To: Stephan Szabo <sszabo@megazone23.bigpanda.com>
Cc: Achilleus Mantzios <achill@matrix.gatewaynet.com>,
	Radu-Adrian Popescu <radu.popescu@aldratech.com>, pgsql-sql@postgresql.org
Subject: Re: SQL function parse error ? 
In-reply-to: <20030109085635.I92199-100000@megazone23.bigpanda.com> 
References: <20030109085635.I92199-100000@megazone23.bigpanda.com>
Comments: In-reply-to Stephan Szabo <sszabo@megazone23.bigpanda.com>
	message dated "Thu, 09 Jan 2003 09:00:31 -0800"
Date: Thu, 09 Jan 2003 12:17:15 -0500
Message-ID: <5570.1042132635@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>

Stephan Szabo <sszabo@megazone23.bigpanda.com> writes:
> Although the rules could be similar to those for + and - at the end of
> operator strings (no $ at the end of an operator unless it contains
> characters not normally in SQL92 operators).  I'm not sure that
> behavior is sensible either, but if someone wanted to
> do it for their own installation it's about a 2 line patch.

It could be done that way.  But given that "$" already has one weird
special case in the operator name rules (ie, it can't be the only
character of an operator name), I feel that we'd be making things overly
complicated.

The proposal back in Aug 2001 was to remove "$" from the set of operator
name characters altogether (which would allow us to use it in
identifiers instead, improving Oracle compatibility).  I originally
objected to that idea on backwards-compatibility grounds, but I'm
leaning more and more to the view that it's the right thing to do.

I've re-opened the thread on pgsql-hackers about this, and we'll see
whether any consensus emerges this time.

			regards, tom lane