dolfin team mailing list archive

[Johan Hake <johan.hake@xxxxxxxxx>]

On Monday April 18 2011 13:09:23 Anders Logg wrote:
> On Mon, Apr 18, 2011 at 01:07:09PM -0700, Johan Hake wrote:
> > On Monday April 18 2011 13:00:44 Anders Logg wrote:
> > > On Mon, Apr 18, 2011 at 12:51:37PM -0700, Johan Hake wrote:
> > > > Hello!
> > > > 
> > > > While doing some cleanup in the site-package directory I stumbled
> > > > over the meshconvert modules which resides in the dolfin
> > > > subdirectory. I think it should reside in dolfin_utils instead of in
> > > > the dolfin directory.
> > > 
> > > What should the distinction be between dolfin and dolfin_utils? It was
> > > originally created as a module for code for which it was important
> > > that it didn't depend on DOLFIN being installed. But I guess it would
> > > also work to say that dolfin_utils is a directory for code that does
> > > not depend on DOLFIN (C++ wrappers + add-ons).
> > 
> > I think that would be reasonable.
> > 
> > > A reason for keeping it in the dolfin module would be that the
> > > converters could potentially depend on DOLFIN, say if the converters
> > > were to import the mesh into DOLFIN and call order().
> > 
> > That is a good point. All of my own mesh tools import dolfin and uses the
> > MeshEditor, as I then get the correct ordering. I am also able to
> > generate a correctly ordered MeshFunctions.
> > 
> > But as it is now a user need to install PyDOLFIN for using meshconverter,
> > which is not logical.
> 
> Agree.

Then I move it to dolfin_utils and we would then require potential future 
dependencies of pydolfin in meshconverter to be included in try: except: 
clauses.

Johan