rxjava遍历文件夹的巨坑

递归遍历文件夹

java 8之前遍历文件夹用File.list()&＃xff0c;而rxjava要想遍历文件夹&＃xff0c;那么最简单的做法是递归遍历。

public Observable recursiveDir(final File parent) {if (!parent.isDirectory()) {return Observable.just(parent);}File[] files &＃61; null;if ((files &＃61; parent.listFiles()) &＃61;&＃61; null || files.length &＃61;&＃61; 0){}return Observable.empty();}return Observable.fromArray(files).flatMap(new Function>() {&＃64;Overridepublic Observable apply(File file) throws Exception {return recursiveDir(file);}});
}

递归的写法简单写起来爽&＃xff0c;不过跑起来就要哭死你了&＃xff0c;StackOverflow!OutOfMemory&＃xff01;。不信试试遍历下手机sdcard下面的tencent目录&＃xff0c;不是巨慢就是崩溃&＃xff01;

堆栈遍历文件夹

rxJava递归遍历文件夹除了本身递归在队栈保存恢复上消耗系统资源&＃xff0c;还因为rxjava在每一层递归中产生一个Observable&＃xff0c;而Observable如果是并发的很可能很快就消耗尽线程池资源。既然递归方式不行&＃xff0c;那就改成自己维护堆栈的方式遍历呗&＃xff0c;直接看代码吧

/*** Created by droidwolf on 2017/11/14.* https://my.oschina.net/droidwolf* 转载请注明*/
public class FileTreeWalker implements Iterable {private ArrayDeque mDirectories &＃61;null;private ArrayDeque mFiles &＃61; null;public FileTreeWalker walk(File path) {if (mDirectories !&＃61; null && !mDirectories.isEmpty()) {mDirectories.clear();}if (mFiles !&＃61; null && !mFiles.isEmpty()) {mFiles.clear();}walkDir(path);return this;}private void walkDir(File path) {if (path &＃61;&＃61; null || !path.exists() || !path.isDirectory()) {return;}final File[] files &＃61; path.listFiles();if (files &＃61;&＃61; null || files.length &＃61;&＃61; 0) {return;}if(mDirectories&＃61;&＃61;null) mDirectories &＃61; new ArrayDeque(256);if(mFiles&＃61;&＃61;null) mFiles &＃61; new ArrayDeque(512);for (File f : files) {if (f.isDirectory()) {mDirectories.push(f);} else {mFiles.addLast(f);}}}&＃64;Overridepublic Iterator iterator() {return mIterator;}private final Iterator mIterator&＃61;new Iterator() {&＃64;Overridepublic boolean hasNext() {return (mFiles!&＃61;null &&!mFiles.isEmpty()) || (mDirectories!&＃61;null&& !mDirectories.isEmpty());}&＃64;Overridepublic File next() {if (mFiles !&＃61; null && !mFiles.isEmpty()) {final File f &＃61; mFiles.pollFirst();return f;} else if (mDirectories !&＃61; null && !mDirectories.isEmpty()) {final File dir &＃61; mDirectories.pop();walkDir(dir);return dir;}return null;}};
}

rxjava调用FileTreeWalker 

Observable.fromIterable(new FileTreeWalker().walk(path)).subscribeOn(Schedulers.io()).filter(file -> {//。。。return false;}).observeOn(AndroidSchedulers.mainThread()).subscribe(fileNext -> {//。。。});

当然也可以在java8中使用FileTreeWalker

StreamSupport.stream(Spliterators.spliteratorUnknownSize(new FileTreeWalker().walk(new File("C:\\")).iterator(),Spliterator.NONNULL),true )
.forEach(file->{if(file.isDirectory()) {System.out.println(file.getName());}else {System.out.println("\t"&＃43;file.getName());}
});

总结

递归很多时候写的爽快但遇到瓶颈时优化是个头疼的问题&＃xff0c;如果语言自带尾递归优化外挂那还好&＃xff0c;没有的话还是得改变原来的思路重来。最后得说一下这个FileTreeWalker实际是不安全的&＃xff0c;因为它持有文件夹和文件队列&＃xff0c;这使得FileTreeWalker在遍历过程中产生状态的变化&＃xff0c;如果在多线程中这是个忌讳。至于怎么改&＃xff0c;自己看着办吧!:(