Tar compression without directory structure

directoryfilenamestar

Under my current directory, I have two sub-directories:

dir_1/
   - file1.png
   - file2.png
   ...
   - fileN.png

dir_2/
   - fileA.txt
   - ...
   - fileZ.txt

When I tar compress the two directories by :

tar -cvzf result.tar.gz dir_1/ dir_2/

I got result.tar.gz but it maintains the directory structure. I mean when I extract the result.tar.gz, I got dir_1 & dir_2 again.

How can I tar compress so that the directory structure is not remained, which means when I extract the tar.gz file, I only get files

result/
   file1.png 
   ... 
   fileN.png 
   fileA.txt
   ...
   fileZ.txt

Best Answer

I think you can do this with the -C option.

From the tar man page:

-C directory, --cd directory, --directory directory
    In c and r mode, this changes the directory before adding the following files.
    In x mode, change directories after opening the archive but before extracting
    entries from the archive.

This means that you should be able to run

tar cvzf result.tar.gz -C /path/to/dir1/ . -C /path/to/dir2/ .

to achieve what you want.

regex using just tar?

method #1

I researched this quite a bit and came up empty. The limiting factor would be that when tar is performing it's excludes, the trailing slash (/) that shows up with directories is not part of the equation when tar is performing its pattern match.

Here's an example:

$ tar -v --create --file=do.tar.gz --auto-compress --no-recursion --exclude={'.','..','.*/'} .*
.qalculate/
.qt/
.qterm/
.qtoctave/
.RapidSVN
.RData
.Rhistory

This variant includes an exclude of .*/ and you can see with the verbose switch turned on to tar, -v, that these directories are passing through that exclude.

method #2

I thought maybe the switches --no-wildcards-match-slash or --wildcards-match-slash would relax the greediness of the .*/ but this had no effect either.

Taking the slash out of the exclude, .* wasn't an option either since that would tell tar to exclude all the dotfiles and dotdirectories:

$ tar -v --create --file=do.tar.gz --auto-compress --no-recursion --exclude={'.','..','.*'} .*
$

method #3 (Ugly!)

So the only other alternative I can conceive is to provide a list of files to tar. Something like this:

$ tar -v --create --file=do.tar.gz --auto-compress --no-recursion --wildcards-match-slash --exclude={'.','..','.*/'} $(ls -aF | grep -E "^\..*[^/]$")
.RapidSVN
.RData
.Rhistory

This approach has issues if the number of files exceeds the maximum amount of space for passing arguments to a command would be one glaring issue. The other is that it's ugly and overly complex.

so what did we learn?

There doesn't appear to be a straight-forward and elegant way to acomplish this using tar & regular expressions. So as to @terdon's comment, find ... | tar ... is really the more appropriate way to do this.

Tar – How to List Files Only with Tar

I don't see a way to do it from the man page, but you can always filter the results. The following assumes no newlines in your file names:

tar tzf your_archive | awk -F/ '{ if($NF != "") print $NF }'

How it works

By setting the field separator to /, the last field awk knows about ($NF) is either the file name if it's processing a file name or empty if it's processing a directory name (tar adds a trailing slash to directory names). So, we're basically telling awk to print the last field if it's not empty.