Replacing directory variables with sed

genderbender · 02-08-2010, 05:40 AM

Hi, this needs to be done with sed and ideally needs to do this without creating temp files.

TMPDIR="/whatever/you/want/"

This directory eventually gets some tar files written to it, these tar's are then stripped of their directories, I also need to remove whatever the $TMPDIR is.

e.g

/whatever/you/want/importantfiles.tar.gz

becomes importantfiles

I can strip everything ($TMPDIR/*.tar.gz | sed 's/.tar.gz//g') but the directory variable - any ideas? I know this is a real challenge and I'm upto using alternatives to sed (e.g tr or awk or something similar, not perl).

Thanks!

druuna · 02-08-2010, 06:31 AM

Hi,

If the creation of tempfiles is not an option, you should use bash internals (even sed -i s/x/y/g file creates a temp file).

Have a look at this:

Code:

$ xxx="/whatever/you/want/importantfiles.tar.gz"
$ echo ${xxx}
/whatever/you/want/importantfiles.tar.gz

$ echo ${xxx%%.*}
/whatever/you/want/importantfiles

$ echo ${xxx##*/}
importantfiles.tar.gz

The first bold part strips the .tar.gz part, the second one strips the directory part.

Combining the 2 will leave you with the stripped filename:

Code:

$ xxx="/whatever/you/want/importantfiles.tar.gz"
$ yyy="${xxx%%.*}"
$ zzz="${yyy##*/}"
$ echo ${zzz}
importantfiles

For more information about this have a look at the bash man page (the Parameter Expansion section).

Hope this helps.

chrism01 · 02-08-2010, 06:30 PM

How about basename http://linux.die.net/man/1/basename

tuxdev · 02-08-2010, 06:37 PM

Quote:

Have a look at this:

PE is indeed the best method, but I'd swap the two, so that the basename-equivalent happens first. Consider what happens if

Code:

xxx="/whatever/you.want/importantfiles.tar.gz"

genderbender · 02-08-2010, 08:34 PM

I like the look of the basename stuff, it's for a script and I'd like to allow the users to specify their own temp directories, so although having a.directory could be problematic, it's the most elegant solution and problems could be put down to user error (I'll write some error checking code).

Actually, I have a small confession to make; the reason I was getting directories in the first place was due to bad coding on my part. I did 'ls -a $DIR/*' and was concerned when the dir's were displayed when really the quickest solution would just be to do 'ls -a $DIR' and then my sed statement to remove the tar.gz. (it's just a standard for loop where each tar has a different action).

Thanks anyway, this basename stuff will probably be used elsewhere as I have numerous similar lines which aren't using ls.

ghostdog74 · 02-08-2010, 09:14 PM

Quote:

Originally Posted by chrism01

How about basename http://linux.die.net/man/1/basename

if OP has only a few files , then its alright. But if going over many files, using shell internals instead of calling basename will be more efficient.

tuxdev · 02-08-2010, 11:38 PM

Quote:

Actually, I have a small confession to make; the reason I was getting directories in the first place was due to bad coding on my part. I did 'ls -a $DIR/*' and was concerned when the dir's were displayed when really the quickest solution would just be to do 'ls -a $DIR' and then my sed statement to remove the tar.gz. (it's just a standard for loop where each tar has a different action).

Are you actually trying to parse 'ls'? If so, that's your biggest problem right there.
http://mywiki.wooledge.org/ParsingLs
Just use a glob:

Code:

for file in "$dir"/* ; do
   name="${file##*/}"
   name="${name%%.*}"
   echo "$name"
done

As for temporary directories, TMPDIR is the standard variable for the top-level temporary directory (usually /tmp). Well-behaved programs should pay attention to TMPDIR.