转自http://aronlulu.iteye.com/blog/1018370
读取文件大小:1.45G
第一种,OldIO:
- public static void oldIOReadFile() throws IOException{
- BufferedReader br = new BufferedReader(new FileReader("G://lily_947.txt"));
- PrintWriter pw = new PrintWriter("G://oldIO.tmp");
- char[] c = new char[100*1024*1024];
- for(;;){
- if(br.read(c)!=-1){
- pw.print(c);
- }else{
- break;
- }
- }
- pw.close();
- br.close();
- }
耗时70.79s
第二种,newIO:
- public static void newIOReadFile() throws IOException{
- FileChannel read = new RandomAccessFile("G://lily_947.txt","r").getChannel();
- FileChannel writer = new RandomAccessFile("G://newIO.tmp","rw").getChannel();
- ByteBuffer bb = ByteBuffer.allocate(200*1024*1024);
- while(read.read(bb)!=-1){
- bb.flip();
- writer.write(bb);
- bb.clear();
- }
- read.close();
- writer.close();
- }
耗时47.24s
第三种,RandomAccessFile:
- public static void randomReadFile() throws IOException{
- RandomAccessFile read = new RandomAccessFile("G://lily_947.txt","r");
- RandomAccessFile writer = new RandomAccessFile("G://random.tmp","rw");
- byte[] b = new byte[200*1024*1024];
- while(read.read(b)!=-1){
- writer.write(b);
- }
- writer.close();
- read.close();
- }
耗时46.65
第四种,MappedByteBuffer:
- public static void mappedBuffer() throws IOException{
- FileChannel read = new FileInputStream("G://lily_947.txt").getChannel();
- FileChannel writer = new RandomAccessFile("G://buffer.tmp","rw").getChannel();
- long i = 0;
- long size = read.size()/30;
- ByteBuffer bb,cc = null;
- while(i<read.size()&&(read.size()-i)>size){
- bb = read.map(FileChannel.MapMode.READ_ONLY, i, size);
- cc = writer.map(FileChannel.MapMode.READ_WRITE, i, size);
- cc.put(bb);
- i+=size;
- bb.clear();
- cc.clear();
- }
- bb = read.map(FileChannel.MapMode.READ_ONLY, i, read.size()-i);
- cc.put(bb);
- bb.clear();
- cc.clear();
- read.close();
- writer.close();
- }
耗时:36
前三种读法对应的资源占用图如下:
相对于最后一种内存直接映射方式前面的测试其实无意义,基本秒杀。。。。。
对于很大的文件直接分块映射时内存会不够,这是因为MappedByteBuffer未被释放造成的,sun未提供直接回收MappedByteBuffer区域的方法,这个时候有两种方法解决,第一种比较愚笨的:
- System.gc();
- System.runFinalization();
- try {
- Thread.sleep(3000);
- } catch (InterruptedException e) {
- e.printStackTrace();
- }
第二种网上找来的,利用反射调用clean方法:
- public static void unmap(final MappedByteBuffer buffer) {
- if (buffer == null) {
- return;
- }
- AccessController.doPrivileged(new PrivilegedAction<Object>() {
- public Object run() {
- try {
- Method getCleanerMethod = buffer.getClass().getMethod("cleaner", new Class[0]);
- if (getCleanerMethod != null) {
- getCleanerMethod.setAccessible(true);
- Object cleaner = getCleanerMethod.invoke(buffer, new Object[0]);
- Method cleanMethod = cleaner.getClass().getMethod("clean", new Class[0]);
- if (cleanMethod != null) {
- cleanMethod.invoke(cleaner, new Object[0]);
- }
- }
- } catch (Exception e) {
- e.printStackTrace();
- }
- return null;
- }
- });
- }
以上两种方法感觉都别扭,还有就是可以自己分割成物理文件再循环调用,这个也不太美观。
速度也会减慢好多。