Re: WIP: hooking parser

Sam Mason <sam@samason.me.uk> writes:
> On Wed, Feb 18, 2009 at 10:30:12AM -0500, Tom Lane wrote:
>> AFAIK, the Oracle behavior is just about entirely unrelated to the
>> parser --- it's a matter of runtime comparison behavior.  It is
>> certainly *not* restricted to literal NULL/'' constants, which is the
>> only case that a parser hack can deal with.

> How about introducing a "varchar2" type as in Oracle?

Maybe.  I think right now we don't allow input functions to decide
that a non-null input string should be converted to a NULL, but
that might be fixable.  It'd still be an ugly mess though, since
I suspect you'd have to introduce a whole structure of varchar2
functions/operators paralleling text.  For example, what is Oracle's
handling of || ?  AFAICS they can't be standards compliant there,
which means you need a varchar2-specific nonstrict implementation
of ||, and then to make that work the way Oracle users would expect,
varchar2-ness rather than text-ness would have to propagate through
anything else that might be done to a column before it reaches the ||.
        regards, tom lane