To delete files older than 5 days from a directory, we can use command.

find /directory -type f -mtime +5 -delete

But, in my case, I want to delete only those files having 'YYYY-MM-DD' in their names and which are older than 5 days.

Below are some example of filenames:


Only files having name in format 'YYYY-MM-DD' and older than 5 days should get deleted.

How to match filenames in find command using regex ??

  • 4
    What version of find do you have? GNU and BSD finds at least supports regex matching with -regex – Eric Renouf Dec 7 '16 at 14:19
  • test this find -type f -mtime +5 -regextype "egrep" -regex '.*[[:digit:]]{4}-[[:digit:]]{2}-[[:digit:]]{2}.* – Baba Dec 7 '16 at 14:26
  • 1
    @Babyy you should write that up as an answer with some exposition – Eric Renouf Dec 7 '16 at 16:02
  • @EricRenouf my answer is not e prefect! because my regex match with dddd-dd-dddddd* or 2016-99-44 or ... ; if you can edit and resolve this problem post it for answer – Baba Dec 7 '16 at 19:38
  • Has the literal date representation found to be 5 or more days old, or does the mtime of the file matter? – user unknown Mar 17 '18 at 18:06

Using GNU find:

find . -type f -mtime +5 \
    -regextype egrep -regex '.*[0-9]{4}-[0-9]{2}-[0-9]{2}[^/]*$' \

The regular expression will match any string in the basename of a pathname that contains a date on the form YYYY-MM-DD. Note that we may also match XXYYYY-MM-DDZZ where XX and ZZ are some other characters.

The [^/]*$ at the end makes sure that we're actually matching the expression against the basename of the current pathname, and means "no / for the rest of the string, please".

Using a shell wildcard pattern instead (easier to maintain):

find . -type f -mtime +5 \
    -name '*[0-9][0-9][0-9][0-9]-[01][0-9]-[0-3][0-9]*' \

Note that -mtime +5 is for files whose age as an integer number of days is strictly greater than 5, so 6 days and over. For files 5 days old or over, you'd need -mtime +4.

| improve this answer | |

Assuming you are working with GNU tools, this should prevent the false-positives spoken of by Babyy:

find /directory -type f -mtime +5 -print | while read f; do
    n=`basename "$f"`
    d=`expr "X$n" : 'X.*\([0-9]\{4,\}-[0-9]\{2\}-[0-9]\{2,\}\)'` # (1)
    test -n "$d" || continue                                     # (2)
    date -d "$d" >/dev/null 2>&1 && rm "$f"                      # (3)


  • (1) assigns the date part of the file name to d. It includes leading and trailing digits. If there is no date part, then the empty string is assigned to d.
  • (2) will skip to the next iteration if "$d" is the empty string. This is needed to prevent (3) from removing the file, since date -d will succeed if given the empty string as its argument. (This seems like a bug in GNU date, but maybe that behavior was intended.)
  • (3) will remove the file if "$d" contains a valid date. Note that date -d will succeed if the year part of the date contains more than 4 digits, as long as the month and day parts are valid. Now that's really future-proofing your code!
| improve this answer | |

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.