Re: New design for FK-based join selectivity estimation

Tomas Vondra <tomas(dot)vondra(at)2ndquadrant(dot)com> writes:
> On 06/18/2016 06:52 PM, Tom Lane wrote:
>> I concur, actually, but I feel that it's out of scope for this
>> particular patch, which is only trying to replace the functionality
>> that was committed previously. If you want to come up with a patch on
>> top of this that adds some accounting for NULLs, I'd be willing to
>> consider it as a post-beta2 improvement.

> Sure, fair enough. By post-beta2 you mean for 9.7, or still for 9.6?

I think it'd be legitimate to address the NULLs question for 9.6, as long
as the patch is not very large. If it does turn out to be invasive or
otherwise hard to review, waiting for 9.7 might be more prudent. But
I argued upthread that failing to consider nulls was a bug in the original
patch, so dealing with them could be considered a bug fix.

> If I could wish one more thing - could you briefly explain why you
> rewrote the patch the way you did? I'd like to learn from this and while
> I think I kinda understand most of the changes, I'm not really sure I
> got it right.

I don't at the moment recall everything I changed, but I'm happy to
answer questions that are more specific than that one ...

regards, tom lane