Re: BUG #4949: NOT IN is prohibitive slower than the rewrite for medium to large sets

[Robert Haas <robertmhaas(at)gmail(dot)com>]

On Jul 28, 2009, at 9:47 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> "Ole Tange" <postgresql(dot)org(at)tange(dot)dk> writes:
> > (modulo NULLs which seem to always cause problems in NOT INs).
>
> > Because it can be rewritten, NOT IN should never be much slower  
> > than the
> > rewritten solution, as PostgreSQL should simply rewrite NOT IN to the
> > above.
>
> Let's see, you understand that the rewrite violates the SQL standard
> semantics of NOT IN, but you think we should do it anyway?

If the subquery can't return NULLs, the rewrite is valid. I know  
you've rejected the idea of checking for this in the past, but perhaps  
you should consider this a user vote in favor of doing so.

I learned the hard way not to do this >5 years ago but it seemed  
strange to me, too.

...Robert