dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Mon, May 30, 2011 at 10:16:48PM +0100, Garth N. Wells wrote:
>
>
> On 30/05/11 22:03, Anders Logg wrote:
> > On Mon, May 30, 2011 at 09:39:17PM +0100, Garth N. Wells wrote:
> >>
> >>
> >> On 30/05/11 21:22, Johan Hake wrote:
> >>> On Monday May 30 2011 12:54:14 noreply@xxxxxxxxxxxxx wrote:
> >>>> ------------------------------------------------------------
> >>>> revno: 5907
> >>>> committer: Garth N. Wells <gnw20@xxxxxxxxx>
> >>>> branch nick: dolfin-all
> >>>> timestamp: Mon 2011-05-30 20:52:14 +0100
> >>>> message:
> >>>>   Comment out parts of test. There is something fishy with
> >>>> DofMap::tabulate_coordinates and the double pointers.
> >>>
> >>> Yes you are probably right. The double** typemap might be the problem. But
> >>> doesn't it work for other similare situations, like set and get double**?
> >>>
> >>> It would be nice to get rid of all foo** arguments as these pose some troubles
> >>> for the SWIG interface. I think we discussed this wrt the ufc interface too.
> >>>
> >>
> >> I'm going to remove them. The double pointer for coords is a UFC
> >> bug/design flaw.
> >
> > I wouldn't call it a design flaw. It was natural at the point of
> > designing UFC to use an array of arrays for a list of coordinates.
> > Designing for possible difficulties in creating SWIG typemaps was not
> > part of the design process.
> >
>
> I'd call it a flaw (as does Martin). It imposes significant constraints
> on the user. See
>
>   https://bugs.launchpad.net/ufc/+bug/787405
>
> UFC uses plain pointers for other square arrays.

Call it what you want.

> > But I agree it would be natural to replace the double pointers by
> > simple pointers (or perhaps even std::vector?).
> >
>
> I suggest a plain pointer in UFC since this provides maximum
> flexibility. My intention is to use boost::multi_array in the DOLFIN
> interface.
>
> > But is that something we should do now before the release of 1.0?
> >
>
> Why not? Even if we don't get it working in Python, it's not working
> properly anyway.

It adds to the list of things we need to implement, but it's a big
enough change that it would be good to have it in place before 1.0,
and it's a big enough change to call it UFC 2.2.

--
Anders


> Garth
>
> >> Garth
> >>
> >>> Johan
> >>>
> >>>> modified:
> >>>>   dolfin/swig/fem_pre.i
> >>>>   test/unit/fem/python/test.py
> >>>>
> >>>>
> >>>>
> >>>> Your team DOLFIN Core Team is subscribed to branch lp:dolfin.
> >>>> To unsubscribe from this branch go to
> >>>> https://code.launchpad.net/~dolfin-core/dolfin/main/+edit-subscription
> >>>
> >>> _______________________________________________
> >>> Mailing list: https://launchpad.net/~dolfin
> >>> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> >>> Unsubscribe : https://launchpad.net/~dolfin
> >>> More help   : https://help.launchpad.net/ListHelp
> >>
> >> _______________________________________________
> >> Mailing list: https://launchpad.net/~dolfin
> >> Post to     : dolfin@xxxxxxxxxxxxxxxxxxx
> >> Unsubscribe : https://launchpad.net/~dolfin
> >> More help   : https://help.launchpad.net/ListHelp