Complex PG regex query - sql

Question

I have the following working function which is used in check constraint (I'll only post the relevant SQL part):
-- a comma should always be followed by a space
-- a period should always be followed by a space, except if it is the last character of the string OR the string contains 'caporal'
-- a question mark should always be followed by a space, except if it is the last character of the string
-- must not contain 2 or more spaces in a row
-- must not contain ((
-- must not contain ))
-- any open parenthesis should be closed: number of '(' should equal to number of ')'
SELECT
($1 !~ ',(?!\s)|\s{2}|[?](?!\s(?!$)|$)|[()]{2,}') AND
((array_length(string_to_array($1, '('), 1) - 1) = (array_length(string_to_array($1, ')'), 1) - 1)) AND
($1 ~ 'caporal' OR $1 !~ '[.](?!\s(?!$)|$)')
Overtime I realized that I need to allow a period without a following space for the cases:
.fr
.com
.net
.co.uk
Also, I realized that I need to allow float numbers to be written with comma/period as separator. The following cases should be valid:
2,5cm
10.4l
I was trying multiple things but it seems I'm just breaking the existing rules instead of adding "exceptions" to them.
My latter attempt was the following:
SELECT
($1 !~ '[[a-zA-Z]àâçéèêëîïôûùüÿæœ],(?!\s)|\s{2}|[?](?!\s(?!$)|$)|[()]{2,}') AND
((array_length(string_to_array($1, '('), 1) - 1) = (array_length(string_to_array($1, ')'), 1) - 1)) AND
($1 ~ 'caporal' OR $1 !~ '[[a-zA-Z]àâçéèêëîïôûùüÿæœ][.](?!\s(?!$)|(?!fr)|(?!com)|$)')
But it clearly isn't what I want. Thank you in advance for hints and advices!

Answer 1

You should change the first regex to
,(?!\d(?<=\d,\d)|\s)|\s{2}|\?(?!\s(?!$)|$)|[()]{2,}
and the last one to
\.(?!\d(?<=\d\.\d)|(?:fr|com|co\.uk|(?<=\yco\.)uk|net)\y|\s(?!$)|$)
The changes are additions to the negative lookaheads that fail the match if their patterns match immediately to the right of the current location.
In the first case, ,(?!\d(?<=\d,\d)|\s) is used to match any comma that is not followed with a whitespace or any digit that is a fractional digit (as it must be preceded with a digit and a comma).
In the second regex, a similar restriction is added, see \d(?<=\d\.\d) that makes the \. match a dot that is not the first fractional digit in a float number with a period as a decimal separator, and the (?:fr|com|co\.uk|(?<=\yco\.)uk|net)\y part is added to avoid matching a . that is followed with fr, com, co.uk, the second period in co.uk ((?<=\yco\.)uk lookbehind makes sure that the comma before uk not preceded with co. is still matched) or net as whole words (see \y, word boundary).