Re: WIP patch: collation assignment algorithm rewrite

Martijn van Oosterhout <kleptog@svana.org> writes:
> I've looked over the main code and it looks good. While I'm not totally
> conviced it has to be done as a seperate pass, this way is exceedingly
> readable and clear as to what is going on, which makes me much more
> confident of its correctness.

FWIW, I'm not at all confident that it's "correct".  What I do think
is that this method will make it a lot easier to tweak the behavioral
details until we like them ...

			regards, tom lane