BUG #1687: Regular expression problem (II)
The following bug has been logged online:
Bug reference: 1687
Logged by: Halley Pacheco de Oliveira
Email address: halleypo@yahoo.com.br
PostgreSQL version: 7.4 and 8.0
Operating system: Linux and Windows
Description: Regular expression problem (II)
Details:
Maybe it would be easier to see the the problem I'm having with regular
expressions this way:
SELECT '192.168.0.15' SIMILAR TO
'([[:alnum:]_-]+).([[:alnum:]_-]+).([[:alnum:]_]+)';
?column?
----------
t
SELECT '192.168.0.15' SIMILAR TO '([\\w-]+).([\\w-]+).([\\w]+)';
?column?
----------
f
SELECT '192.168.0.15' ~
'^([[:alnum:]_-]+)\\.([[:alnum:]_-]+)\\.([[:alnum:]_]+)$';
?column?
----------
f
SELECT '192.168.0.15' ~ '^(([[:alnum:]_-]+)\\.){2}([[:alnum:]_]+)$';
?column?
----------
f
SELECT '192.168.0.15' ~ '^([\\w-]+)\\.([\\w-]+)\\.([\\w]+)$';
?column?
----------
f
SELECT '192.168.0.15' ~ '^(([\\w-]+)\\.){2}([\\w]+)$';
?column?
----------
f
Why does the first query gives a different output? It is not exactly the
same as the second query and similar to the others?
"Halley Pacheco de Oliveira" <halleypo@yahoo.com.br> writes:
Maybe it would be easier to see the the problem I'm having with regular
expressions this way:
Maybe it would be easier to see the the problem I'm having with regular
expressions this way:
SELECT '192.168.0.15' SIMILAR TO
'([[:alnum:]_-]+).([[:alnum:]_-]+).([[:alnum:]_]+)';
?column?
----------
t
SELECT '192.168.0.15' SIMILAR TO '([\\w-]+).([\\w-]+).([\\w]+)';
?column?
----------
f
SIMILAR TO patterns are required to match the whole data string; so
the above fails because it only matches 3 digit groups not 4. The
others all fail because you put explicit ^ and $ into them.
The reason the first one works is that you put _ into the pattern, which
means "match anything" in SIMILAR-TO land; so it gets translated to "."
to be fed to the regular regexp engine. (Arguably that should not
happen inside square brackets, but similar_escape() isn't smart enough
to distinguish.) And that makes it possible for one of the
[]-expressions to match two digit groups plus the intervening dot.
regards, tom lane