Maybe something like:

awk '
  function isleap(y) {
    return y % 4 == 0 && (y % 100 != 0 || y % 400 == 0)
  }
  $2 == 3 && $3 == 1 && isleap($1) && last_day != 29 {
    print $1, 2, 29, (last_data + $4) / 2
  }
  {print; last_day = $3; last_data = $4}' file

Abstract

The correct code should be:

#!/bin/sh
[ "$#" -eq 0 ] && echo "Usage: $0 command [args]" && exit 1
[ "$(date -d tomorrow +'%d')" = 01 ] || exit 0
exec "$@"

Call this script end_of_month.sh and the call in cron is simply:

00 12 28-31 * * /path/to/script/end_of_month.sh command

That would run the script end_of_month (which internally will check that the day is the last day of the month) only on the days 28, 29, 30 and 31. There is no need to check for end of month on any other day.

Old post.

That is a quote from the book "Linux Command Line and Shell Scripting Bible" by Richard Blum, Christine Bresnahan pp 442, Third Edition, John Wiley & Sons ©2015.

Yes, that is what it says, but that is wrong/incomplete:

• Missing a closing fi.
• Needs space between [ and the following `.
• It is strongly recommended to use $(…) instead of `…`.
• It is important that you use quotes around expansions like "$(…)"
• There is an additional ; after then

How do I know? (well, by experience ☺ ) but you can try Shellcheck. Paste the code from the book (after the asterisks) and it will show you the errors listed above plus a "missing shebang". An script without any errors in Shellcheck is this:

#!/bin/sh if [ "$(date +%d -d tomorrow)" = 01 ] ; then script.sh; fi

That site works because what was written is "shell code". That is a syntax that works in many shells.

Some issues that shellcheck doesn't mention are:

• It is assuming that the date command is the GNU date version. The one with a -d option that accepts tomorrow as a value (busybox has a -d option but doesn't understand tomorrow and BSD has a -d option but is not related to "display" of time).

• It is better to set the format after all the options date -d tomorrow +'%d'.

• The cron start time is always in local time, that may make one job start 1 hour earlier of later than an exact day count if the DST (daylight saving time) got set or unset.

What we got done is a shell script which could be called with cron. We can further modify the script to accept arguments of the program or command to execute, like this (finally, the correct code):

#!/bin/sh
[ "$#" -eq 0 ] && echo "Usage: $0 command [args]" && exit 1
[ "$(date -d tomorrow +'%d')" = 01 ] || exit 0
exec "$@"

Call this script end_of_month.sh and the call in cron is simply:

00 12 28-31 * * /path/to/script/end_of_month.sh command

That would run the script end_of_month (which internally will check that the day is the last day of the month) only on the days 28, 29, 30 and 31. There is no need to check for end of month on any other day.

Make sure the correct path is included. The PATH inside cron will not (not likely) be the same as the user PATH.

Note that there is one end of month script tested (as indicated below) that could call many other utilities or scripts.

This will also avoid the additional problem that cron generates with the full command line:

• Cron splits the command line on any % even if quoted either with ' or " (only a \ works here). That is a common way in which cron jobs fail.

You can test if end_of_month.sh script works correctly on some date (without waiting to the end of the month to discover it doesn't work) by testing it with faketime:

$ faketime 2018/10/31 ./end_of_month echo "Command will be executed...."
Command will be executed....