Unix & Linux Stack Exchange is a question and answer site for users of Linux, FreeBSD and other Un*x-like operating systems. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I can probably write a shell script to find files only, then pass the list to tar, but I am wondering whether there already is a built-in feature in tar that allows doing just that, in a single command line?

For example, I found the --no-recursion switch, but when I do:

tar --no-recursion -cvf mydir.tar mydir

It only archives the names of the entries in the directory (including subdirectories!), but it doesn't archive any files.

I also tried:

 tar --no-recursion -cvf mydir.tar mydir/*

But while it archives files only, it also archives the names of the subdirectories.

Is there a way to tell tar files only, no directories?

share|improve this question
4  
Just to clarify: do you want to create an archive with "flat" structure (i.e. all files mixed up in one directory)? – rozcietrzewiacz Nov 18 '11 at 14:47
1  
You could create a new directory and find mydir -type f |xargs cp -t tempdir and then tar tempdir. – Kevin Nov 18 '11 at 15:05
    
@rozcietrzewiacz Yes, flat, but only from that directory, not from subdirectories. – ateiob Nov 18 '11 at 15:17
1  
OK, I think I see what you're trying to do. How about find mydir -depth 1 -type f | xargs tar cf mydir.tar – Kevin Nov 18 '11 at 15:30
2  
Ah, spaces. Use find's -exec instead: find mydir -maxdepth 1 -type f -exec tar cvf mydir.tar {} +. The + puts all the files on the same command line like xargs. – Kevin Nov 18 '11 at 15:49
up vote 9 down vote accepted

As camh points out, the previous command had a small problem in that given too many file names, it would execute more than once, with later invocations silently wiping out the previous runs. Since we're not compressing too, we can append instead of overwrite:

find mydir -maxdepth 1 -type f -print0 | xargs -0 tar Avf mydir.tar
find mydir -maxdepth 1 -type f -exec tar Avf mydir.tar {} +

Iocnarz's answer of using tar's --null and -T options works as well. If you have cpio installed, camh's answer using it is also fine. And if you have zsh and don't mind using it for a command, Gilles's answer using a zsh glob (*(.)) seems the most straightforward.


The key was the -maxdepth option. Final answer, dealing with spaces appropriately:

find mydir -maxdepth 1 -type f -print0 | xargs -0 tar cvf mydir.tar

This should also work:

find mydir -maxdepth 1 -type f -exec tar cvf mydir.tar {} +
share|improve this answer
3  
Both of these can result in multiple invocations of tar. Both xargs and find with the + variant has a maximum number of arguments that can be passed to tar. This will result in the second invocation of tar overwriting the output of the first. – camh Apr 6 '12 at 22:30
1  
Expanding on the xargs limit, xargs default to a maximum 128KiB command line. If the file list is larger, you get a second (or more) invocation of the command (tar), leading to silent data loss. You can use -x to force xargs to fail instead of losing data, and while better than a silent data loss bug, it's still not ideal. This sort bug is so dangerous because everything seems OK at first, but as the file list grows over time, you start triggering it and may not notice until you try to restore your backup. Then it's too late. – camh Apr 6 '12 at 22:49
    
@camh You're right, thanks for pointing that out; I've updated to reflect that. – Kevin Apr 7 '12 at 3:22
    
The "final answer" is not correct, if you use "tar c" without -T you might get partial results. – eckes May 31 '13 at 17:27

When you want to use find with tar, the best way is to use cpio instead of tar. cpio can write tar archives and is designed to take the list of files to archive from stdin.

find mydir -type f -maxdepth 1 -print0 | cpio -o -H ustar -0 > mydir.tar

Using find and cpio is a more unix-y approach in that you let find do the file selection with all the power that it has, and let cpio do the archiving. It is worth learning this simple use of cpio, as you find it easy to solve problems you bang your ahead against when trying tar.

share|improve this answer
    
Brilliant! I've used cpio in the distant past, but never knew of the ustar option. – Ian McGowan Feb 23 at 18:07

I'm not sure I understand your requirements. If you want to store the regular files in mydir but not its subdirectories, the easiest way is to use zsh, where matching regular files only is the simple matter of using the . glob qualifier:

tar cf mydir.tar mydir/*(.)
share|improve this answer

You may even use find ... -print0 and tar ... --null directly without using xargs at all.

find . -maxdepth 1 -type f -print0 | tar cvf mydir.tar --null -T -

In the given example, the --no-recursion option to tar is not necessary because only paths of files (and not directories) will be passed from find to tar.

Using the --no-recursion option to tar in the following example, however, prevents tar from double archiving directories. find will do the directory tree recursion instead of tar then.

# compare
find . -print0 | tar cf mydir.tar --null -T -
tar -tf mydir.tar | nl

find . -print0 | tar cf mydir.tar --null --no-recursion -T -
tar -tf mydir.tar | nl
share|improve this answer

As the introductory paragraph in man tar says (last sentence),

The use of a directory name always implies that the subdirectories below should be included in the archive.

Which I understand as a "no" answer to your question.

share|improve this answer
    
Good catch, except that tar now includes numerous exclusion meachisms (e.g. --no-recursion, --exclude-tag, etc.). I am looking into --exclude-tag which looks promising but seems to be the exact opposite of what I am looking for. – ateiob Nov 18 '11 at 15:21
    
I tried this and I saw characters being removed from long pathnames. Using tar with -T and --null as camh proposed avoided this problem. – Paul Brannan May 26 '15 at 19:45

I might have found a solution.

find mydir -type f -printf '%P\0'|tar czvf mydir.tar.gz -C mydir --null -T -

This might be costly a little, but anyway it will work, because this doesn't depend on xargs.

share|improve this answer
star -c -C startdir -find . ! -type d > out.tar

Omit -C startdir and replace . with startdir if it should appear in the archive.

This is the most efficient method based on the features of libfind. Libfind also offers primaries -chown -chgrp -chmod that modify struct stat in place and allow to archive different metadata. This also works in list and extract mode and avoids the need to extract the whole archive in many cases.

share|improve this answer
    
Can you use -find in -copy mode. The man page synopsis suggests you can but I've not been able to make it work (while trying to answer How to copy modified files while preserving folder structure) (with schily-2016-02-10) – Stéphane Chazelas Feb 23 at 17:58
    
I did not test this for a longer time. It may be that there currently is a bug. The man page says that the last argument would be the extraction directory, but when I try this, I get: "Path arguments not yet supported in extract mode". Is this what you get when you try? – schily Feb 23 at 18:39
    
yes. The man page also says you can only have find arguments after -find but if I put the destination directory before -find I also get an error – Stéphane Chazelas Feb 23 at 20:06
    
There is code to switch off the find code in the extract process and the error message is from an incorrect check that expects -c but should also permit -copy. In -copy mode, the find parser should not include the last argument. – schily Feb 23 at 22:35
1  
A final solution now has been published in the schily tools at sourceforge.net/projects/schilytools/files/schily-2016-03-11.tar.bz2 this also fixes a problem from a missing setlocale() in star and this enables -C directrory for star -find. – schily Mar 11 at 10:53

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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