Re: pg_class.relpartbound definition overly brittle

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> Also, you're attacking a straw man. Accidentally storing a meaningless
> parse location in the catalog isn't a feature, and we shouldn't
> pretend otherwise.

It's not meaningless, and it is a feature, useful for debugging.
Otherwise you'd have a hard time telling one occurrence of e.g. "+" from
another when you're trying to make sense of a stored tree. We worked out
all this behavior ages ago for other expression node trees that are stored
in the catalogs (default expressions, index expressions, check
constraints, etc etc) and I don't see a reason for partition expressions
to be different.

I agree that this means you can't just strcmp a couple of stored node tree
strings to decide if they're equal, but you couldn't anyway --- a moment's
perusal of equalfuncs.c will show you other special cases, each with their
own rationale. Maybe you'd like to complain that every one of those
rationales is wrongheaded but I do not think you will get far.

I think the best advice for Mark is to look at pg_get_expr() output and
see if that matches.

regards, tom lane