Re: any way for ORDER BY x to imply NULLS FIRST in 8.3?

On Wed, 2007-11-07 at 11:39 -0500, Tom Lane wrote:
> Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> > On Wed, 2007-11-07 at 10:23 -0500, Tom Lane wrote:
> >> I put this in the same category as altering the identifier case-folding
> >> rules.
>
> > That has much less effect on application portability,
>
> Really? Try counting the number of requests for that in the archives,
> vs the number of requests for this.

I think you're arguing in favour of both changes, not burying my point.

Most applications don't hit the case folding issue for identifiers.
Certainly people have, but those are people doing things with metadata
like trying to write tools that work with both. They're database savvy
people who come on list and try and fix things.

Almost all applications have NULLs and use ORDER BY and indexes. That
doesn't mean everybody is effected by NULL sorting, but they might be
and probably don't realise.

I think you're right in identifying there are other issues for
portability. My list would be:

1. statement level abort
2. equivalent performance of identical SQL (e.g. NOT IN)
3. case insensitive searches
4. NULL ordering
5. case folding identifiers

Those differ depending upon the database.

--
Simon Riggs
2ndQuadrant http://www.2ndQuadrant.com