Re: New pg_lsn type doesn't have hash/btree opclasses

On 2014-05-11 06:02:23 +0900, Fujii Masao wrote:
> On Sat, May 10, 2014 at 9:51 AM, Fabrízio de Royes Mello
> <fabriziomello(at)gmail(dot)com> wrote:
> >
> > On Fri, May 9, 2014 at 9:30 PM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> >> [ shrug... ] "proactive" would have been doing this a month ago.
> >> If we're going to ship a release, we have to stop taking new features
> >> at some point, and we are really past that point for 9.4.

We *couldn't* do it a month ago since we didn't know about it a month
ago. My delorean is getting a bit old.

> > I agree with you that is too late to add *new features*.
> >
> > But I agree with Andres when he said this is a regression introcuced in the
> > pg_lsn patch.
> >
> > So we'll release a version that break a simple query like that:
> >
> > fabrizio=# SELECT DISTINCT (g.i||'/0')::pg_lsn f FROM generate_series(1,
> > 100) g(i), generate_series(1, 5);
> > ERROR: could not identify an equality operator for type pg_lsn
> > LINE 1: SELECT DISTINCT (g.i||'/0')::pg_lsn f FROM generate_series(1...
> > ^
>
> I agree that this is not new feature but just the fix of oversight of
> the pg_lsn patch.
> Without such opclass, we cannot execute even such simple query.

I don't even understand why it's questionable whether this should be
fixed. It's an almost trivial patch for an oversight in a newly
introduced feature. We gain absolutely nothing by patching this in the
next cycle, except that things need to be tested twice.

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services