RegEx for PostgreSQL 'interval' function - regex

Question

I need to implement regex validation for value that will be used in my server side to get data where certain timestamp is older (smaller) than now() - interval 'myValue'.
pSQL interval function is explained here, and in short it can have values like 2 days,3 years,12 hours, but also you can nest more different values like 2 days 6 hours 30 minutes etc.
I currently have a regex /^\d+\s(seconds?|minutes?|hours?|days?|weeks?|months?|years?)$/i that accepts only one value (e.g. 2 days), but can't figure out how to allow multiple values, and set a rule that a certain string from this group can only be repeated once or not at all.
This regex /^\d+\s(seconds?|minutes?|hours?|days?|weeks?|months?|years?)(\s\d+\s(seconds?|minutes?|hours?|days?|weeks?|months?|years?))*$/i allows nesting but also allows repetition of values e.g. 2 days 12 hours 6 hours 2 minutes which will result in a fatal error in pSQL query.
I tried restricting repetition of values in this group with \1 and {0,1} combination of regex operators but I just can't nail it precisely enough.
NOTE: Regex is unfortunately only way I can validate this value, since I don't have access to server-side controller which receives this value nor do I have access to client-side frontend of this form. I can't just throw exceptions or skip query because it is a part of important cron-job, and must be stable at all time.
(All I have access to is json schema of this value, and therefore can only define regex pattern for it)
Any help is appreciated, thanks.

Answer 1

You can use
^(?!.*(second|minute|hour|day|week|month|year).*\1)\d+\s+(?:second|minute|hour|day|week|month|year)s?(?:\s+\d+\s+(?:second|minute|hour|day|week|month|year)s?)*$
See the regex demo
Details
^ - start of string
(?!.*(second|minute|hour|day|week|month|year).*\1) - no second, minute, hour day, week, monthoryear` string repetition allowed in the whole string
\d+\s+(?:second|minute|hour|day|week|month|year)s? - 1 or more digits, one or more whitespaces, then either second, minute, hour, day, week, month or year, and then an optional s letter
(?:\s+\d+\s+(?:second|minute|hour|day|week|month|year)s?)* - zero or more repetition of one or more whitespaces followed with the pattern described above
$ - end of string.

Answer 2

Forget it. The only complete documentation of the supported values for interval is the implementation (the guts are in ParseDateTime).
Consider these:
SELECT INTERVAL '12 00:12:00';
interval
══════════════════
12 days 00:12:00
(1 row)
SELECT INTERVAL '12 d 12 mins';
interval
══════════════════
12 days 00:12:00
(1 row)
SELECT INTERVAL '3-2';
interval
════════════════
3 years 2 mons
(1 row)
What I would do in your place is to write a function that casts the string to interval and catches and reports an error:
CREATE FUNCTION interval_ok(text) RETURNS boolean
LANGUAGE plpgsql AS
$$BEGIN
PERFORM CAST ($1 AS interval);
RETURN TRUE;
EXCEPTION
WHEN invalid_datetime_format THEN
RETURN FALSE;
END;$$;