Would like to delete all files in a directory except the 7 latest ones

vbekker · 12-04-2010, 01:25 AM

Hi,
Here is how i am isolating the 7 latest files

Code:

ls -ltr | tail -7

is there a way to give that negated list to the rm command?
or should i be using the find command with an exec {}?
Thank YOu

barriehie · 12-04-2010, 01:40 AM

Quote:

Originally Posted by vbekker

Hi,
Here is how i am isolating the 7 latest files

Code:

ls -ltr | tail -7

is there a way to give that negated list to the rm command?
or should i be using the find command with an exec {}?
Thank YOu

Here's a way:

Code:

#
#  Leave 7 session files
#
# Specify the target directory and file names to operate on.
target_files=/home/barrie/.nautilus/saved-session-*

# Calculate the total number of files matching the target criteria. 
total_files=$(ls -t1 $target_files | wc --lines)

# Specify the number of files to retain.
retained_files=7

# If there are surplus files, delete them.
if [ $total_files -gt $retained_files ]
  then
  rm $(ls -t1 $target_files | tail --lines=$((total_files-retained_files)))
fi

gd2shoe · 12-04-2010, 03:10 AM

I don't think your rm line will work right. You should probably use xargs with the -d '\n' option. That will help when dealing with spaces in file names.

Additionally, directories included in $target_files cause things to go screwy. We don't know for sure that he's dealing with session information or log files yet. He might have errant subdirectories that need avoiding.

I suggest something closer to this:

Code:

#
#  Leave 7 newest files
#
# Specify the target directory and file names to operate on.
target_files=~/target_dir/*

#remove directories(etc) from target
target_files=$(ls -t1d --file-type $target_files | grep -v '[/=>@|]$')  #there's probably a better way

# Calculate the total number of files matching the target criteria. 
total_files=$(ls -t1 $target_files | wc --lines)

# Specify the number of files to retain.
retained_files=7

# If there are surplus files, delete them.
if [ $total_files -gt $retained_files ]
  then
  ls -t1d $target_files | tail --lines=$((total_files-retained_files)) | xargs -d '\n' rm
fi

(may still contain bugs)

druuna · 12-04-2010, 03:27 AM

Hi,

Assuming there are no directories present in the output (they won't be deleted, but an error will be shown), this is probably the simplest way:

Code:

ls -Qtr | tail -7 | xargs rm

(misread the original question.... Sorry!) See post #7

The -l isn't needed (it will only complicate things), the -Q puts double quotes around the found entries (takes care of spaces and the lot).

Hope this helps.

gd2shoe · 12-04-2010, 03:33 AM

Thanks for -Q. That's nifty.

Note, though, that he wants to delete all the files except the most recent 7. That's what the jumping through hoops and wc are about.

edit: also, that's not a -l, that's a -1. It's barriehie's way of making absolutely certain each file is on a separate line when it hits wc.

grail · 12-04-2010, 03:51 AM

Here's 2 cents worth:

Code:

target_dir=~/target_dir

(( NUM_FILES = $(find $target_dir -maxdepth 1 -mindepth 1 -type f | wc -l) - 7 ))

if (( NUM_FILES > 0 ))
then
    find $target_dir -maxdepth 1 -mindepth 1 -type f -exec ls -rt {} \; 2> /dev/null | head -n $NUM_FILES | xargs rm
fi

Probably needs tweaking but you get the idea.

druuna · 12-04-2010, 04:14 AM

Hi again,

Sorry for the incorrect answer in post #4!

Here's a one-liner that should work:

Code:

ls -1Qtr | head -$(($(ls -1 | wc -l)-7)) | xargs rm

Hope this helps.

gd2shoe · 12-04-2010, 05:05 AM

Neat. A little buggy, but very compact. (Fewer than 7 files or the presence of subdirectories cause problems.)

Here's my attempt without wc:

Code:

target=~/target/*
retained=2
ls -1tQd --file-type $target | grep -v '[/=>@|]$' | (while test $((retained--)) -ne 0; do read; done ; xargs -r rm)

I'd have preferred to use "head -$retained", but head consumes all input.

Of course, mine has its own drawbacks. Setting retained to anything other than an integer causes all files to be deleted.

barriehie · 12-04-2010, 10:03 AM

In regards to the 7 latest files are they generated any way such that find can be used? Only asking because find may have fewer CLI contortions than using ls.

GrapefruiTgirl · 12-04-2010, 10:07 AM

I haven't read every word of every post in this thread, but wanted to point the OP to this other thread:
http://www.linuxquestions.org/questi...-files-839315/
where it appears that the OP there wanted to do the same (or a very similar) thing. Maybe you can learn something there that will help, and apply it here.

grail · 12-04-2010, 10:40 AM

I like it

Be interesting to see how the running times would be affected if you put the rm into your awk ... but nice all the same.

GrapefruiTgirl · 12-04-2010, 10:46 AM

Quote:

Originally Posted by grail

I like it

....

If you're referring to my little script - well, I suppose you should like it a little bit at least- it exists thanks to some code that you posted priorly somewhere, using similar `find -printf` functionality. So, thank you!

vbekker · 12-04-2010, 01:14 PM

Thank you all very much for this....Just a follow up question.

Code:

tail +10 Output every line from line 10 onwards.

This is true in UNIX (on Solaris, korn shell), how would i achieve the same functionality in Linux bash?
If might be alot simpler to sort the list of files by time (ls -1t) and then use the tail +7 command to show all lines that are not the first 7 and do manipulation (rm) on those lines. Please correct me if i wrong.
Thank You

vbekker · 12-04-2010, 04:02 PM

this might be the simplest solution, what do you think?

Code:

rm -rf $(ls -1t | tail -n +7)

GrapefruiTgirl · 12-04-2010, 04:35 PM

Quote:

Originally Posted by vbekker

this might be the simplest solution, what do you think?

Code:

rm -rf $(ls -1t | tail -n +7)

One problem with the above, is that filenames containing spaces cause problems. To demonstrate:

Code:

root@reactor: ls -1t
crap
hello there
root@reactor: echo rm -rf $(ls -1t)
rm -rf crap hello there
root@reactor:

So you see, `rm` will try to delete 3 files, 2 of which do not exist; the result: the file with spaces in the name, didn't get deleted.