Re: using xmin in a query?

[Michael Nolan <htfoot@xxxxxxxxx>]

On Thu, Jul 28, 2011 at 5:36 PM, Tom Lane <tgl@xxxxxxxxxxxxx> wrote:

Michael Nolan <htfoot@xxxxxxxxx> writes:

> It also appears you cannot group on a column of type xid.

You can in 8.4 and up.  Previous versions only know how to GROUP BY
sortable columns, which requires a btree opclass, which xid doesn't
have and really can't have because it doesn't have a linear ordering.
There is a hash opclass for it, though, so in versions that know how to
GROUP BY using hashing, it'll work.

:sigh:  I thought I had done all the tests on my 9.0.4 testbed server too.  One of
these days I hope to get the production and development servers off 8.2.  

I more or less understand why xid types don't have a linear ordering from Robert Hass's tutorial at  PGCON11.

So, a <> operator (either xid,xid or xid,integer) would need to be implemented using the hash opclass, correct? 

(I don't have a use case for it yet, though.) 

> Would adding a <> operator enable that?

No, it's pretty irrelevant ...

OK, thanks for putting up with my noobie questions.
--
Mike Nolan