一行程序解决方案(Java8)，递归删除所有文件和目录，包括起始目录:

try (var dirStream = Files.walk(Paths.get("c:/dir_to_delete/"))) {
    dirStream
        .map(Path::toFile)
        .sorted(Comparator.reverseOrder())
        .forEach(File::delete);
}

我们使用一个颠倒顺序的比较器，否则File::delete将不能删除可能非空的目录。所以，如果你想保留目录，只删除文件，只需在sorted()中删除比较器或完全删除排序，并添加文件过滤器:

try (var dirStream = Files.walk(Paths.get("c:/dir_to_delete/"))) {
    dirStream
        .filter(Files::isRegularFile)
        .map(Path::toFile)
        .forEach(File::delete);
}

for(Path p : Files.walk(directoryToDelete).
        sorted((a, b) -> b.compareTo(a)). // reverse; files before dirs
        toArray(Path[]::new))
{
    Files.delete(p);
}

或者如果你想处理IOException:

Files.walk(directoryToDelete).
    sorted((a, b) -> b.compareTo(a)). // reverse; files before dirs
    forEach(p -> {
        try { Files.delete(p); }
        catch(IOException e) { /* ... */ }
      });

为了更安全的使用，我编写了有3个安全标准的程序。

package ch.ethz.idsc.queuey.util;

import java.io.File;
import java.io.IOException;

/** recursive file/directory deletion
 * 
 * safety from erroneous use is enhanced by three criteria
 * 1) checking the depth of the directory tree T to be deleted
 * against a permitted upper bound "max_depth"
 * 2) checking the number of files to be deleted #F
 * against a permitted upper bound "max_count"
 * 3) if deletion of a file or directory fails, the process aborts */
public final class FileDelete {
    /** Example: The command
     * FileDelete.of(new File("/user/name/myapp/recordings/log20171024"), 2, 1000);
     * deletes given directory with sub directories of depth of at most 2,
     * and max number of total files less than 1000. No files are deleted
     * if directory tree exceeds 2, or total of files exceed 1000.
     * 
     * abort criteria are described at top of class
     * 
     * @param file
     * @param max_depth
     * @param max_count
     * @return
     * @throws Exception if criteria are not met */
    public static FileDelete of(File file, int max_depth, int max_count) throws IOException {
        return new FileDelete(file, max_depth, max_count);
    }

    // ---
    private final File root;
    private final int max_depth;
    private int removed = 0;

    /** @param root file or a directory. If root is a file, the file will be deleted.
     *            If root is a directory, the directory tree will be deleted.
     * @param max_depth of directory visitor
     * @param max_count of files to delete
     * @throws IOException */
    private FileDelete(final File root, final int max_depth, final int max_count) throws IOException {
        this.root = root;
        this.max_depth = max_depth;
        // ---
        final int count = visitRecursively(root, 0, false);
        if (count <= max_count) // abort criteria 2)
            visitRecursively(root, 0, true);
        else
            throw new IOException("more files to be deleted than allowed (" + max_count + "<=" + count + ") in " + root);
    }

    private int visitRecursively(final File file, final int depth, final boolean delete) throws IOException {
        if (max_depth < depth) // enforce depth limit, abort criteria 1)
            throw new IOException("directory tree exceeds permitted depth");
        // ---
        int count = 0;
        if (file.isDirectory()) // if file is a directory, recur
            for (File entry : file.listFiles())
                count += visitRecursively(entry, depth + 1, delete);
        ++count; // count file as visited
        if (delete) {
            final boolean deleted = file.delete();
            if (!deleted) // abort criteria 3)
                throw new IOException("cannot delete " + file.getAbsolutePath());
            ++removed;
        }
        return count;
    }

    public int deletedCount() {
        return removed;
    }

    public void printNotification() {
        int count = deletedCount();
        if (0 < count)
            System.out.println("deleted " + count + " file(s) in " + root);
    }
}

在Java 7中，我们终于可以通过可靠的符号链接检测来做到这一点。(我不认为Apache的commons-io现在有可靠的符号链接检测，因为它不处理Windows上用mklink创建的链接。)

出于历史原因，这里有一个java 7之前的答案，它遵循符号链接。

void delete(File f) throws IOException {
  if (f.isDirectory()) {
    for (File c : f.listFiles())
      delete(c);
  }
  if (!f.delete())
    throw new FileNotFoundException("Failed to delete file: " + f);
}

rm -rf的性能比FileUtils.deleteDirectory要好得多。

经过大量的基准测试后，我们发现使用rm -rf比使用FileUtils.deleteDirectory快几倍。

当然，如果您有一个小的或简单的目录，这并不重要，但在我们的例子中，我们有多个gb和深度嵌套的子目录，使用FileUtils.deleteDirectory需要10分钟以上，使用rm -rf只需1分钟。

下面是我们粗略的Java实现:

// Delete directory given and all subdirectories and files (i.e. recursively).
//
static public boolean deleteDirectory( File file ) throws IOException, InterruptedException {

    if ( file.exists() ) {

        String deleteCommand = "rm -rf " + file.getAbsolutePath();
        Runtime runtime = Runtime.getRuntime();

        Process process = runtime.exec( deleteCommand );
        process.waitFor();

        return true;
    }

    return false;

}

如果您正在处理大型或复杂的目录，值得一试。

一行程序解决方案(Java8)，递归删除所有文件和目录，包括起始目录:

try (var dirStream = Files.walk(Paths.get("c:/dir_to_delete/"))) {
    dirStream
        .map(Path::toFile)
        .sorted(Comparator.reverseOrder())
        .forEach(File::delete);
}

我们使用一个颠倒顺序的比较器，否则File::delete将不能删除可能非空的目录。所以，如果你想保留目录，只删除文件，只需在sorted()中删除比较器或完全删除排序，并添加文件过滤器:

try (var dirStream = Files.walk(Paths.get("c:/dir_to_delete/"))) {
    dirStream
        .filter(Files::isRegularFile)
        .map(Path::toFile)
        .forEach(File::delete);
}