On Tue, Jun 18, 2019 at 9:24 PM Thomas Munro <thomas.munro@gmail.com> wrote:
> That points to the problem: for JOIN_UNIQUE_INNER we plan a Parallel
> Hash Join, but that's nonsense, there is no code to unique-ify the
> partial inner side (because that's not possible). There may be
> something better we can do here (like understanding that this should
> really be a semi-join), but this works for me to prevent the bad plan:
Here's a tidier version with a comment and commit message.
--
Thomas Munro
https://enterprisedb.com