At 20:56 +0200 on 28/03/1999, Richi Plana wrote:
>
> 1) I'm not sure the regex thing is ANSI SQL and would port to other SQL
> systems easilly
That's true. ANSI only has LIKE, but it's oh-so-limited.
> 2) Isn't there a way to do [^{ws}](word)[{ws}$], where [^{ws}] means
> either the start or a whitespace and [{ws}$] means either a whitespace or
> EOL?
Hehe... Actually, there is. This is what I came up with:
SELECT * FROM {class} WHERE '{phrase}' ~ ( '(^|[[:space:]])' || ( rtrim( FieldN ) || '([[:space:]]|$)' ) );
It seems a bit complicated. The [[:space:]] combination is the set of all
whitespace characters. Like perl's \s, but locale-sensitive.
The combination '(^|[[:space:]])' uses the regexp alternatives syntax. This
means "Either the beginning of the phrase (^) or a space character". The
combination '([[:space:]]|$)' is, along the same lines, "either a space
character or the end of the phrase".
All the rest of the mess are concatenation operators, and the parentheses
required for their precedence and associativity.
> 3) When I try the concat strings operator (||), I get the following psql
> error:
Yes, you are right about the precedence (I don't know why I didn't
encounter it yesterday). One must also remember that the || operator is
also non-associative (no real reason for it - it's a minor bug), so you
have to put parentheses around it if you concat more than two strings.
Herouth
--
Herouth Maoz, Internet developer.
Open University of Israel - Telem project
http://telem.openu.ac.il/~herutma
