Re: trouble caused by change in 7.3 handling of '' in

On 18 Dec 2002, Vivek Khera wrote:

> I see in the 7.3 HISTORY file this:
>
>      * An empty string ('') is no longer allowed as the input into an
>        integer field. Formerly, it was silently interpreted as 0.
>
> This is causing major issues with the Request Tracker program
> <http://www.bestpractical.com/rt>.
>
> The author has gone so far as to notify all current users that
> Postgres 7.3 is not usable with RT, and to stick to 7.2, or to use
> MySQL.

This is backwards.  RT isn't usable with Postgresql 7.3.  There's a SQL
standard (several actually, but 92/99 are mostly it) .  If RT had been
programmed to them, it wouldn't even notice this change.  I know that
sounds cold hearted, but it isn't Postgresql's job to work with
noncompliant SQL.

> Apparently, there is a lot of work that will be necessary to retrofit
> RT to this change.  Is there any way possible to make this a logged
> warning rather than a fatal so that there is time to transition the RT
> code?  I didn't see any notice in any prior release of the upcoming
> change.

the problem is that it is not allowed by spec, and allowing it is against
spec, so it is likely that if it was implemented, it would have to be
similar to the transform_null_equals switch which is off by default.

> This unfortunately locks me into the 7.2 line until (if) RT can be
> fixed up, and I could really use the 7.3 improvements for my other
> applications.

Please be clear here.  The problem is with RT, not Postgresql.  I know
that doesn't comfort you personally, but it's an important point.  If RT
wants to be portable, it's the kind of thing they need to fix, as well as
probably many more things in their SQL code.

Maybe now would be a good time for a code review of their SQL to see what
other possible non-sql spec problems are lurking there.

I'm downloading it now, and I'll take a look at the SQL code in it to see
what can be done.