一直困扰我的是,在Java中复制文件的唯一方法包括打开流,声明缓冲区,读入一个文件,循环遍历它,并将其写入到另一个steam。网络上充斥着类似的,但仍然略有不同的这类解决方案的实现。
是否有一种更好的方法能够保持在Java语言的范围内(即不涉及执行特定于操作系统的命令)?也许在一些可靠的开源实用程序包中,它至少会模糊这个底层实现,并提供一行解决方案?
一直困扰我的是,在Java中复制文件的唯一方法包括打开流,声明缓冲区,读入一个文件,循环遍历它,并将其写入到另一个steam。网络上充斥着类似的,但仍然略有不同的这类解决方案的实现。
是否有一种更好的方法能够保持在Java语言的范围内(即不涉及执行特定于操作系统的命令)?也许在一些可靠的开源实用程序包中,它至少会模糊这个底层实现,并提供一行解决方案?
当前回答
public static void copyFile(File src, File dst) throws IOException
{
long p = 0, dp, size;
FileChannel in = null, out = null;
try
{
if (!dst.exists()) dst.createNewFile();
in = new FileInputStream(src).getChannel();
out = new FileOutputStream(dst).getChannel();
size = in.size();
while ((dp = out.transferFrom(in, p, size)) > 0)
{
p += dp;
}
}
finally {
try
{
if (out != null) out.close();
}
finally {
if (in != null) in.close();
}
}
}
其他回答
虽然有点晚了,但是这里比较了使用各种文件复制方法复制一个文件所花费的时间。我循环使用这些方法10次,然后取平均值。使用IO流的文件传输似乎是最糟糕的候选者:
方法如下:
private static long fileCopyUsingFileStreams(File fileToCopy, File newFile) throws IOException {
FileInputStream input = new FileInputStream(fileToCopy);
FileOutputStream output = new FileOutputStream(newFile);
byte[] buf = new byte[1024];
int bytesRead;
long start = System.currentTimeMillis();
while ((bytesRead = input.read(buf)) > 0)
{
output.write(buf, 0, bytesRead);
}
long end = System.currentTimeMillis();
input.close();
output.close();
return (end-start);
}
private static long fileCopyUsingNIOChannelClass(File fileToCopy, File newFile) throws IOException
{
FileInputStream inputStream = new FileInputStream(fileToCopy);
FileChannel inChannel = inputStream.getChannel();
FileOutputStream outputStream = new FileOutputStream(newFile);
FileChannel outChannel = outputStream.getChannel();
long start = System.currentTimeMillis();
inChannel.transferTo(0, fileToCopy.length(), outChannel);
long end = System.currentTimeMillis();
inputStream.close();
outputStream.close();
return (end-start);
}
private static long fileCopyUsingApacheCommons(File fileToCopy, File newFile) throws IOException
{
long start = System.currentTimeMillis();
FileUtils.copyFile(fileToCopy, newFile);
long end = System.currentTimeMillis();
return (end-start);
}
private static long fileCopyUsingNIOFilesClass(File fileToCopy, File newFile) throws IOException
{
Path source = Paths.get(fileToCopy.getPath());
Path destination = Paths.get(newFile.getPath());
long start = System.currentTimeMillis();
Files.copy(source, destination, StandardCopyOption.REPLACE_EXISTING);
long end = System.currentTimeMillis();
return (end-start);
}
我在使用NIO通道类时看到的唯一缺点是,我似乎仍然无法找到一种方法来显示中间文件复制进度。
这些方法是经过性能设计的(它们与操作系统本机I/O集成)。 这些方法适用于文件、目录和链接。 提供的每个选项都可以省略——它们是可选的。
实用工具类
package com.yourcompany.nio;
class Files {
static int copyRecursive(Path source, Path target, boolean prompt, CopyOptions options...) {
CopyVisitor copyVisitor = new CopyVisitor(source, target, options).copy();
EnumSet<FileVisitOption> fileVisitOpts;
if (Arrays.toList(options).contains(java.nio.file.LinkOption.NOFOLLOW_LINKS) {
fileVisitOpts = EnumSet.noneOf(FileVisitOption.class)
} else {
fileVisitOpts = EnumSet.of(FileVisitOption.FOLLOW_LINKS);
}
Files.walkFileTree(source[i], fileVisitOpts, Integer.MAX_VALUE, copyVisitor);
}
private class CopyVisitor implements FileVisitor<Path> {
final Path source;
final Path target;
final CopyOptions[] options;
CopyVisitor(Path source, Path target, CopyOptions options...) {
this.source = source; this.target = target; this.options = options;
};
@Override
FileVisitResult preVisitDirectory(Path dir, BasicFileAttributes attrs) {
// before visiting entries in a directory we copy the directory
// (okay if directory already exists).
Path newdir = target.resolve(source.relativize(dir));
try {
Files.copy(dir, newdir, options);
} catch (FileAlreadyExistsException x) {
// ignore
} catch (IOException x) {
System.err.format("Unable to create: %s: %s%n", newdir, x);
return SKIP_SUBTREE;
}
return CONTINUE;
}
@Override
public FileVisitResult visitFile(Path file, BasicFileAttributes attrs) {
Path newfile= target.resolve(source.relativize(file));
try {
Files.copy(file, newfile, options);
} catch (IOException x) {
System.err.format("Unable to copy: %s: %s%n", source, x);
}
return CONTINUE;
}
@Override
public FileVisitResult postVisitDirectory(Path dir, IOException exc) {
// fix up modification time of directory when done
if (exc == null && Arrays.toList(options).contains(COPY_ATTRIBUTES)) {
Path newdir = target.resolve(source.relativize(dir));
try {
FileTime time = Files.getLastModifiedTime(dir);
Files.setLastModifiedTime(newdir, time);
} catch (IOException x) {
System.err.format("Unable to copy all attributes to: %s: %s%n", newdir, x);
}
}
return CONTINUE;
}
@Override
public FileVisitResult visitFileFailed(Path file, IOException exc) {
if (exc instanceof FileSystemLoopException) {
System.err.println("cycle detected: " + file);
} else {
System.err.format("Unable to copy: %s: %s%n", file, exc);
}
return CONTINUE;
}
}
正在复制目录或文件
long bytes = java.nio.file.Files.copy(
new java.io.File("<filepath1>").toPath(),
new java.io.File("<filepath2>").toPath(),
java.nio.file.StandardCopyOption.REPLACE_EXISTING,
java.nio.file.StandardCopyOption.COPY_ATTRIBUTES,
java.nio.file.LinkOption.NOFOLLOW_LINKS);
移动目录或文件
long bytes = java.nio.file.Files.move(
new java.io.File("<filepath1>").toPath(),
new java.io.File("<filepath2>").toPath(),
java.nio.file.StandardCopyOption.ATOMIC_MOVE,
java.nio.file.StandardCopyOption.REPLACE_EXISTING);
递归复制目录或文件
long bytes = com.yourcompany.nio.Files.copyRecursive(
new java.io.File("<filepath1>").toPath(),
new java.io.File("<filepath2>").toPath(),
java.nio.file.StandardCopyOption.REPLACE_EXISTING,
java.nio.file.StandardCopyOption.COPY_ATTRIBUTES
java.nio.file.LinkOption.NOFOLLOW_LINKS );
要复制文件并将其保存到目标路径,您可以使用下面的方法。
public void copy(File src, File dst) throws IOException {
InputStream in = new FileInputStream(src);
try {
OutputStream out = new FileOutputStream(dst);
try {
// Transfer bytes from in to out
byte[] buf = new byte[1024];
int len;
while ((len = in.read(buf)) > 0) {
out.write(buf, 0, len);
}
} finally {
out.close();
}
} finally {
in.close();
}
}
在Java 7中作为标准提供,path.copyTo: http://openjdk.java.net/projects/nio/javadoc/java/nio/file/Path.html http://java.sun.com/docs/books/tutorial/essential/io/copy.html
我不敢相信他们花了这么长时间来标准化像文件复制这样常见而简单的事情:(
根据我的测试,带缓冲区的NIO复制是最快的。下面的工作代码来自我在https://github.com/mhisoft/fastcopy上的一个测试项目
import java.io.Closeable;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.nio.channels.FileChannel;
import java.text.DecimalFormat;
public class test {
private static final int BUFFER = 4096*16;
static final DecimalFormat df = new DecimalFormat("#,###.##");
public static void nioBufferCopy(final File source, final File target ) {
FileChannel in = null;
FileChannel out = null;
double size=0;
long overallT1 = System.currentTimeMillis();
try {
in = new FileInputStream(source).getChannel();
out = new FileOutputStream(target).getChannel();
size = in.size();
double size2InKB = size / 1024 ;
ByteBuffer buffer = ByteBuffer.allocateDirect(BUFFER);
while (in.read(buffer) != -1) {
buffer.flip();
while(buffer.hasRemaining()){
out.write(buffer);
}
buffer.clear();
}
long overallT2 = System.currentTimeMillis();
System.out.println(String.format("Copied %s KB in %s millisecs", df.format(size2InKB), (overallT2 - overallT1)));
}
catch (IOException e) {
e.printStackTrace();
}
finally {
close(in);
close(out);
}
}
private static void close(Closeable closable) {
if (closable != null) {
try {
closable.close();
} catch (IOException e) {
if (FastCopy.debug)
e.printStackTrace();
}
}
}
}