Java 如何在两个或多个流上执行外部联接
在我的应用程序中,我使用了几个提供表单元素(ID、value)的流。元素由以下类定义:Java 如何在两个或多个流上执行外部联接,java,join,java-stream,Java,Join,Java Stream,在我的应用程序中,我使用了几个提供表单元素(ID、value)的流。元素由以下类定义: static final class Element<T> implements Comparable<Element<T>> { final long id; final T value; Element(int id, T value) { this.id = id; this.value = value;
static final class Element<T> implements Comparable<Element<T>> {
final long id;
final T value;
Element(int id, T value) {
this.id = id;
this.value = value;
}
@Override
public int compareTo(Element o) {
return Long.compare(id, o.id);
}
}
因为到目前为止,我对Java的流式API的经验非常基础,所以我通常使用迭代器来完成这些任务
有没有一种惯用(且有效)的方法来执行这种与流的连接?有我可以使用的实用程序库吗
旁注:示例已简化。应用程序从一个类似于面向列的数据存储(没有真正的DMB)的地方接收数据,该数据存储的大小为几GB,并且不容易放入内存。这种连接操作也没有内置支持。最简单的解决方案是编写迭代器,然后使用StreamSupport::stream从迭代器创建流。但是,如果要使用并行流,您可能会发现性能方面存在一些问题。最简单的解决方案是编写迭代器,然后使用StreamSupport::stream从迭代器创建流。但是,如果要使用并行流,您可能会发现性能方面存在一些问题。为了构建完整的外部连接流实现,我使用了两个阻塞队列。队列与每个流相关联,填充类(可运行的实现)从流中读取数据并将其写入队列。当filler类的数据用完时,它会将流结束标记写入队列。然后,我从AbstractSpliterator构造了一个拆分器。tryAdvance方法实现从左队列和右队列中获取一个值,并根据比较器结果使用或保留这些值。我使用元素类的变体。请参阅以下代码:
import java.util.ArrayList;
import java.util.Collection;
public final class Element<T> implements Comparable<Element<T>> {
final long id;
final Collection<T> value;
public Element(int id, T value) {
this.id = id;
// Order preserving
this.value = new ArrayList<T>();
this.value.add(value);
}
Element(long id, Element<T> e1, Element<T> e2) {
this.id = id;
this.value = new ArrayList<T>();
add(e1);
add(e2);
}
private void add(Element<T> e1) {
if(e1 == null) {
this.value.add(null);
} else {
this.value.addAll(e1.value);
}
}
/**
* Used as End-of-Stream marker
*/
Element() {
id = -1;
value = null;
}
@Override
public int compareTo(Element<T> o) {
return Long.compare(id, o.id);
}
}
import java.util.ArrayList;
导入java.util.Collection;
公共final类元素实现了可比较的{
最终长id;
最终收款价值;
公共元素(int id,T值){
this.id=id;
//保序
this.value=新的ArrayList();
这个.价值.增加(价值);
}
元素(长id、元素e1、元素e2){
this.id=id;
this.value=新的ArrayList();
添加(e1);
添加(e2);
}
专用无效添加(元素e1){
if(e1==null){
this.value.add(null);
}否则{
this.value.addAll(e1.value);
}
}
/**
*用作流结束标记
*/
元素(){
id=-1;
值=空;
}
@凌驾
公共整数比较(元素o){
返回Long.compare(id,o.id);
}
}
加入实现
import java.util.Comparator;
import java.util.Spliterator;
import java.util.Spliterators;
import java.util.concurrent.ArrayBlockingQueue;
import java.util.concurrent.BlockingQueue;
import java.util.function.Consumer;
import java.util.stream.Stream;
import java.util.stream.StreamSupport;
public class OuterJoinSpliterator<T> extends Spliterators.AbstractSpliterator<Element<T>> {
private final class Filler implements Runnable {
private final Stream<Element<T>> stream;
private final BlockingQueue<Element<T>> queue;
private Filler(Stream<Element<T>> stream, BlockingQueue<Element<T>> queue) {
this.stream = stream;
this.queue = queue;
}
@Override
public void run() {
stream.forEach(x -> {
try {
queue.put(x);
} catch (final InterruptedException e) {
e.printStackTrace();
}
});
try {
queue.put(EOS);
} catch (final InterruptedException e) {
e.printStackTrace();
}
}
}
public final Element<T> EOS = new Element<T>();
private final int queueSize;
private final BlockingQueue<Element<T>> leftQueue;
private final BlockingQueue<Element<T>> rightQueue;
protected Element<T> leftValue;
protected Element<T> rightValue;
private OuterJoinSpliterator(long estSize, int characteristics, int queueSize,
Stream<Element<T>> leftStream, Stream<Element<T>> rightStream) {
super(estSize, characteristics);
this.queueSize = queueSize;
leftQueue = createQueue();
rightQueue = createQueue();
createFillerThread(leftStream, leftQueue).start();
createFillerThread(rightStream, rightQueue).start();
}
private Element<T> acceptBoth(long id, Element<T> left, Element<T> right) {
return new Element<T>(id, left, right);
}
private final Element<T> acceptLeft(Element<T> left) {
return acceptBoth(left.id, left, null);
}
private final Element<T> acceptRight(Element<T> right) {
return acceptBoth(right.id, null, right);
}
private final Thread createFillerThread(Stream<Element<T>> leftStream, BlockingQueue<Element<T>> queue) {
return new Thread(new Filler(leftStream, queue));
}
private final ArrayBlockingQueue<Element<T>> createQueue() {
return new ArrayBlockingQueue<>(queueSize);
}
@Override
public Comparator<? super Element<T>> getComparator() {
return null;
}
private final boolean isFinished() {
return leftValue == EOS && rightValue == EOS;
}
@Override
public final boolean tryAdvance(Consumer<? super Element<T>> action) {
try {
updateLeft();
updateRight();
if (isFinished()) {
return false;
}
if (leftValue == EOS) {
action.accept(acceptRight(rightValue));
rightValue = null;
} else if (rightValue == EOS) {
action.accept(acceptLeft(leftValue));
leftValue = null;
} else {
switch (leftValue.compareTo(rightValue)) {
case -1:
action.accept(acceptLeft(leftValue));
leftValue = null;
break;
case 1:
action.accept(acceptRight(rightValue));
rightValue = null;
break;
default:
action.accept(acceptBoth(leftValue.id, leftValue, rightValue));
leftValue = null;
rightValue = null;
}
}
} catch (final InterruptedException e) {
return false;
}
return true;
}
private final void updateLeft() throws InterruptedException {
if (leftValue == null) {
leftValue = leftQueue.take();
}
}
private final void updateRight() throws InterruptedException {
if (rightValue == null) {
rightValue = rightQueue.take();
}
}
public static <T> Stream<Element<T>> join(long estSize, int characteristics, int queueSize, boolean parallel, Stream<Element<T>> leftStream, Stream<Element<T>> rightStream) {
Spliterator<Element<T>> spliterator = new OuterJoinSpliterator<>(estSize, characteristics, queueSize, leftStream, rightStream);
return StreamSupport.stream(spliterator, parallel);
}
}
import java.util.Comparator;
导入java.util.Spliterator;
导入java.util.Spliterators;
导入java.util.concurrent.ArrayBlockingQueue;
导入java.util.concurrent.BlockingQueue;
导入java.util.function.Consumer;
导入java.util.stream.stream;
导入java.util.stream.StreamSupport;
公共类OuterJoinSpliterator扩展Spliterators.AbstractSpliterator{
私有最终类填充器实现可运行{
私有最终流;
私有最终阻塞队列;
专用填充程序(流、阻塞队列){
this.stream=流;
this.queue=队列;
}
@凌驾
公开募捐{
stream.forEach(x->{
试一试{
队列。put(x);
}捕获(最终中断异常e){
e、 printStackTrace();
}
});
试一试{
队列放置(EOS);
}捕获(最终中断异常e){
e、 printStackTrace();
}
}
}
公共最终元素EOS=新元素();
私有最终整数队列大小;
私有最终阻塞队列leftQueue;
私有最终阻塞队列rightQueue;
保护元素左值;
受保护元素值;
专用OuterJoinSpliterator(长estSize、int-characteristics、int-queueSize、,
流左流、流右流){
超级(大小、特性);
this.queueSize=queueSize;
leftQueue=createQueue();
rightQueue=createQueue();
createFillerThread(leftStream,leftQueue).start();
createFillerThread(rightStream,rightQueue).start();
}
私有元素acceptBoth(长id、元素左、元素右){
返回新元素(id、左、右);
}
私有最终元素acceptLeft(元素左){
返回acceptBoth(left.id,left,null);
}
私有最终元素接受权限(元素权限){
返回acceptBoth(right.id,null,right);
}
私有最终线程createFillerThread(流leftStream、BlockingQueue队列){
返回新线程(新填充(leftStream,queue));
}
私有最终ArrayBlockingQueue createQueue(){
返回新的ArrayBlockingQueue(queueSize);
}
@凌驾
公共比较器为了构造一个完整的外部连接流实现,我使用了两个阻塞队列。一个队列与每个流和一个填充类(可运行的实现)相关联从流中读取数据并将其写入队列。当filler类耗尽数据时,它会将流结束标记写入队列。然后,我从AbstractSpliterator中构造一个拆分器。tryAdvance方法实现从左队列和右队列中获取一个值,并根据比较使用或保留这些值或结果。我使用元素类的变体。请参见以下代码:
import java.util.ArrayList;
import java.util.Collection;
public final class Element<T> implements Comparable<Element<T>> {
final long id;
final Collection<T> value;
public Element(int id, T value) {
this.id = id;
// Order preserving
this.value = new ArrayList<T>();
this.value.add(value);
}
Element(long id, Element<T> e1, Element<T> e2) {
this.id = id;
this.value = new ArrayList<T>();
add(e1);
add(e2);
}
private void add(Element<T> e1) {
if(e1 == null) {
this.value.add(null);
} else {
this.value.addAll(e1.value);
}
}
/**
* Used as End-of-Stream marker
*/
Element() {
id = -1;
value = null;
}
@Override
public int compareTo(Element<T> o) {
return Long.compare(id, o.id);
}
}
import java.util.ArrayList;
导入java.util.Collection;
公共final类元素实现了可比较的{
最终长id;
最终收款价值;
公共元素(int id,T值){
this.id=id;
//保序
this.value=新的ArrayList();
这个.价值.增加(价值);
}
元素(长id、元素e1、元素e2){
this.id=id;
这
1, "red", null, 87.9
2, "green", 28, 21.0
3, null, 9, 107
4, "red" 17, null
6, "blue", 11, null
import java.util.ArrayList;
import java.util.Collection;
public final class Element<T> implements Comparable<Element<T>> {
final long id;
final Collection<T> value;
public Element(int id, T value) {
this.id = id;
// Order preserving
this.value = new ArrayList<T>();
this.value.add(value);
}
Element(long id, Element<T> e1, Element<T> e2) {
this.id = id;
this.value = new ArrayList<T>();
add(e1);
add(e2);
}
private void add(Element<T> e1) {
if(e1 == null) {
this.value.add(null);
} else {
this.value.addAll(e1.value);
}
}
/**
* Used as End-of-Stream marker
*/
Element() {
id = -1;
value = null;
}
@Override
public int compareTo(Element<T> o) {
return Long.compare(id, o.id);
}
}
import java.util.Comparator;
import java.util.Spliterator;
import java.util.Spliterators;
import java.util.concurrent.ArrayBlockingQueue;
import java.util.concurrent.BlockingQueue;
import java.util.function.Consumer;
import java.util.stream.Stream;
import java.util.stream.StreamSupport;
public class OuterJoinSpliterator<T> extends Spliterators.AbstractSpliterator<Element<T>> {
private final class Filler implements Runnable {
private final Stream<Element<T>> stream;
private final BlockingQueue<Element<T>> queue;
private Filler(Stream<Element<T>> stream, BlockingQueue<Element<T>> queue) {
this.stream = stream;
this.queue = queue;
}
@Override
public void run() {
stream.forEach(x -> {
try {
queue.put(x);
} catch (final InterruptedException e) {
e.printStackTrace();
}
});
try {
queue.put(EOS);
} catch (final InterruptedException e) {
e.printStackTrace();
}
}
}
public final Element<T> EOS = new Element<T>();
private final int queueSize;
private final BlockingQueue<Element<T>> leftQueue;
private final BlockingQueue<Element<T>> rightQueue;
protected Element<T> leftValue;
protected Element<T> rightValue;
private OuterJoinSpliterator(long estSize, int characteristics, int queueSize,
Stream<Element<T>> leftStream, Stream<Element<T>> rightStream) {
super(estSize, characteristics);
this.queueSize = queueSize;
leftQueue = createQueue();
rightQueue = createQueue();
createFillerThread(leftStream, leftQueue).start();
createFillerThread(rightStream, rightQueue).start();
}
private Element<T> acceptBoth(long id, Element<T> left, Element<T> right) {
return new Element<T>(id, left, right);
}
private final Element<T> acceptLeft(Element<T> left) {
return acceptBoth(left.id, left, null);
}
private final Element<T> acceptRight(Element<T> right) {
return acceptBoth(right.id, null, right);
}
private final Thread createFillerThread(Stream<Element<T>> leftStream, BlockingQueue<Element<T>> queue) {
return new Thread(new Filler(leftStream, queue));
}
private final ArrayBlockingQueue<Element<T>> createQueue() {
return new ArrayBlockingQueue<>(queueSize);
}
@Override
public Comparator<? super Element<T>> getComparator() {
return null;
}
private final boolean isFinished() {
return leftValue == EOS && rightValue == EOS;
}
@Override
public final boolean tryAdvance(Consumer<? super Element<T>> action) {
try {
updateLeft();
updateRight();
if (isFinished()) {
return false;
}
if (leftValue == EOS) {
action.accept(acceptRight(rightValue));
rightValue = null;
} else if (rightValue == EOS) {
action.accept(acceptLeft(leftValue));
leftValue = null;
} else {
switch (leftValue.compareTo(rightValue)) {
case -1:
action.accept(acceptLeft(leftValue));
leftValue = null;
break;
case 1:
action.accept(acceptRight(rightValue));
rightValue = null;
break;
default:
action.accept(acceptBoth(leftValue.id, leftValue, rightValue));
leftValue = null;
rightValue = null;
}
}
} catch (final InterruptedException e) {
return false;
}
return true;
}
private final void updateLeft() throws InterruptedException {
if (leftValue == null) {
leftValue = leftQueue.take();
}
}
private final void updateRight() throws InterruptedException {
if (rightValue == null) {
rightValue = rightQueue.take();
}
}
public static <T> Stream<Element<T>> join(long estSize, int characteristics, int queueSize, boolean parallel, Stream<Element<T>> leftStream, Stream<Element<T>> rightStream) {
Spliterator<Element<T>> spliterator = new OuterJoinSpliterator<>(estSize, characteristics, queueSize, leftStream, rightStream);
return StreamSupport.stream(spliterator, parallel);
}
}