Patryk Obara <patryk.ob...@gmail.com> writes: > Ah! I presumed two separate loops, one throwing away oids and second > one actually filling a table - this makes more sense. I was just about > to send v4, but will rewrite the last patch and we'll see how it looks > like.
Yeah, it is understandable if you missed my "a loop that runs exactly twice", as that pattern, while we do use it in a few places in our codebase, is of limited applicability in general---the cost of discarded computation in the first pass need to be low enough for the improved maintainability to make sense.
