while loop compare timestamp

Question

I want to generate a series of timestamp and compare with the max limit. I wanted to use the while loop to check this.

Here is my code.

start_ts='2021-08-07 15:00:00'
end_ts='2021-08-10 05:00:00'
while [ "$start_ts" > "$end_ts" ]
do 
  echo $start_ts
    start_ts=$(date -ud "$start_ts +0 + 30 minutes" +'%Y-%m-%d %H:%M:%S')
done

But this loop is going and going, not stopping at $end_ts value.

Am I missing something?

Answer 1

When used with [ ... ], the unquoted > is a redirection operator. This means that you now probably have a file named 2021-08-10 05:00:00 in your current directory, created by redirecting the output of the [ command to file.

The command

[ "$start_ts" > "$end_ts" ]

may also be written

[ "$start_ts" ] >"$end_ts"

(see e.g. How is this command legal? "> file1 < file2 cat"). This is a test for whether the string "$start_ts" is non-empty, and the output of that test (which is nothing because [ does not output anything) is redirected to the file $end_ts.

When used with [[ ... ]], the > is a string comparison operator that tests whether two strings are sorted in a particular way. This is likely what you want to use, but to get the logic right you'd use <, or swap the arguments around >, or use until instead of while.

---

Another approach would be to skip the string comparisons completely and instead use integer comparisons of Unix timestamps:

start_ts=$( date -d '2021-08-07 15:00:00' +%s )
end_ts=$( date -d '2021-08-10 05:00:00' +%s )
while [ "$start_ts" -lt "$end_ts" ]; do 
    printf '%(%Y-%m-%d %H:%M:%S)T\n' "$start_ts"
    start_ts=$(( start_ts + 1800 ))
done

Since there is no call to GNU date inside the loop, this runs many times faster than the original code (about 50 times faster on my system).