Greg Stark <stark@mit.edu> writes:
> On Thu, Sep 29, 2011 at 4:22 AM, Peter Eisentraut <peter_e@gmx.net> wrote:
>>> Not sure what to do about this. Is it okay to suppose that collation
>>> can be ignored when matching to a collation-less index?
>> That sounds correct on first reading.
> Doesn't this depend on the semantics of the ? operator?
Well, yeah. But if the operator requires a particular collation, what's
it doing in an operator class for a collation-less indexed datatype?
I think the operator could actually depend on its input collation with
respect to some part of the processing it does on its collatable
right-hand input. But if the left-hand column (the indexed column) is
of a non-collatable type, it's hard to see how the operator could depend
on the index having a collation.
Also, I've now tested a patch along these lines and it passes core and
contrib regression tests, so there's not anything too obviously broken
about the idea.
regards, tom lane