On Wed, 2006-03-01 at 14:18, Bruce Momjian wrote:
> Stephan Szabo wrote:
> > > justify_days doesn't currently do anything with this result --- it
> > > thinks its charter is only to reduce day components that are >= 30 days.
> > > However, I think a good case could be made that it should normalize
> > > negative days too; that is, the invariant on its result should be
> > > 0 <= days < 30, not merely days < 30.
> >
> > What about cases like interval '1 month -99 days', should that turn into
> > interval '-3 mons +21 days' or '-2 mons -9 days'?
>
> I think it should be the later. It is best to have a single sign, and I
> think it is possible in all cases:
>
> '2 mons -1 days'
>
> could be adjusted to '1 mons 29 days'.
There's a part of me that thinks the WHOLE THING should be positive or
negative:
-(2 months 1 day)