rsync N newest files in a directory

Question

What would be the easiest way to go about rsyncing the n newest files in a directory to a remote server?

Solution 1

The easiest way is to run zsh instead of bash.

rsync -a -- /path/to/directory/*(om[1,42]) remote-server:

In the parentheses, om orders files by reverse modification time (i.e. by increasing age), and [1,42] selects the first 42 matches.

If you want only regular files and not directories, add a . after the opening parenthesis. For more possibilities, look under “glob qualifiers” in the zshexpn man page.

If you want to match files in the directory tree rooted at /path/to/directory, rather than just inside that directory, use /path/to/directory/**/*(.om[1,42]). This will send all the files to the same directory on the target, though.

Solution 2

Assuming you want to send files from the current working directory:

rsync `ls -tp | grep -v / | head -n <n>` <destination> <options>

will do the trick. For example:

rsync `ls -tp | grep -v / | head -n 10` user@host:/dest/dir/ --progress --compress

This will give an error if there are no files to be found in the current working directory, or if any of the top files contain spaces or other special characters.

The ` characters around ls -tp | grep -v / | head -n <n> tell bash to run the commands and replaced them with the resulting file list as a space separated list. The -t option tells ls to sort by timestamp, the -p tells it to add a / after directory names and the grep part screens out lines ending / so you don't end up sending directories over. Add -c to the ls options if you want the newest files to be judged by creation time instead of modification time (though note that some programs will remove and replace files instead of updating them so ctime and mtime can be the same even though a file seems to have been around longer).

I'll not claim it is without doubt the easiest way but it would be the way I'd first think of.

---

(mwm's sub-question-in-a-comment)

what about the other way around? server to local?

That would be a little more complex as I don't think you can have multiple remote file sources on the rsync command line like you can local ones. You could run rsync on the server via over ssh, opening a tunnel back to your local sshd (with the same ssh invocation you run rsync through) to receive the connection, essentially used the same command from the other side.

Comments

[QuickishFM]

If you want to update, then rsync has an --update switch. You could also use find with head if you want the N newest files, and pipe this into rsync. You may have to use xargs but I don't know for sure.

[Gilles 'SO- stop being evil']

Do not use this command unless you know that the file names cannot contain any “special” character (whitespace, \[?*, control characters or non-ASCII characters).

[c 2]

FYI, I had to add single quote around '*(om[1,42])' to make it work. I was copying files from remote server to local though.

[Gilles 'SO- stop being evil']

@c2 The quotes cause the wildcards to be expanded on the remote machine. They are needed on a remote path (and will work only if your login shell on the remote machine in zsh), and would not work on a local path.

[mwm]

what about the other way around? server to local?

[David Spillett]

@mwm - more faf, but possible. See note added to answer.

[gaoithe]

ls -tr sorts by oldest first so you should do ls -tr | tail -n 10 to get to 10 newest files.

[gaoithe]

And when you use --files-from you must still specify source and destination so full command should be ls -tr |tail -n 10 |rsync --files-from=- . user@host:/dest/dir/