dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Dec 12, 2008 at 09:23:21AM +0100, Johan Hake wrote:
> On Friday 12 December 2008 07:58:39 Johan Hake wrote:
> > On Friday 12 December 2008 00:20:49 Anders Logg wrote:
> > > On Thu, Dec 11, 2008 at 09:43:22PM +0100, Johan Hake wrote:
> > > > Hello!
> > > >
> > > > To be able to read a function from file in PyDOLFIN we have changed the
> > > > signature of generated code for a finite_element. It now equals:
> > > >
> > > >   'FiniteElement("Lagrange","Triangle",1)' aso
> > >
> > > But with lower-case "t" in "triangle"?
> >
> > Yes of course.
> >
> > > > It is maybe not as readable as the other but now one can use that
> > > > string to create a ffc(ufl).FiniteElement in python given a compiled
> > > > finite_element, with just:
> > > >
> > > >   repr(fe.signature())
> > > >
> > > > You will need the newest version of ffc to be able to utelise DOLFIN
> > > > functionality that uses this signature(), e.g. read/write Functions
> > > > from/to a file.
> > > >
> > > > Also if you previously have stored a function to file, in an xml
> > > > format, you need to either save it again to be able to read it in
> > > > (sic.) or change the lines defining the signatures (finite_element and
> > > > dof_map) in the end of the file.
> > >
> > > Impressive!
> > >
> > > It seems to work fine here, but the plotting from C++ still doesn't
> > > work. The only thing that seems to be needed is to make it possible to
> > > do
> > >
> > >   f = Function("function.xml")
> > >
> > > that is, without the filename= argument. It would simplify the
> > > plotting (since it can can then handle Functions and Meshes in the
> > > same way), simplify the interface (no need to write "filename=") and
> > > make it uniform (both with the C++ interface for Function and with
> > > other classes in Python like the Mesh class).
> > >
> > > Since you already check for the .xml suffix in function.py, this
> > > should be rather easy. I was tempted to make the fix myself but was
> > > unsure I might break some logic. Could you make the fix?
> >
> > I'll do.
> 
> Done.

Excellent!

The plotting from C++ seems to work nicely now.

-- 
Anders

Attachment: signature.asc
Description: Digital signature