wenhui qiu <qiuwenhuifx(at)gmail(dot)com> writes:
> I found this path https://commitfest.postgresql.org/patch/3235/ already
> supports anti join , But I've found that in many cases it doesn't work.It
> always uses SubPlan Here's my testing process.
NOT IN is not convertible to an anti-join: the semantics are wrong
because of the way it treats nulls in the sub-select's output.
In principle you could do it if you could prove that the sub-select's
output is never null, but we need to do this transformation long
before we have enough information to make such a deduction.
regards, tom lane