Re: Regexp confusion

Alvaro Herrera <alvherre(at)commandprompt(dot)com> writes:
> Doug Gorley escribi:
>> Trying to match some numbers, and I'm having some regexp problems.
>> I've boiled it down to the following:
>>
>> /* (1) */ select '3.14' similar to E'^\\d+\\.\\d+$'; -- true
>> /* (2) */ select '3.14' similar to E'^\\d+(\\.\\d+)$'; -- true
>> /* (3) */ select '3.14' similar to E'^\\d+(\\.\\d+)*$'; -- true
>> /* (4) */ select '3.14' similar to E'^\\d+(\\.\\d+)?$'; -- false
>> /* (5) */ select '3.14' similar to E'^\\d+(\\.\\d+)+$'; -- true
>>
>> So, based on (1) and (2), the pattern '\.\d+' occurs once. So why
>> does (4) return false? between (3), (4), and (5), it appears as
>> though the group is matching multiple times.

> I think the confusion is about what SIMILAR TO supports. ? it doesn't.
> See here:
> http://www.postgresql.org/docs/8.4/static/functions-matching.html#FUNCTIONS-SIMILARTO-REGEXP

> You probably want to use ~ instead of SIMILAR TO.

> (SIMILAR TO is a weird beast that the SQL committee came up with,
> vaguely based on regular expressions.)

Hmm ... actually I think *none* of those should have succeeded, because
^ and $ are not supposed to be metacharacters in SIMILAR TO. We are
failing to quote them, but apparently we need to --- it looks like the
regexp engine processes ^^ at the start of the pattern the same as ^,
and likewise for $$ at the end.

regards, tom lane