dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, Apr 28, 2008 at 02:26:08PM +0200, Martin Sandve Alnæs wrote:
> 2008/4/28 Anders Logg <logg@xxxxxxxxx>:
> >
> > On Mon, Apr 28, 2008 at 02:18:53PM +0200, Martin Sandve Alnæs wrote:
> >  > 2008/4/28 Anders Logg <logg@xxxxxxxxx>:
> >  > > On Mon, Apr 28, 2008 at 02:08:14PM +0200, Martin Sandve Alnæs wrote:
> >  > >  > 2008/4/28 Anders Logg <logg@xxxxxxxxx>:
> >  > >  > > On Mon, Apr 28, 2008 at 12:17:07PM +0200, Ola Skavhaug wrote:
> >  > >  > >  > Anders Logg skrev den 28/04-2008 følgende:
> >  > >  > >  > > Are there any outstanding issues we need to fix before the release?
> >  > >  > >  > >
> >  > >  > >  > > (Other than adding "virtual" in LA subclasses if there are no further
> >  > >  > >  > > opinions on that.)
> >  > >  >
> >  > >  > Can we add min() and max() to GenericVector?
> >  > >
> >  > >  Yes, why not, but why do you need those?
> >  > >
> >  > >  There is VecMin and VecMax in PETSc so that's not a problem.
> >  >
> >  > max() is the same as norm(linf), so only min() is needed.
> >  >
> >  > Right now, I needed them to adjust the color range of plots.
> >  > But I'm sure there are other uses as well for getting the
> >  > range of values in a vector.
> >
> >  ok, feel free to add them to GenericVector (both min and max). I guess
> >  a suitable location would be after the declaration of norm()?
> 
> Ok! Should I add max() when we already have norm(linf)?
> It may be more intuitive in user code to write
>   umin, umax = u.min(), u.max()
> than
>   umin, umax = u.min(), u.norm(linf)
> but I see no other reason to have both.

Yes, I think it's ok to add both. It doesn't hurt.

-- 
Anders