On Tue, Jan 06, 2009 at 11:45:46AM -0300, Ola Skavhaug wrote: > On Mon, Jan 5, 2009 at 5:09 AM, Anders Logg <[email protected]> wrote: > > On Sat, Jan 03, 2009 at 09:27:04PM -0300, Ola Skavhaug wrote: > >> What's the purpose of dolfin_plot? It just calls viper_plot for all > >> types, except when the object is unknown, in which case it tries > >> to project something, without knowing what it is. This should be put > >> into viper instead of in this almost empty wrapper. What is the type > >> of object if project can handle it? > > > > The type of object is any expression like grad(u), u + u, div(u) etc. > > but it can also be anything else (like an int or float) that FFC can > > interpret as a Form. The easiest test is probably the one that's there > > now in plot.py (try: project). > > I don't understand. I don't see any tests where project is used. Can > you point me to the appropriate test? > > Ola
I mean the following tests:
# Plot mesh
if isinstance(object, cpp.Mesh):
return viper_plot(object, *args, **kwargs)
# Plot mesh function
if isinstance(object, cpp.MeshFunctionUInt) or \
isinstance(object, cpp.MeshFunctionInt) or \
isinstance(object, cpp.MeshFunctionReal):
return viper_plot(object, *args, **kwargs)
# Plot function
if isinstance(object, cpp.Function):
return viper_plot(object, *args, **kwargs)
# Try projecting function
from project import project
try:
print "Object cannot be plotted directly, projecting to
piecewise linears."
u = project(object)
return plot(u, *args, **kwargs)
except:
raise RuntimeError, ("Don't know how to plot given object and
projection failed: " + str(object))
This tests for any object that DOLFIN knows that Viper can handle and
otherwise tries to project the object to a Function and then calls
Viper with the Function.
So: first check for everything that Viper knows how to plot and
otherwise try to project it.
--
Anders
> >
> >> The case viper is not available, I guess you need some logic anyway.
> >> Plot muting when an environment variable is set could be placed in
> >> viper, but that would only be a duplicate of what you need anyway, so
> >> I suggest you keep that (or a better version of the IgnorePlot).
> >
> > It would be helpful if there were an environment variable, perhaps
> > NOPLOT that Viper looked at. Then we can remove that part too from the
> > wrapper. Maybe it's also good for Viper to have a plotless mode for
> > unit testing?
> >
> > The release of DOLFIN 0.9.0 is planned to come out today, so it would
> > be good if we could wait with any changes in the interface until after
> > the release.
> >
> > Can you make a Viper release today that matches the current DOLFIN?
> > There have been a few bug fixes since last time I think. There will
> > also be a new FFC release. Then Johannes can make a release of
> > fenics-2009-01. :-)
> >
> >
> > -----BEGIN PGP SIGNATURE-----
> > Version: GnuPG v1.4.9 (GNU/Linux)
> >
> > iEYEARECAAYFAklhwFMACgkQTuwUCDsYZdGUFACbB3vaOadgI8OcqKFcnXdSeXCS
> > Jd8AniNeO/k3JAPv0CLlOU1UmGdn6h9q
> > =9+nD
> > -----END PGP SIGNATURE-----
> >
> > _______________________________________________
> > Viper-dev mailing list
> > [email protected]
> > http://fenics.org/mailman/listinfo/viper-dev
> >
> >
>
>
>
signature.asc
Description: Digital signature
_______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
