[SOLVED] remove directories that only contain .txt and .log files?

jmark · 08-08-2010, 02:42 AM

I'd like to remove all directories of a certain depth that don't contain .txt or .log files -- is this possible? So far I have:

find ~ -mindepth 3 -maxdepth 4 -type d -exec rm -r '{}' \;

Is it possible to add in "only if the directory doesn't contain .txt and/or .log files"? Or do I have to start learning perl to do that?

For example:

dir 1:
hello.txt
runme.sh

dir 2:
runme.sh
oct12.log

dir 3:
runme.sh

dir 4:
<empty>

It would delete directories 3 & 4

r3sistance · 08-08-2010, 03:26 AM

As a quick method of finding out if a directory contains files that are not just .txt or .log... "ls -a | grep -v .txt | grep -v .log". there might be an easier way to do both in a single grep but -a is important since there might be hidden files to consider, however they may also be common hidden files, if you don't care to bother to check for them or of them then just do a normal ls with the grepping.

a very quick bash script I think would be

Directory = <some directory>
Filecheck = 'ls ${Directory} | grep -v .*\.txt$ | grep -v .*\.log$'
# more advanced version of the greps.. I am still unable to merge these into a single grep =/.

if [ ${Filecheck} = "" ]
then
rm ${Directory}
fi

however I only tentatively know bash scripting so this likely could use a lot of improvement, I am more of a c, php, vb, javascript, java... kinda man.

druuna · 08-08-2010, 03:55 AM

@r3sistance:

Quote:

# more advanced version of the greps.. I am still unable to merge these into a single grep =/.

egrep -v "\.txt$|\.log$" or grep -Ev "\.txt$|\.log$"

The pipe character (|) separates the regular expressions.

Hope this helps.

r3sistance · 08-08-2010, 04:06 AM

Sure, I don't usually do fantastic things with grep myself... but always useful to know. I realize I made another mistake above.

should have been rmdir ${Directory} not rm${Directory} anyways.

David the H. · 08-08-2010, 04:08 AM

You can also use -e to define multiple expressions.

Code:

egrep -v -e ".txt$" -e ".log$"

grail · 08-08-2010, 04:15 AM

How about a while loop:

Code:

while read dir
do
    [[ -n $(ls $dir/*{txt,log} &> /dev/null) ]] || echo rm -r $dir
done< <(find $HOME -mindepth 3 -maxdepth 4 -type d)

jmark · 08-08-2010, 04:47 AM

Wow, that was quick!

I mishmashed everyone's ideas and this seems to work:

while read dir
do
Filecheck=$(ls ${dir} | egrep -e ".txt$" -e ".log$")
if [ -z ${Filecheck} ]
then
echo rm -r ${dir}
fi

done< <(find $HOME -mindepth 3 -maxdepth 4 -type d)

I tried to get this to work in the while loop:
[[ -n $(ls $dir/*{txt,log} &> /dev/null) ]] || echo rm -r $dir
It seemed to pick every directory though, and I don't understand the syntax enough to figure out what the problem was.

Thanks!

jmark · 08-08-2010, 04:54 AM

Doh, Filecheck=$(ls ${dir} | egrep -e ".txt$" -e ".log$") doesn't seem to work when dir has a space in it.

grail · 08-08-2010, 05:09 AM

So quote anything you need to preserve the space, ie "$dir"

The test I wrote had 2 issues (my bad) I redirected all input and should have only been stderror and used the wrong control

Try:

Code:

while read dir
do
    [[ -n $(ls "$dir/*{txt,log}" 2> /dev/null) ]] && echo rm -r "$dir"
done< <(find $HOME -mindepth 3 -maxdepth 4 -type d)

jmark · 08-08-2010, 05:33 AM

Thanks -- I played around with putting quotes, but couldn't find the right place until your suggestion.

Code:

while read dir
do
    [[ -n $(ls "$dir/*{txt,log}" 2> /dev/null) ]] && echo rm -r "$dir"
done< <(find $HOME -mindepth 3 -maxdepth 4 -type d)

Doesn't print anything now -- I'll have to play around with it.

With the quote changes for directories containing spaces, the following works well, thanks!
(I had to subtract 1 from the depth arguments because of the ls in the while loop)

Code:

while read dir
do
Filecheck=$(ls "${dir}" | egrep -e ".txt$" -e ".log$")
if [ -z ${Filecheck} ]
then
echo rm -r "${dir}"
fi
   
done< <(find $HOME -mindepth 2 -maxdepth 3 -type d)

ghostdog74 · 08-08-2010, 06:11 AM

using shell expansion instead of calling ls

Code:

shopt -s nullglob
find ~ -mindepth 3 -maxdepth 4 -type d | while read -r DIR
do
  v=$(echo $DIR/*.txt $DIR/*.log)
  [[ -z $v ]] && rm -fr "$DIR" 
done

David the H. · 08-09-2010, 12:18 AM

Quote:

Originally Posted by jmark

Code:

ls "$dir/*{txt,log}"

Doesn't print anything now -- I'll have to play around with it.

Quoting protects all special characters(1), so your *, {, and } also become literal parts of the string here. So be careful to quote only the parts you need to protect, usually the variable and other literal strings, and not those characters the shell needs to process as commands.

Code:

ls "$dir"/*{txt,log}

(1) There's a difference between double and single quotes here. Double-quotes do not protect $, `, and \, which means you can still use variables, embedded commands, and escape sequences inside them. Single quotes, however, protect *everything* (except for another single quote, of course).

Correct quoting is a very important part of scripting; one small mistake can cost you many hours in debugging a large script. More on it here:
http://www.tldp.org/LDP/abs/html/quoting.html

grail · 08-09-2010, 12:34 AM

Quote:

(I had to subtract 1 from the depth arguments because of the ls in the while loop)

Of course this change would need to be applied to my script as well ... otherwise it will not find the files.

jthill · 08-09-2010, 03:55 AM

What happens if there are deeper subdirectories?

Code:

$ find . -mindepth 3 -iname '*.txt' -o -iname '*.log'
./a/b/c/d/testdata.txt

should directory c be removed or not? For safety, I'm guessing not, though I can read it both ways.

Let's ignore the possibility that directory names might contain newlines or single quotes.

Code:

$ find . -depth -mindepth 3 -type d \! -path '*/*/*/*/*' -o -type f \( -iname '*.txt' -o -iname '*.log' \) |
awk -F/ 'NF!=4 {k=1} NF==4 {if (k==0) print "rm -rf '\''" $0 "'\''";} NF==4 {k=0}'

will print out the rm commands for the depth 3 directories that contain no *.txt or *.log files anywhere underneath.

The find "-depth" option tells find to run depth-first, so it prints everything it finds in a directory before printing that directory. So I'm telling it to print any reason to keep a directory before printing that directory, then telling awk to print directories it doesn't see any reason to keep.