Working with dates

Question

I have a task to do for which i am able to think of logic but unable to implement in Unix.

I have a directory in which its sub-directories are having there naming pattern as 'mmddyy'. So my task here is to delete all the directories older than X days. The Date-stamp in question is not Unix Timestamp but the Date-stamp as resembeled by directory name.

My Approach:

1. Get list of dir name in array
2. If (SYSDATE - array > 30) Then Remove Else Continue.

How do we solve using Unix?

Answer 1

reformat the date details in the file into yy-mm-dd and then use date and the seconds since epoch

X=30 - 30 days ago

epc=$(date -d -"$X"days +%s)

epc is the reference epoch string to check against

Then process the date in the file name

example date in file name

 dat="010188" 
 dat1=awk '{print substr($0,5,2)"-"substr($0,3,2)"-"substr($0,1,2)}' <<< $dat
 refepc=$(date -d "$dat1" +%s)

refepc is then your reference epoch that can be checked against epc

Answer 2

then=$( date -d '30 days ago' +%y%m%d )

for d in *; do
    if [[ -d "$d" ]] && [[ "$d" =~ ^([0-9][0-9])([0-9][0-9])([0-9][0-9])$ ]]; then
        datestring="${BASH_REMATCH[3]}${BASH_REMATCH[1]}${BASH_REMATCH[2]}"

        if [[ "$datestring" < "$then" ]]; then
            echo rm -rf "$d"
        fi
    fi
done

This bash script would first figure out the date of 30 days ago using GNU date and storing it in the YYMMDD format in $then.

It then iterates over the names found in the current directory. If any of the found names is a directory that matches the given regular expression, it picks out the parts of the directory name and arranges them in the YYMMDD format in $datestring.

If $datestring is lexicographically less than $then, then this directory should be deleted. The echo prevents this from actually happening, so run it first a few times to make sure that it works before removing it.

Testing:

$ mkdir {01..12}0117  # one directory for the 1st of every month this year
$ bash ./script.sh
rm -rf 010117
rm -rf 020117
rm -rf 030117
rm -rf 040117
rm -rf 050117
rm -rf 060117
rm -rf 070117
rm -rf 080117

(today's date is 170922 and $then has the value 170823)

This would obviously fail for directory names containing dates in the last century (these would not be deleted).

Answer 3

With zsh:

#! /usr/bin/env zsh
days=${1-30}

zmodload zsh/datetime || exit

strftime -s cutoff %Y%m%d $((EPOCHSECONDS - days * 24 * 60 * 60))

old_enough() {
  year=$REPLY[5,6]
  ((year += 1900 + (year < 70) * 100))
  [[ $year$REPLY[1,4] < $cutoff ]]
}

set -o extendedglob
echo rm -rf [0-9](#c6)(/+old_enough)

Remove the echo or pipe to sh if happy.

With ksh93, an equivalent could be:

#! /usr/bin/env ksh
days=${1-30}

cutoff=$(printf '%(%Y%m%d)T' "$days days ago")

old_enough() {
  year=${1:4}
  ((year += 1900 + (year < 70) * 100))
  [[ $year${1:0:4} < $cutoff ]]
}
dirs=()
for f in {6}([0-9]); do
  [ -d "$f" ] && [ ! -L "$f" ] && old_enough "$f" && dirs+=("$f")
done
if ((${#dirs[@]})); then
  echo rm -rf "${dirs[@]}"
else
  echo >&2 "No match"
  exit 1
fi

With bash, an equivalent could be:

#! /usr/bin/env bash
days=${1-30}

printf -v now '%(%s)T' -1
printf -v cutoff '%(%Y%m%d)T' "$((now - days * 24 * 60 * 60))"

old_enough() {
  year=${1:4}
  year=${year#0} # work around bug where bash complains about 08 and 09
  ((year += 1900 + (year < 70) * 100))
  [[ $year${1:0:4} < $cutoff ]]
}
shopt -s nullglob
dirs=()
for f in [0-9][0-9][0-9][0-9][0-9][0-9]; do
  [ -d "$f" ] && [ ! -L "$f" ] && old_enough "$f" && dirs+=("$f")
done
if ((${#dirs[@]})); then
  echo rm -rf "${dirs[@]}"
else
  echo >&2 "No match"
  exit 1
fi

Years above 70 are assumed to be from the last century (123199 means 1999-12-31, 092217 means 2017-09-22).

If you have neither of those shells and are on a Unix (not GNU (UNG)) system, then your best bet may be to resort to perl:

#! /usr/bin/env perl
use POSIX;
$cutoff = time - ($ARGV[0] // 30) * 24 * 60 * 60;
$, = " "; $\ = "\n";
if (@dirs = grep {
    /^(\d\d)(\d\d)(\d\d)$/ && ! -l && -d && 
    mktime(0,0,0,$2,$1-1,$3+100*($3<70)) < $cutoff} <*>) {
  print "rm -rf", @dirs;
} else {
  die "No match";
}