This is probably very simple, but I can't figure it out. I have a directory structure like this (dir2 is inside dir1):

        --- file1
        --- file2

What is the best way to 'flatten' this director structure in such a way to get file1 and file2 in dir1 not dir2.


You can do this with GNU find and GNU mv:

find /dir1 -mindepth 2 -type f -exec mv -t /dir1 -i '{}' +

Basically, the way that works if that find goes through the entire directory tree and for each file (-type f) that is not in the top-level directory (-mindepth 2), it runs a mv to move it to the directory you want (-exec mv … +). The -t argument to mv lets you specify the destination directory first, which is needed because the + form of -exec puts all the source locations at the end of the command. The -i makes mv ask before overwriting any duplicates; you can substitute -f to overwrite them without asking (or -n to not ask or overwrite).

As Stephane Chazelas points out, the above only works with GNU tools (which are standard on Linux, but not most other systems). The following is somewhat slower (because it invokes mv multiple times) but much more universal:

find /dir1 -mindepth 2 -type f -exec mv -i '{}' /dir1 ';'
  • 3
    Edited to use -exec + so that it doesn't execute a large number of processes of mv – Random832 Oct 24 '12 at 19:32
  • 1
    @Random832 And going to revert again, because + doesn't work. mv needs the destination as the final argument, but + would have the sources as the final argument. Find wont even accept the syntax you changed it to (find: missing argument to `-exec') – derobert Oct 24 '12 at 19:34
  • 1
    @Random832 but I suppose mv has a -t we can use, so I'll change it to that. – derobert Oct 24 '12 at 19:36
  • @Random832 See How can I use two bash commands in -exec of find command? – Gilles Oct 24 '12 at 22:10
  • 1
    @Dom find prints hidden (dot) files by default. The depth is relative to the directory you pass to find. – derobert Jul 15 '13 at 15:32

In zsh:

mv dir1/*/**/*(.D) dir1

**/ traverses subdirectories recursively. The glob qualifier . matches regular files only, and D ensures that dot files are included (by default, files whose name starts with a . are excluded from wildcard matches). To clean up now-empty directories afterwards, run rmdir dir1/**/*(/Dod)/ restricts to directories, and od orders the matches depth first so as to remove dir1/dir2/dir3 before dir1/dir2.

If the total length of the file names is very large, you may run into a limitation on the command line length. Zsh has builtins for mv and rmdir which are not affected by this limitation: run zmodload zsh/files to enable them.

With only POSIX tools:

find dir1 -type f -exec mv {} dir1 \;
find dir1 -depth -exec rmdir {} \;

or (faster because it doesn't have to run a separate process for each file)

find dir1 -type f -exec sh -c 'mv "$@" dir1' _ {} +
find dir1 -depth -exec rmdir {} +
  • 1
    This should be the accepted answer! Especially with the concise zsh version. – Adamski Sep 6 '17 at 10:18

Try doing this :

cp /dir1/dir2/file{1,2} /another/place

or for each files matching file[0-9]* in the subdir :

cp /dir1/dir2/file[0-9]* /another/place

See http://mywiki.wooledge.org/glob

  • I should have indicated this, but I have to many files to use {} in my real problem. – turtle Oct 24 '12 at 19:30
  • See my second solution – Gilles Quenot Oct 24 '12 at 19:31
  • Bingo. Thanks for the help. This is definitely the best solution. – turtle Oct 24 '12 at 19:34

I wrote two functions you can use together that do just that, you can limit the directory level by adding a -maxdepth $VAL parameter.

# This scripts flattens the file directory
# Run this script with a folder as parameter:
# $ path/to/script path/to/folder


    local DIR="$1"
    for dir in "$DIR"/*/
        [ -d "${dir}" ] || continue # if not a directory, skip
        if [ "$(ls -A "$dir")" ]; then
            rmEmptyDirs "$dir"
            rmdir "$dir"
    if [ "$(ls -A "$DIR")" ]; then
        rmEmptyDirs "$DIR"

    local DIR="$1"
    find "$DIR" -mindepth 2 -type f -exec mv -i '{}' "$DIR" ';'

read -p "Do you wish to flatten folder: ${1}? " -n 1 -r
echo    # (optional) move to a new line
if [[ $REPLY =~ ^[Yy]$ ]]
    flattenDir "$1" &
    rmEmptyDirs "$1" &
    echo "Done";
  • Man, I just misused your script by forgetting the path argument, that just really fucked up my server. Ok, I'm the guy who copy paste things and misuse them, but guys, be wise and add checks / confirmations on scripts that delete / move stuffs like that... – dulgan Aug 1 '18 at 8:32
  • Whoops! I am sorry to hear that. Hope you have a backup... I added a confirmation for future protection. – Bruno Aug 2 '18 at 10:34
  • Thanks @Bruno that's much better this way. My server is still running flawlessly, i commented the "flatten" part to just delete empty directories recursively from (and that was my error) root, until I saw an error that made me stop running the script. – dulgan Aug 2 '18 at 14:05

Expanding on the popular answer for this question, since I had a use-case for flattening a directory containing files of the same name.

├── dir2
│   └── file
└── dir3
    └── file

In this case, the -i (--interactive) option passed to mv wouldn't yield the desired result to flatten the directory structure and handle name conflicts. So it's simply replaced with --backup=t (equivalent to --backup=numbered). More documentation on the -b (--backup) option available at https://www.gnu.org/software/coreutils/manual/coreutils.html#Backup-options.

Resulting in:

find dir1/ -mindepth 2 -type f -exec mv -t dir1/ --backup=t '{}' +

Which yields:

├── dir2
├── dir3
├── file
└── file.~1~

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.