*/
public Set<BaseWork> getRoots() {
return new LinkedHashSet<BaseWork>(roots);
}
/**
* getLeaves returns all nodes that do not have a child
*/
public Set<BaseWork> getLeaves() {
return new LinkedHashSet<BaseWork>(leaves);
}
public void setRequiredCounterPrefix(Map<String, List<String>> requiredCounterPrefix) {
this.requiredCounterPrefix = requiredCounterPrefix;
}
public Map<String, List<String>> getRequiredCounterPrefix() {
return requiredCounterPrefix;
}
/**
* getParents returns all the nodes with edges leading into work
*/
public List<BaseWork> getParents(BaseWork work) {
Preconditions.checkArgument(invertedWorkGraph.containsKey(work),
"AssertionError: expected invertedWorkGraph.containsKey(work) to be true");
Preconditions.checkArgument(invertedWorkGraph.get(work) != null,
"AssertionError: expected invertedWorkGraph.get(work) to be not null");
return new LinkedList<BaseWork>(invertedWorkGraph.get(work));
}
/**
* getChildren returns all the nodes with edges leading out of work
*/
public List<BaseWork> getChildren(BaseWork work) {
Preconditions.checkArgument(workGraph.containsKey(work),
"AssertionError: expected workGraph.containsKey(work) to be true");
Preconditions.checkArgument(workGraph.get(work) != null,
"AssertionError: expected workGraph.get(work) to be not null");
return new LinkedList<BaseWork>(workGraph.get(work));
}
/**
* remove removes a node from the graph and removes all edges with
* work as start or end point. No change to the graph if the node
* doesn't exist.
*/
public void remove(BaseWork work) {
if (!workGraph.containsKey(work)) {
return;
}
List<BaseWork> children = getChildren(work);
List<BaseWork> parents = getParents(work);
for (BaseWork w: children) {
edgeProperties.remove(new ImmutablePair<BaseWork, BaseWork>(work, w));
invertedWorkGraph.get(w).remove(work);
if (invertedWorkGraph.get(w).size() == 0) {
roots.add(w);
}
}
for (BaseWork w: parents) {
edgeProperties.remove(new ImmutablePair<BaseWork, BaseWork>(w, work));
workGraph.get(w).remove(work);
if (workGraph.get(w).size() == 0) {
leaves.add(w);
}
}
roots.remove(work);
leaves.remove(work);
workGraph.remove(work);
invertedWorkGraph.remove(work);
}
/**
* returns the edge type connecting work a and b
*/
public SparkEdgeProperty getEdgeProperty(BaseWork a, BaseWork b) {
return edgeProperties.get(new ImmutablePair<BaseWork, BaseWork>(a, b));
}
/**
* connect adds an edge between a and b. Both nodes have
* to be added prior to calling connect.
* @param
*/
public void connect(BaseWork a, BaseWork b, SparkEdgeProperty edgeProp) {
workGraph.get(a).add(b);
invertedWorkGraph.get(b).add(a);
roots.remove(b);
leaves.remove(a);
ImmutablePair<BaseWork, BaseWork> workPair = new ImmutablePair<BaseWork, BaseWork>(a, b);
edgeProperties.put(workPair, edgeProp);
}
/*
* Dependency is a class used for explain
*/
public class Dependency implements Serializable, Comparable<Dependency> {
public BaseWork w;
public SparkEdgeProperty prop;
@Explain(displayName = "Name")
public String getName() {
return w.getName();
}
@Explain(displayName = "Shuffle Type")
public String getShuffleType() {
return prop.getShuffleType();
}
@Explain(displayName = "Number of Partitions")
public String getNumPartitions() {
return Integer.toString(prop.getNumPartitions());
}
@Override
public int compareTo(Dependency o) {
int compare = getName().compareTo(o.getName());
if (compare == 0) {
compare = getShuffleType().compareTo(o.getShuffleType());
}
return compare;
}
}
/**
* Task name is usually sorted by natural order, which is the same
* as the topological order in most cases. However, with Spark, some
* tasks may be converted, so have new names. The natural order may
* be different from the topological order. This class is to make
* sure all tasks to be sorted by topological order deterministically.
*/
private static class ComparableName implements Comparable<ComparableName> {
private final Map<String, String> dependencies;
private final String name;
ComparableName(Map<String, String> dependencies, String name) {
this.dependencies = dependencies;
this.name = name;
}
/**
* Check if task n1 depends on task n2
*/
boolean dependsOn(String n1, String n2) {
for (String p = dependencies.get(n1); p != null; p = dependencies.get(p)) {
if (p.equals(n2)) {
return true;
}
}
return false;
}
/**
* Get the number of parents of task n
*/
int getDepth(String n) {
int depth = 0;
for (String p = dependencies.get(n); p != null; p = dependencies.get(p)) {
depth++;
}
return depth;
}
@Override
public int compareTo(ComparableName o) {
if (dependsOn(name, o.name)) {
// this depends on o
return 1;
}
if (dependsOn(o.name, name)) {
// o depends on this
return -1;
}
// No dependency, check depth
int d1 = getDepth(name);
int d2 = getDepth(o.name);
if (d1 == d2) {
// Same depth, using natural order
return name.compareTo(o.name);
}
// Deep one is bigger, i.e. less to the top
return d1 > d2 ? 1 : -1;
}
@Override
public String toString() {
return name;
}
}
@Explain(displayName = "Edges", explainLevels = { Explain.Level.USER, Explain.Level.DEFAULT, Explain.Level.EXTENDED },
vectorization = Vectorization.SUMMARY_PATH)
public Map<ComparableName, List<Dependency>> getDependencyMap() {
Map<String, String> allDependencies = new HashMap<String, String>();
Map<ComparableName, List<Dependency>> result =
new LinkedHashMap<ComparableName, List<Dependency>>();
for (BaseWork baseWork : getAllWork()) {
if (invertedWorkGraph.get(baseWork) != null && invertedWorkGraph.get(baseWork).size() > 0) {
List<Dependency> dependencies = new LinkedList<Dependency>();
for (BaseWork d : invertedWorkGraph.get(baseWork)) {
allDependencies.put(baseWork.getName(), d.getName());
Dependency dependency = new Dependency();
dependency.w = d;
dependency.prop = getEdgeProperty(d, baseWork);
dependencies.add(dependency);
}
if (!dependencies.isEmpty()) {
Collections.sort(dependencies);
result.put(new ComparableName(allDependencies,
baseWork.getName()), dependencies);
}
}
}
return result;
}
/**
* @return all reduce works of this spark work, in sorted order.
*/
public List<ReduceWork> getAllReduceWork() {
List<ReduceWork> result = new ArrayList<ReduceWork>();
for (BaseWork work : getAllWork()) {
if (work instanceof ReduceWork) {
result.add((ReduceWork) work);
}
}
return result;
}
public Map<BaseWork, BaseWork> getCloneToWork() {
return cloneToWork;
}
public void setCloneToWork(Map<BaseWork, BaseWork> cloneToWork) {
this.cloneToWork = cloneToWork;
}
}