On Thu, Jul 1, 2021 at 6:10 PM Thomas Munro <thomas.mu...@gmail.com> wrote:
> One thing I'm wondering about is whether it's worth having stuff to > support future experimentation like ST_SORT_SMALL_THRESHOLD and > ST_COMPARE_RET_TYPE in the tree, or whether we should pare it back to > the minimal changes that definitely produce results. I think I'd like > to keep those changes: even if it may be some time, possibly an > infinite amount, before we figure out how to tune the thresholds > profitably, giving them names instead of using magic numbers seems > like progress. I suspect if we experiment on two extremes of type "heaviness" (accessing and comparing trivial or not), such as uint32 and tuplesort, we'll have a pretty good idea what the parameters should be, if anything different. I'll do some testing along those lines. (BTW, I just realized I lied and sent a .patch file after all, oops) -- John Naylor EDB: http://www.enterprisedb.com
