> ... best not to do an EXPLAIN in the regress test, because I don't want
> to have to touch the tests every time the cost functions are tweaked...
At some point we really should have an "optimizer" regression test, so
y'all *do* have to touch some regression test when the cost functions
are tweaked. If it were isolated into a single test, with appropriate
comments to keep it easy to remember *why* a result should be a certain
way, then it should be manageable and make it easier to proactively
evaluate changes.
It likely would have have full coverage, but at least some over/under
test cases would help...
- Thomas