Java 計算資料夾大小六種實現方案：從遞歸到 ForkJoin 全景解析

需求：遞歸遍歷目錄樹，累加所有檔案大小

測試目錄：C:\Documents and Settings\Administrator\桌面\monkeytalk

方案一：單執行緒遞歸（基準版）

private long getTotalSizeOfFilesInDir(final File file) {
    if (file.isFile()) return file.length();
    final File[] children = file.listFiles();
    long total = 0;
    if (children != null)
        for (final File child : children)
            total += getTotalSizeOfFilesInDir(child);
    return total;
}

• 優點：程式碼最簡潔、易理解
• 缺點：無法利用多核、深層目錄棧溢位風險

方案二：FixedThreadPool + Callable + Future（任務分發版）

class SubDirectoriesAndSize {
    final long size;
    final List subDirectories;
    // 建構子...
}

private SubDirectoriesAndSize getTotalAndSubDirs(final File file) {
    long total = 0;
    List subDirectories = new ArrayList<>();
    if (file.isDirectory()) {
        for (File child : file.listFiles()) {
            if (child.isFile()) total += child.length();
            else subDirectories.add(child);
        }
    }
    return new SubDirectoriesAndSize(total, subDirectories);
}

private long getTotalSizeOfFilesInDir(final File file) throws ... {
    ExecutorService service = Executors.newFixedThreadPool(100);
    try {
        long total = 0;
        List directories = new ArrayList<>();
        directories.add(file);
        while (!directories.isEmpty()) {
            List> partialResults = new ArrayList<>();
            for (File directory : directories) {
                partialResults.add(service.submit(() -> getTotalAndSubDirs(directory)));
            }
            directories.clear();
            for (Future f : partialResults) {
                SubDirectoriesAndSize r = f.get(100, TimeUnit.SECONDS);
                directories.addAll(r.subDirectories);
                total += r.size;
            }
        }
        return total;
    } finally { service.shutdown(); }
}

策略： 層級並行，每層目錄並發處理，收集子目錄供下一輪。

方案三：遞歸提交 Callable（簡潔版）

private long getTotalSizeOfFilesInDir(ExecutorService service, File file) throws ... {
    if (file.isFile()) return file.length();
    long total = 0;
    File[] children = file.listFiles();
    if (children != null) {
        List> futures = new ArrayList<>();
        for (File child : children) {
            futures.add(service.submit(() -> getTotalSizeOfFilesInDir(service, child)));
        }
        for (Future f : futures) total += f.get(100, TimeUnit.SECONDS);
    }
    return total;
}

缺點： 遞歸深度大時會建立大量 Future，記憶體壓力大。

方案四：CountDownLatch + AtomicLong（計數器協調版）

final AtomicLong pendingFileVisits = new AtomicLong();
final AtomicLong totalSize = new AtomicLong();
final CountDownLatch latch = new CountDownLatch(1);

private void updateTotalSizeOfFilesInDir(final File file) {
    long fileSize = 0;
    if (file.isFile()) fileSize = file.length();
    else {
        for (File child : file.listFiles()) {
            if (child.isFile()) fileSize += child.length();
            else {
                pendingFileVisits.incrementAndGet();
                service.execute(() -> updateTotalSizeOfFilesInDir(child));
            }
        }
    }
    totalSize.addAndGet(fileSize);
    if (pendingFileVisits.decrementAndGet() == 0) latch.countDown();
}

機制： pendingFileVisits 追蹤進行中任務數，歸零時觸發 latch.countDown()。

方案五：BlockingQueue + AtomicLong（生產者消費者版）

final BlockingQueue fileSizes = new ArrayBlockingQueue<>(500);
final AtomicLong pendingFileVisits = new AtomicLong();

private void exploreDir(final File file) {
    long fileSize = 0;
    if (file.isFile()) fileSize = file.length();
    else {
        for (File child : file.listFiles()) {
            if (child.isFile()) fileSize += child.length();
            else startExploreDir(child); // 遞歸提交子目錄
        }
    }
    fileSizes.put(fileSize); // 生產大小
    pendingFileVisits.decrementAndGet();
}

// 消費者迴圈
while (pendingFileVisits.get() > 0 || fileSizes.size() > 0) {
    Long size = fileSizes.poll(10, TimeUnit.SECONDS);
    totalSize += size;
}

優勢： 解耦遍歷與累加，背壓自然形成（隊列滿即阻塞）。

方案六：JDK 7 ForkJoinPool（分治法標準實作）

private static class FileSizeFinder extends RecursiveTask {
    final File file;
    public FileSizeFinder(File file) { this.file = file; }
    
    @Override
    protected Long compute() {
        long size = 0;
        if (file.isFile()) return file.length();
        File[] children = file.listFiles();
        if (children != null) {
            List> tasks = new ArrayList<>();
            for (File child : children) {
                if (child.isFile()) size += child.length();
                else tasks.add(new FileSizeFinder(child));
            }
            for (ForkJoinTask task : invokeAll(tasks)) size += task.join();
        }
        return size;
    }
}

// 使用
ForkJoinPool pool = new ForkJoinPool();
long total = pool.invoke(new FileSizeFinder(new File(path)));

核心： RecursiveTask 分治，invokeAll 並行執行子任務，join 合併結果。工作竊取演算法自動負載均衡。

方案對比總結

生產建議

1. JDK 7+ 直接用 ForkJoinPool 或 Files.walkFileTree (NIO.2)
2. 超大目錄考慮 Files.walk 串流 + 並行 parallel()
3. 注意 File.listFiles() 可能回傳 null（權限不足）
4. 生產環境加入超時、異常重試、進度回報