I have a lot of daily backup archives. To manage disk usage, I need a bash script that will delete all files older than 1 month, but keep all files created on Mondays, even if they are older than 1 month.
For example, this will delete all files last modified more than 30 days ago:
find /path/to/files* -type f -mtime +30 -delete
But I don't really know how to keep files created on Mondays.
Slightly simpler and more cautious version of @JoSo's answer:
find /path/to/files -type f -mtime +30 \
-exec sh -c 'test $(date +%a -r "$1") = Mon || echo rm "$1"' -- {} \;
The differences:
- Using
date -r
to get the last modification date of a file directly
- Using
%a
to work with more comprehensible weekday names
- Just echo the
rm "$1"
first to review what will be deleted. If looks good, then either stick | sh
at the end to really execute, or remove the echo
However, @JoSo is right to point out that date +%a
is locale dependent, so these versions would be indeed safer:
find /path/to/files -type f -mtime +30 \
-exec sh -c 'test $(date +%u -r "$1") = 1 || echo rm "$1"' -- {} \;
find /path/to/files -type f -mtime +30 \
-exec sh -c 'test $(LC_TIME=C date +%a -r "$1") = Mon || echo rm "$1"' -- {} \;
As find
to my knowledge has no weekday check, you need to call an external program.
find /path/to/files -type f -mtime +30 \
-exec sh -c \
'[ "$(date +%u -d @"$(stat -c %Y "$1")")" != 1 ] && rm "$1"' -- {} \;
Update: Using the -r
switch to date
(Kudos to Janos) and only testing, not deleting inside the shell command probably yields the cleanest possible version:
find /path/to/files -type f -mtime +30 \
-exec sh -c 'test "$(date +%u -r "$1")" != 1' -- {} \; \
-print # or -delete instead
you can use
stat -c %y yourfile
to get the date the file was created.
Then extract the date with cut to have a var like myvar=yyyymmdd
and finally use
date +%u --d $myvar
date +%u --d 20130822
it will return the day of week, if it returns 1 it was monday.
In my case it return 4 because 22/08/2013 was a Thursday
Edit : if you can get all these working as a simple command line as Jo So suggest, it's better !
What about avoiding subprocesses in loop:
find /path/to/files -type f -mtime +30 -printf '%Ta %p\n' \
| grep -v ^Mon | cut -c5- | tr "\n" "\0" | xargs -0 rm -v