Message-Id: <6.2.1.2.0.20060325094356.04de84d8@localhost>
Date: Sat, 25 Mar 2006 09:48:53 +0800
To: Tom Lane <tgl@sss.pgh.pa.us>
From: K C Lau <kclau60@netvigator.com>
Subject: Re: limitation using LIKE on ANY(array) 
Cc: pgsql-performance@postgresql.org
In-Reply-To: <5275.1143217555@sss.pgh.pa.us>
References: <200603170625.k2H6PJ926955@pillette.com>
	<1143112188.2810.27.camel@theo-laptop.int.coza.net.za>
	<20060323132448.GB4881@surnet.cl> <4423E90E.2090607@rentalia.com>
	<20060324125245.GB90527@pervasive.com>
	<4423EDA9.3090502@rentalia.com>
	<20060324130619.GF90527@pervasive.com>
	<20060324133252.GA22688@winnie.fuhr.org>
	<6.2.1.2.0.20060324225743.04d49ac0@localhost>
	<5275.1143217555@sss.pgh.pa.us>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"; format=flowed

Thank you very much, Tom. We'll try it and report if there is any 
significant impact performance-wise.

Best regards,
KC.

At 00:25 06/03/25, Tom Lane wrote:
>K C Lau <kclau60@netvigator.com> writes:
> > Indeed, I get rejected even with:
> > .. WHERE ANY(array) = 'xx'
>
> > It would only work as documented in the manual (8.10.5):
> > SELECT * FROM sal_emp WHERE 10000 = ANY (pay_by_quarter);
>
>That's not changing any time soon; the SQL spec defines only the second
>syntax for ANY, and I believe there would be syntactic ambiguity if we
>tried to allow the other.
>
> > With 8.1.3, I get an error when trying to do this on a Text[] column :
> > .. WHERE ANY(array) LIKE 'xx%'
>
>If you're really intent on doing that, make an operator for "reverse
>LIKE" and use it with the ANY on the right-hand side.
>
>regression=# create function rlike(text,text) returns bool as
>regression-# 'select $2 like $1' language sql strict immutable;
>CREATE FUNCTION
>regression=# create operator ~~~ (procedure = rlike, leftarg = text,
>regression(# rightarg = text, commutator = ~~);
>CREATE OPERATOR
>regression=# select 'xx%' ~~~ any(array['aaa','bbb']);
>  ?column?
>----------
>  f
>(1 row)
>
>regression=# select 'xx%' ~~~ any(array['aaa','xxb']);
>  ?column?
>----------
>  t
>(1 row)
>
>regression=#
>
>                         regards, tom lane