Re: [sqlsmith] Failed to generate plan on lateral subqueries

Andreas Seltenreich <seltenreich(at)gmx(dot)de> writes:
> Tom Lane writes:
>> [2. transitive-lateral-fixes-1.patch]

> I was about to write that sqlsmith likes the patch, but after more than
> 10^8 ok queries the attached ones were generated.

Ah-hah --- the new check I added in join_is_legal understood about
chains of LATERAL references, but it forgot that we could also have chains
of outer-join ordering constraints. When we're looking to see if joining
on the basis of a LATERAL reference would break some later outer join, we
have to look at outer joins to the outer joins' inner relations, too.

Fixed in the attached. I also stuck all of join_is_legal's
lateral-related checks inside an "if (root->hasLateralRTEs)" block,
which will save some time in typical queries with no LATERAL. That
makes that section of the patch a bit bigger than before, but it's
mostly just reindentation.

Many thanks for the work you've done with this tool! Who knows how
long it would've taken us to find these problems otherwise ...

regards, tom lane