Re: Make foo=null a warning by default.

On Mon, Jul 16, 2018 at 11:37:28AM -0400, Tom Lane wrote:
> Heikki Linnakangas <hlinnaka(at)iki(dot)fi> writes:
> > On 16/07/18 18:10, Tom Lane wrote:
> >> TBH I'm not really excited about investing any work in this area at all.
> >> Considering how seldom we hear any questions about transform_null_equals
> >> anymore[1], I'm wondering if we couldn't just rip the "feature" out
> >> entirely.
>
> > Yeah, I was wondering about that too. But Fabien brought up a completely
> > new use-case for this: people learning SQL. For beginners who don't
> > understand the behavior of NULLs yet, I can see a warning or error being
> > useful training wheels. Perhaps a completely new "training_wheels=on"
> > option, which could check may for many other beginner errors, too, would
> > be better for that.
>
> Agreed --- but what we'd want that to do seems only vaguely related to
> the existing behavior of transform_null_equals. As an example, we
> intentionally made transform_null_equals *not* trigger on
>
> CASE x WHEN NULL THEN ...
>
> but a training-wheels warning for that would likely be reasonable.
>
> For that matter, many of the old threads about this are complaining
> about nulls that aren't simple literals in the first place. I wonder
> whether a training-wheels feature that whined *at runtime* about null
> WHERE-qual or case-test results would be more useful than a parser
> check.

Am I understanding correctly that this would be looking for bogus
conditions in the plan tree?

Best,
David.
--
David Fetter <david(at)fetter(dot)org> http://fetter.org/
Phone: +1 415 235 3778

Remember to vote!
Consider donating to Postgres: http://www.postgresql.org/about/donate