I have a folder containing approximately 320116 .pdb.gz files. I want to uncompress them all. If I use gunzip *.gz it gives me an error i.e. argument list too long. The folder is about 2GB. Please give me an appropriate suggestion.

  • If you will have to work on this directory structure on the long term, split this directory in many ones. For example based on files modification time or on files name. – dan Mar 19 '15 at 7:35
  • Yes I do have to work on long term. they have been extracted now I want to subdivide and classify them into three folders on the basis of their names. Is there a shell script to do so? – Lily Sharpton Mar 19 '15 at 10:21
  • I suggest you to search for similar questions there. If you don't find one to suit your needs, ask your own new question. – dan Mar 19 '15 at 12:11
find . -name '*.pdb.gz' -exec gunzip {} +

-exec gunzip {} + will provide gunzip with many but not too many file names on its command line. This is more efficient than -exec gunzip {} \; which starts a new gunzip process for each and every file.

  • 3
    One find, fewer gunzip! – dan Mar 19 '15 at 7:32
  • 2
    Note that the "+" is a GNUism and so will not work on non-GNU systems like *BSD. – Reinstate Monica - M. Schröder Mar 20 '15 at 8:03
  • 3
    Later versions of BSD find allow the "+" notation. See, for example, the find man page for BSD 10.1. Also applies to OS X (10.9 and later at least, maybe earlier). – plasma Mar 25 '15 at 19:20

Whenever you get "argument list too long" errors you can work around it by invoking the desired command multiple times, each time with a subset of the arguments you want to use. xargs is a tool that helps you do that automatically.

find . -type f -a -name \*.pdb.gz -print0 | xargs -0 gunzip
  • doesn't this have the same inefficiency as -execdir gunzip "{}" \; i.e. that xargs will invoke gunzip separately for each file? That is my reading of man page. – gogoud Mar 19 '15 at 7:45
  • 5
    No, xargs will stuff as many filenames as will fit on the gunzip command line. Try it! echo a b c d e f | xargs echo only invokes echo once with all 6 arguments so you see one line of output (pretty useless command to execute though!!!!) while if you force xargs to only supply up to 3 arguments per invocation of the command using echo a b c d e f | xargs -n 3 echo then you get 2 lines of output. – Celada Mar 19 '15 at 7:49
  • 4
    Another advantage of using xargs is that, with the -P option, you can run multiple gunzip processes in parallel, which (depending on the exact parameters of your system) may go faster. – psmears Mar 19 '15 at 14:44
  • thanks for the pointer to -P, @psmears. Now I learned something too! – Celada Mar 20 '15 at 1:00

I think this should work, it passes the path/name of each file individually to gunzip for processing:

find /my/dir -name "*.pdb.gz" -execdir gunzip "{}" \;
  • 1
    That will execute gunzip once per file. See John1024's answer for a slightly different way that avoids that inefficiency. – Celada Mar 19 '15 at 7:29
  • @Celada This was deliberate; my concern was that using + might again lead to an error message because of overloading gunzip. If John1024's method works, it is technically more efficient, but mine should work if his doesn't. – gogoud Mar 19 '15 at 7:34
  • 1
    find with + and xargs are expressly designer with exactly that issue in mind. They will always supply as many arguments as they can, while not exceeding the operating system's limit. Because, by the way, it is an operating system limit, nothing to do with gunzip. – Celada Mar 19 '15 at 7:37
  • 1
    @Celada ok thanks for that info, so presumably with '+' gunzip might get invoked more than once, but less than 320,000 times? – gogoud Mar 19 '15 at 7:38
  • 1
    correct‍‍‍‍‍‍‍. – Celada Mar 19 '15 at 7:55

Try this way:

find . -name '*.gz' -exec gunzip {} \;
  • 3
    That will execute gunzip once per file. See John1024's answer for a slightly different way that avoids that inefficiency. – Celada Mar 19 '15 at 7:27
  • Make sure to escape the * in *.gz... – user253751 Mar 19 '15 at 8:15

If you have a multi-core machine you will probably see that using gunzip is not going to max out the capabilities of your machine. For that you would need to run multiple gunzips in parallel. To keep track of which are done in which terminal by hand is is cumbersome, but you can easily do that with GNU parallel:

find . -name "*.gz" | parallel -X gunzip {}
  • 1
    Won't that fail because the argument list to parallel is too long? – user253751 Mar 19 '15 at 8:14
  • @immibis Yes, I forgot the original problem, I'll update my post – Anthon Mar 19 '15 at 8:16
  • Won't that still fail because the argument list to find is too long? – user253751 Mar 19 '15 at 8:27
  • 1
    yes but you're passing all the filenames on find's command line. – user253751 Mar 19 '15 at 8:34
  • Seems this is not a good day to answer questions, I forgot to quote the argument to -name – Anthon Mar 19 '15 at 8:49

It is unnecessary to use find for this, since you did not mention subfolders. What you need to do is:

for f in *.gz;do gunzip $f;done
  • 4
    You do need find if you don't want to spawn 320116 gunzip processes, just like this loop does. – John WH Smith Mar 19 '15 at 17:29

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.