在Java中,似乎有不同的方式来读取和写入文件数据。

我想从一个文件读取ASCII数据。可能的方式和它们的区别是什么?


当前回答

我必须以不同的方式进行基准测试。我将对我的发现进行评论,但简而言之,最快的方法是在FileInputStream上使用普通的BufferedInputStream。如果必须读取许多文件,那么三个线程将使总执行时间减少大约一半,但是增加更多的线程将逐渐降低性能,直到使用20个线程完成的时间比仅使用一个线程完成的时间长3倍。

假设您必须读取文件并对其内容执行一些有意义的操作。在示例中,我们从日志中读取行,并计算包含超过一定阈值的行。因此,我假设一行Java 8 Files.lines(Paths.get("/path/to/file.txt"))。Map (line -> line.split(";"))不是一个选项。

我在Java 1.8、Windows 7以及SSD和HDD驱动器上进行了测试。

我写了六个不同的实现:

rawParse:在FileInputStream上使用BufferedInputStream,然后一个字节一个字节地读取切行。这优于任何其他单线程方法,但对于非ascii文件可能非常不方便。

lineReaderParse:在FileReader上使用BufferedReader,逐行读取,通过调用String.split()分割行。这比rawParse慢了大约20%。

linereaderparsepar列:这与lineReaderParse相同,但它使用几个线程。在所有情况下,这是最快的选择。

nioFilesParse:使用java.nio.files.Files.lines()

nioAsyncParse:使用带有完成处理程序和线程池的AsynchronousFileChannel。

nioMemoryMappedParse:使用内存映射文件。这确实是一个糟糕的想法,导致执行时间比任何其他实现都要长至少三倍。

这是在四核i7和SSD驱动器上读取204个4mb文件的平均时间。这些文件是动态生成的,以避免磁盘缓存。

rawParse                11.10 sec
lineReaderParse         13.86 sec
lineReaderParseParallel  6.00 sec
nioFilesParse           13.52 sec
nioAsyncParse           16.06 sec
nioMemoryMappedParse    37.68 sec

我发现在SSD或HDD驱动器上运行的差异比我预期的要小,SSD大约快15%。这可能是因为文件是在一个未分片的硬盘上生成的,并且它们是按顺序读取的,因此旋转驱动器可以像SSD一样执行。

我对nioAsyncParse实现的低性能感到惊讶。要么是我以错误的方式实现了某些东西,要么是使用NIO和完成处理程序的多线程实现与使用java的单线程实现执行相同(甚至更差)。io API。此外,使用CompletionHandler的异步解析代码行要比在旧流上直接实现的代码行长得多,而且要正确实现也比较棘手。

现在,这六个实现后面跟着一个包含它们的类,再加上一个可参数化的main()方法,该方法允许处理文件数量、文件大小和并发度。注意,文件的大小变化为正负20%。这是为了避免由于所有文件大小完全相同而造成的任何影响。

rawParse

public void rawParse(final String targetDir, final int numberOfFiles) throws IOException, ParseException {
    overrunCount = 0;
    final int dl = (int) ';';
    StringBuffer lineBuffer = new StringBuffer(1024);
    for (int f=0; f<numberOfFiles; f++) {
        File fl = new File(targetDir+filenamePreffix+String.valueOf(f)+".txt");
        FileInputStream fin = new FileInputStream(fl);
        BufferedInputStream bin = new BufferedInputStream(fin);
        int character;
        while((character=bin.read())!=-1) {
            if (character==dl) {

                // Here is where something is done with each line
                doSomethingWithRawLine(lineBuffer.toString());
                lineBuffer.setLength(0);
            }
            else {
                lineBuffer.append((char) character);
            }
        }
        bin.close();
        fin.close();
    }
}

public final void doSomethingWithRawLine(String line) throws ParseException {
    // What to do for each line
    int fieldNumber = 0;
    final int len = line.length();
    StringBuffer fieldBuffer = new StringBuffer(256);
    for (int charPos=0; charPos<len; charPos++) {
        char c = line.charAt(charPos);
        if (c==DL0) {
            String fieldValue = fieldBuffer.toString();
            if (fieldValue.length()>0) {
                switch (fieldNumber) {
                    case 0:
                        Date dt = fmt.parse(fieldValue);
                        fieldNumber++;
                        break;
                    case 1:
                        double d = Double.parseDouble(fieldValue);
                        fieldNumber++;
                        break;
                    case 2:
                        int t = Integer.parseInt(fieldValue);
                        fieldNumber++;
                        break;
                    case 3:
                        if (fieldValue.equals("overrun"))
                            overrunCount++;
                        break;
                }
            }
            fieldBuffer.setLength(0);
        }
        else {
            fieldBuffer.append(c);
        }
    }
}

lineReaderParse

public void lineReaderParse(final String targetDir, final int numberOfFiles) throws IOException, ParseException {
    String line;
    for (int f=0; f<numberOfFiles; f++) {
        File fl = new File(targetDir+filenamePreffix+String.valueOf(f)+".txt");
        FileReader frd = new FileReader(fl);
        BufferedReader brd = new BufferedReader(frd);

        while ((line=brd.readLine())!=null)
            doSomethingWithLine(line);
        brd.close();
        frd.close();
    }
}

public final void doSomethingWithLine(String line) throws ParseException {
    // Example of what to do for each line
    String[] fields = line.split(";");
    Date dt = fmt.parse(fields[0]);
    double d = Double.parseDouble(fields[1]);
    int t = Integer.parseInt(fields[2]);
    if (fields[3].equals("overrun"))
        overrunCount++;
}

lineReaderParseParallel

public void lineReaderParseParallel(final String targetDir, final int numberOfFiles, final int degreeOfParalelism) throws IOException, ParseException, InterruptedException {
    Thread[] pool = new Thread[degreeOfParalelism];
    int batchSize = numberOfFiles / degreeOfParalelism;
    for (int b=0; b<degreeOfParalelism; b++) {
        pool[b] = new LineReaderParseThread(targetDir, b*batchSize, b*batchSize+b*batchSize);
        pool[b].start();
    }
    for (int b=0; b<degreeOfParalelism; b++)
        pool[b].join();
}

class LineReaderParseThread extends Thread {

    private String targetDir;
    private int fileFrom;
    private int fileTo;
    private DateFormat fmt = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
    private int overrunCounter = 0;

    public LineReaderParseThread(String targetDir, int fileFrom, int fileTo) {
        this.targetDir = targetDir;
        this.fileFrom = fileFrom;
        this.fileTo = fileTo;
    }

    private void doSomethingWithTheLine(String line) throws ParseException {
        String[] fields = line.split(DL);
        Date dt = fmt.parse(fields[0]);
        double d = Double.parseDouble(fields[1]);
        int t = Integer.parseInt(fields[2]);
        if (fields[3].equals("overrun"))
            overrunCounter++;
    }

    @Override
    public void run() {
        String line;
        for (int f=fileFrom; f<fileTo; f++) {
            File fl = new File(targetDir+filenamePreffix+String.valueOf(f)+".txt");
            try {
            FileReader frd = new FileReader(fl);
            BufferedReader brd = new BufferedReader(frd);
            while ((line=brd.readLine())!=null) {
                doSomethingWithTheLine(line);
            }
            brd.close();
            frd.close();
            } catch (IOException | ParseException ioe) { }
        }
    }
}

nioFilesParse

public void nioFilesParse(final String targetDir, final int numberOfFiles) throws IOException, ParseException {
    for (int f=0; f<numberOfFiles; f++) {
        Path ph = Paths.get(targetDir+filenamePreffix+String.valueOf(f)+".txt");
        Consumer<String> action = new LineConsumer();
        Stream<String> lines = Files.lines(ph);
        lines.forEach(action);
        lines.close();
    }
}


class LineConsumer implements Consumer<String> {

    @Override
    public void accept(String line) {

        // What to do for each line
        String[] fields = line.split(DL);
        if (fields.length>1) {
            try {
                Date dt = fmt.parse(fields[0]);
            }
            catch (ParseException e) {
            }
            double d = Double.parseDouble(fields[1]);
            int t = Integer.parseInt(fields[2]);
            if (fields[3].equals("overrun"))
                overrunCount++;
        }
    }
}

nioAsyncParse

public void nioAsyncParse(final String targetDir, final int numberOfFiles, final int numberOfThreads, final int bufferSize) throws IOException, ParseException, InterruptedException {
    ScheduledThreadPoolExecutor pool = new ScheduledThreadPoolExecutor(numberOfThreads);
    ConcurrentLinkedQueue<ByteBuffer> byteBuffers = new ConcurrentLinkedQueue<ByteBuffer>();

    for (int b=0; b<numberOfThreads; b++)
        byteBuffers.add(ByteBuffer.allocate(bufferSize));

    for (int f=0; f<numberOfFiles; f++) {
        consumerThreads.acquire();
        String fileName = targetDir+filenamePreffix+String.valueOf(f)+".txt";
        AsynchronousFileChannel channel = AsynchronousFileChannel.open(Paths.get(fileName), EnumSet.of(StandardOpenOption.READ), pool);
        BufferConsumer consumer = new BufferConsumer(byteBuffers, fileName, bufferSize);
        channel.read(consumer.buffer(), 0l, channel, consumer);
    }
    consumerThreads.acquire(numberOfThreads);
}


class BufferConsumer implements CompletionHandler<Integer, AsynchronousFileChannel> {

        private ConcurrentLinkedQueue<ByteBuffer> buffers;
        private ByteBuffer bytes;
        private String file;
        private StringBuffer chars;
        private int limit;
        private long position;
        private DateFormat frmt = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");

        public BufferConsumer(ConcurrentLinkedQueue<ByteBuffer> byteBuffers, String fileName, int bufferSize) {
            buffers = byteBuffers;
            bytes = buffers.poll();
            if (bytes==null)
                bytes = ByteBuffer.allocate(bufferSize);

            file = fileName;
            chars = new StringBuffer(bufferSize);
            frmt = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss");
            limit = bufferSize;
            position = 0l;
        }

        public ByteBuffer buffer() {
            return bytes;
        }

        @Override
        public synchronized void completed(Integer result, AsynchronousFileChannel channel) {

            if (result!=-1) {
                bytes.flip();
                final int len = bytes.limit();
                int i = 0;
                try {
                    for (i = 0; i < len; i++) {
                        byte by = bytes.get();
                        if (by=='\n') {
                            // ***
                            // The code used to process the line goes here
                            chars.setLength(0);
                        }
                        else {
                                chars.append((char) by);
                        }
                    }
                }
                catch (Exception x) {
                    System.out.println(
                        "Caught exception " + x.getClass().getName() + " " + x.getMessage() +
                        " i=" + String.valueOf(i) + ", limit=" + String.valueOf(len) +
                        ", position="+String.valueOf(position));
                }

                if (len==limit) {
                    bytes.clear();
                    position += len;
                    channel.read(bytes, position, channel, this);
                }
                else {
                    try {
                        channel.close();
                    }
                    catch (IOException e) {
                    }
                    consumerThreads.release();
                    bytes.clear();
                    buffers.add(bytes);
                }
            }
            else {
                try {
                    channel.close();
                }
                catch (IOException e) {
                }
                consumerThreads.release();
                bytes.clear();
                buffers.add(bytes);
            }
        }

        @Override
        public void failed(Throwable e, AsynchronousFileChannel channel) {
        }
};

所有案例的完全可运行实现

https://github.com/sergiomt/javaiobenchmark/blob/master/FileReadBenchmark.java

其他回答

我编写的这段代码对于非常大的文件要快得多:

public String readDoc(File f) {
    String text = "";
    int read, N = 1024 * 1024;
    char[] buffer = new char[N];

    try {
        FileReader fr = new FileReader(f);
        BufferedReader br = new BufferedReader(fr);

        while(true) {
            read = br.read(buffer, 0, N);
            text += new String(buffer, 0, read);

            if(read < N) {
                break;
            }
        }
    } catch(Exception ex) {
        ex.printStackTrace();
    }

    return text;
}

如果您有一个大文件,您可以使用Apache Commons IO迭代处理该文件,而不会耗尽可用内存。

try (LineIterator it = FileUtils.lineIterator(theFile, "UTF-8")) {
    while (it.hasNext()) {
        String line = it.nextLine();
        // do something with line
    }
}

Java 11 Files.readString中引入了最直观的方法

import java.io.*;
import java.nio.file.Files;
import java.nio.file.Paths;

public class App {
    public static void main(String args[]) throws IOException {
        String content = Files.readString(Paths.get("D:\\sandbox\\mvn\\my-app\\my-app.iml"));
        System.out.print(content);
    }
}

PHP几十年来一直享有这种特权!☺

这基本上与Jesus Ramos的回答完全相同,除了使用File而不是FileReader加上迭代来逐级遍历文件的内容。

Scanner in = new Scanner(new File("filename.txt"));

while (in.hasNext()) { // Iterates each line in the file
    String line = in.nextLine();
    // Do something with line
}

in.close(); // Don't forget to close resource leaks

... 抛出FileNotFoundException

在Java中从文件中读取数据的最简单方法是使用file类读取文件,使用Scanner类读取文件的内容。

public static void main(String args[])throws Exception
{
   File f = new File("input.txt");
   takeInputIn2DArray(f);
}

public static void takeInputIn2DArray(File f) throws Exception
{
    Scanner s = new Scanner(f);
    int a[][] = new int[20][20];
    for(int i=0; i<20; i++)
    {
        for(int j=0; j<20; j++)
        {
            a[i][j] = s.nextInt();
        }
    }
}

PS:别忘了导入java.util.*;扫描仪的工作。