Joe Van Dyk wrote:
>
> Perhaps I fat-fingered something somewhere... I tried that and I got
> this:
> https://gist.github.com/joevandyk/4958906/raw/5561f95ef2b5d82f81ab14913c4d36f6aac3ee0a/gistfile1.txt
>
> The with_filters view uses a different plan.
Interesting. It is avoiding the hash join, but it is still evaluating
the exists column even when it is not referenced at all in the select. I
would have expected the optimizer to remove it entirely.