作者：gnuhpc
出处：http://www.cnblogs.com/gnuhpc/
1.引言
Java I/O系统是建立在数据流概念之上的，而在UNIX/Linux中有一个类似的概念，就是管道，它具有将一个程序的输出当作另一个程序的输入的能力。在Java中，可以使用管道流进行线程之间的通信，输入流和输出流必须相连接，这样的通信有别于一般的Shared Data通信，其不需要一个共享的数据空间。
2.相关类及其关系
1）字节流：
分为管道输出流（PipedOutputStream）和管道输入流（PipedInputStream），利用 java.io.PipedOutputStream和java.io.PipedInputStream可以实现线程之间的二进制信息传输。如果要进行管道输出，则必须把输出流连在输入流上。 java.io.PipedOutputStream是java.io.OutputStream的直接子类，而java.io. PipedInputStream是java.io.InputStream的直接子类。PipedOutputStream和 PipedInputStream往往成对出现、配合使用。举例说明：
TestPipe.Java
import java.io.IOException;
public class TestPipe {
public static void main(String[] args) {
Send s = new Send();
Receive r = new Receive();
try {
s.getPos().connect(r.getPis()); // 连接管道
} catch (IOException e) {
e.printStackTrace();
}
new Thread(s).start(); // 启动线程
new Thread(r).start(); // 启动线程
}
}
Receive.java
import java.io.IOException;
import java.io.PipedInputStream;
class Receive implements Runnable { // 实现Runnable接口
private PipedInputStream pis = null;
public Receive() {
this.pis = new PipedInputStream(); // 实例化输入流
}
public void run() {
byte b[] = new byte[1024];
int len = 0;
try {
len = this.pis.read(b); // 接收数据
} catch (IOException e) {
e.printStackTrace();
}
try {
this.pis.close();
} catch (IOException e) {
e.printStackTrace();
}
System.out.println("接收的内容为：" + new String(b, 0, len));
}
public PipedInputStream getPis() {
return pis;
}
}
Send.java
import java.io.IOException;
import java.io.PipedOutputStream;
class Send implements Runnable {
// 实现Runnable接口
private PipedOutputStream pos = null; // 管道输出流
public Send() {
this.pos = new PipedOutputStream();// 实例化输出流
}
public void run() {
String str = "Hello World!!!";
try {
this.pos.write(str.getBytes()); // 输出信息
} catch (IOException e) {
e.printStackTrace();
}
try {
this.pos.close(); // 关闭输出流
} catch (IOException e) {
e.printStackTrace();
}
}
public PipedOutputStream getPos() { // 通过线程类得到输出流
return pos;
}
}
我们可以看到使用管道流，通过connect方法进行连接，实现了Send线程和Receive线程之间的通信。
注意：
PipedInputStream中实际是用了一个1024字节固定大小的循环缓冲区。写入PipedOutputStream的数据实际上保存到对应的 PipedInputStream的内部缓冲区。从PipedInputStream执行读操作时，读取的数据实际上来自这个内部缓冲区。如果对应的 PipedInputStream输入缓冲区已满，任何企图写入PipedOutputStream的线程都将被阻塞。而且这个写操作线程将一直阻塞，直至出现读取PipedInputStream的操作从缓冲区删除数据。这也就是说往PipedOutputStream写数据的线程Send若是和从PipedInputStream读数据的线程Receive是同一个线程的话，那么一旦Send线程发送数据过多（大于1024字节），它就会被阻塞，这就直接导致接受数据的线程阻塞而无法工作（因为是同一个线程嘛），那么这就是一个典型的死锁现象，这也就是为什么javadoc中关于这两个类的使用时告诉大家要在多线程环境下使用的原因了。
应用：过滤器模式
使用这个模式的典型例子是Unix的shell命令。这个模式的好处在于过滤器无需知道它与何种东西进行连接，并且这可以实现并行，而且系统的可扩展性可以根据添加删除或者改变Filter进行增强。
在这举一个不断计算平均值的例子，producer作为前端的数据源，不断产生随机数，通过pipe进入filter进行数据处理，然后通过第二个pipe就行后端处理。
import java.util.*;
import java.io.*;
public class PipeTest
/* 建立3个线程（Producer、Filter、Consumer）类和两组通信管道，通过多线程将管道1的数据传送到管道2中，实现管道的通信。
* Producer => pout1->pin1 => Filter(pin1->pout2) => pout2->pin2 =>Consumer
*/
{
public static void main(String args[]) {
try {
PipedOutputStream pout1 = new PipedOutputStream();
PipedInputStream pin1 = new PipedInputStream(pout1);
PipedOutputStream pout2 = new PipedOutputStream();
PipedInputStream pin2 = new PipedInputStream(pout2);
/* construct threads */
Producer prod = new Producer(pout1);
Filter filt = new Filter(pin1, pout2);
Consumer cons = new Consumer(pin2);
/* start threads */
prod.start();
filt.start();
cons.start();
} catch (IOException e) {
}
}
}
// 前端：该类的作用是产生随机数，并将其放到管道1的输出流中
class Producer extends Thread {
private DataOutputStream out;// DataOutputStream是用于写入一些基本类型数据的类，此类的实例用于生成伪随机数流
private Random rand = new Random();
public Producer(OutputStream os) {
out = new DataOutputStream(os);
}
public void run() {
while (true) {
try {
double num = rand.nextDouble();
// 将double值直接写入流
out.writeDouble(num);
System.out.println("写入流中的值是 ：" + num);
out.flush();
sleep(Math.abs(rand.nextInt()%10));//随机休眠一段时间
} catch (Exception e) {
System.out.println("Error: " + e);
}
}
}
}
// 过滤器，起数据处理作用，读取管道1中输入流的内容，并将其放到管道2的输出流中
class Filter extends Thread {
private DataInputStream in;
private DataOutputStream out;
private double total = 0;
private int count = 0;
public Filter(InputStream is, OutputStream os) {
in = new DataInputStream(is);
out = new DataOutputStream(os);
}
public void run() {
for (;;) {
try {
double x = in.readDouble(); // 读取流中的数据
total += x;
count++;
if (count != 0) {
double d = total / count;
out.writeDouble(d); // 将得到的数据平均值写入流
}
} catch (IOException e) {
System.out.println("Error: " + e);
}
}
}
}
// 后端：读取管道2输入流的内容
class Consumer extends Thread {
private double old_avg = 0;
private DataInputStream in;
public Consumer(InputStream is) {
in = new DataInputStream(is);
}
public void run() {
for (;;) {
try {
double avg = in.readDouble();
if (Math.abs(avg - old_avg) > 0.01) {
System.out.println("现在的平均值是： " + avg);
System.out.println();
old_avg = avg;
}
} catch (IOException e) {
System.out.println("Error: " + e);
}
}
}
}
2）字符流
Java利用 java.io.PipedWriter和java.io.PipedReader在线程之间传输字符信息。与 java.io.PipedOutputStream和java.io.PipedInputStream类似，java.io.PipedWriter 是java.io.Writer的直接子类，java.io.PipedReader是java.io.Reader的直接子类。PipedWriter拥有一个允许指定输入管道字符流的构造方法，而PipedReader拥有一个允许指定输出管道字符流的构造方法。从而使得PipedWriter和PipedReader往往成对出现、配合使用。
以典型KWIC系统为例，下边的代码演示了如何使用字符流并且使用了过滤器模式：ReadLineThread --Pipe1 --> ShiftThread -- Pipe2 --> SortLinesThread
import java.io.BufferedReader;
import java.io.BufferedWriter;
import java.io.FileNotFoundException;
import java.io.FileReader;
import java.io.FileWriter;
import java.io.IOException;
import java.io.PipedReader;
import java.io.PipedWriter;
import java.util.ArrayList;
import java.util.Collections;
import java.util.StringTokenizer;
public class KwicPipe {
public static void main(String[] args) {
try {
//get the input and output path
String src = args[0];
String dest = args[1];
//(writeToShiftThread => readFromShiftThread) = Pipe1
PipedReader readFromShiftThread = new PipedReader();
PipedWriter writeToShiftThread = new PipedWriter(readFromShiftThread);
//(writeToSortLinesThread => readFromSortLinesThread) = Pipe2
PipedReader readFromSortLinesThread = new PipedReader();
PipedWriter writeToSortLinesThread = new PipedWriter(readFromSortLinesThread);
//ReadLineThread --Pipe1 --> ShiftThread -- Pipe2 --> SortLinesThread
ReadLineThread R1 = new ReadLineThread(writeToShiftThread,src);
ShiftThread R2 = new ShiftThread(readFromShiftThread,writeToSortLinesThread);
SortLinesThread R3 = new SortLinesThread(readFromSortLinesThread,dest);
//Start the three processing thread
R1.start();
R2.start();
R3.start();
}
catch (IOException e) {
System.out.println("NO I/O");
}
}
}
// read the content of kwici.dat and send the lines to another thread
class ReadLineThread extends Thread {
PipedWriter PipeIn;
String InputFilename= null;
ReadLineThread(PipedWriter PlaceInPipe, String InputFilename) {
PipeIn = PlaceInPipe;
this.InputFilename = InputFilename;
}
private BufferedReader fileopen(String InputFilename) {
BufferedReader input_file = null;
try {
input_file = new BufferedReader(new FileReader(InputFilename));
} catch (IOException e) {
System.err.println(("File not open" + e.toString()));
System.exit(1);
}
return input_file;
}
public void run() {
try {
String Input;
BufferedReader TheInput = fileopen(InputFilename);
while ( (Input = TheInput.readLine()) != null) {
System.out.println(Input);
PipeIn.write(Input + "/n"); // Read from the file and then write to the pipe1
}
}
catch (FileNotFoundException e) {
System.out.println("NO FILE ");
}
catch (IOException e) {
System.out.println("NO I/O");
}
}
}
// read the lines from ReadLineThread and shift them. Send all the shifted lines to SortLinesThread
class ShiftThread extends Thread {
PipedReader PipeOut;
PipedWriter PipeIn;
ShiftThread(PipedReader ReadFromPipe, PipedWriter WriteToPipe) {
PipeOut = ReadFromPipe;
PipeIn = WriteToPipe;
}
public void run() {
char[] cbuf = new char[80];
int i, j;
StringBuffer linebuff = new StringBuffer();
try {
// read from ReadLineThread
i = PipeOut.read(cbuf, 0, 80);
while (i != -1) {
for (j = 0; j < i; j++) {
//if new line
if (cbuf[j]=='/n'){
// When reach the end of line,shift it
shiftline(linebuff.toString());
// empty the buffer
linebuff.delete(0, linebuff.length());
}
else {
linebuff.append(cbuf[j]);
}
}
i = PipeOut.read(cbuf, 0, 80); //get next buffer's worth
}
}
catch (FileNotFoundException e) {
System.out.println("NO FILE ");
}
catch (IOException e) {
System.out.println("NO I/O or end of stream (ShiftThread terminated)");
}
/* BECAUSE
* If a thread was providing data characters to the connected piped output,
* but the thread is no longer alive, then an IOException is thrown. (javadoc)
*/
}
private void shiftline( String line )
{
String onetoken = new String ();
StringTokenizer tokens =
new StringTokenizer( line );
ArrayList<String> Tokens = new ArrayList<String> ();
int count = tokens.countTokens();
for ( int i = 0; i < count; i++)
{
onetoken = tokens.nextToken();
if (!((onetoken.compareTo( "a" ) == 0) && (onetoken.compareTo( "an" ) == 0) && (onetoken.compareTo( "and" ) == 0) && (onetoken.compareTo( "the" ) == 0)))
{
Tokens.add(onetoken);
}
}
for ( int tokencount = 0; tokencount < count; tokencount++ )
{
StringBuffer linebuffer = new StringBuffer ();
int index = tokencount;
for ( int i = 0; i< count; i++ )
{
if (index >= count)
index = 0;
linebuffer.append ( Tokens.get(index) );
linebuffer.append (" ");
index++;
} //for i
line = linebuffer.toString();
// send the line to the SortLinesThread
try {
PipeIn.write(line+ "/n");
} catch (IOException e) {
e.printStackTrace();
}
} // for token count
return;
}
}
class SortLinesThread extends Thread {
PipedReader PipeOut;
String OutputFilename;
ArrayList<String> KwicList = new ArrayList<String>();
SortLinesThread(PipedReader ReadFromPipe, String OutputFilename) {
PipeOut = ReadFromPipe;
this.OutputFilename = OutputFilename;
}
public void run() {
char[] cbuf = new char[80];
int i, j;
StringBuffer linebuff = new StringBuffer();
try {
// read from ShiftLineThread
i = PipeOut.read(cbuf, 0, 80);
while (i != -1) { // I don't know we're using that (The method Read blocks until at least one character of input is available.)
for (j = 0; j < i; j++) {
//if new line
if (cbuf[j]=='/n'){
// add it to the ArrayList
KwicList.add(linebuff.toString());
// adn empty the buffer
linebuff.delete(0, linebuff.length());
}
else {
//append the character to the line
linebuff.append(cbuf[j]);
}
}
i = PipeOut.read(cbuf, 0, 80); //get next buffer's worth
}
}
catch (FileNotFoundException e) {
System.out.println("NO FILE ");
}
catch (IOException e) {
System.out.println("NO I/O or end of stream (SortLinesThread terminated)");
}
/* BECAUSE
* If a thread was providing data characters to the connected piped output,
* but the thread is no longer alive, then an IOException is thrown. (javadoc)
*/
// when the reading is finished, sort the ArrayList and diplay
Collections.sort(KwicList);//sort when added
displaylist ( KwicList );//Standard Output
//Export to file
try {
export(KwicList, OutputFilename);
} catch (Exception e) {
System.out.println("Error Output File ");
}
}
private void displaylist (ArrayList<String> KwicList )
{
System.out.println ("/nList : " );
for ( int count = 0; count < KwicList.size(); count++ )
System.out.println (KwicList.get (count) );
}
private void export(ArrayList<String> List, String oufFilename) throws Exception{
BufferedWriter writer = null;
try {
writer = new BufferedWriter(new FileWriter(oufFilename));
} catch (FileNotFoundException e) {
System.err.println(("File not open" + e.toString()));
System.exit(1);
}
for (int count = 0; count < List.size(); count++) {
writer.write(List.get(count));
writer.write("/r/n");
}
writer.flush();
writer.close();
System.out.println("Processed Finished");
}
}