Re: Implied Functional index use (redux)

Gregory Stark <stark(at)enterprisedb(dot)com> writes:
> I've thought further about this and I believe the problem is simpler than we
> were thinking previously. All we need is one boolean flag on the equality
> operator for the data type (or perhaps it would be more convenient to have it
> on the operator class) that indicates that two objects will never compare
> equal unless they're binary equal.

Well, we could simplify it that far, but that lets out float, numeric,
most or all of the geometry types, and I'm not too sure I care to
promise it for text either (think Unicode combining characters...).
So really that's too simple IMHO.

regards, tom lane