Noah Misch <noah@leadboat.com> writes:
> Describing in those terms illuminates much. While the concepts do suggest 2^N
> worst-case planning cost, my artificial test case showed a rigid 4^N pattern;
> what could explain that?
Well, the point of the 2^N concept is just that adding one more relation
multiplies the planning work by a constant factor. It's useful data
that you find the factor to be about 4, but I wouldn't have expected the
model to tell us that.
regards, tom lane