dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

On May 6 2009, Martin Sandve Alnæs wrote:

> On Wed, May 6, 2009 at 8:45 PM, Johan Hake <hake@xxxxxxxxx> wrote:
> > On Wednesday 06 May 2009 14:07:11 DOLFIN wrote:
> > > One or more new changesets pushed to the primary dolfin repository.
> > > A short summary of the last three changesets is included below.
> > >
> > > changeset:   6098:4b3c43f72124bd0be03b6eaa0d7d2dd1f7b03d65 tag:       
> > >   tip user:        "Garth N. Wells <gnw20@xxxxxxxxx>" date:        Wed 
> > > May 06 13:06:55 2009 +0100 files:       dolfin/la/BlockVector.cpp 
> > > dolfin/la/BlockVector.h dolfin/la/EpetraVector.cpp 
> > > dolfin/la/EpetraVector.h dolfin/la/GenericVector.h 
> > > dolfin/la/MTL4Vector.cpp dolfin/la/MTL4Vector.h 
> > > dolfin/la/PETScLUSolver.cpp dolfin/la/PETScMatrix.cpp 
> > > dolfin/la/PETScMatrix.h dolfin/la/PETScVector.cpp 
> > > dolfin/la/PETScVector.h dolfin/la/Vector.h dolfin/la/enums_la.h 
> > > dolfin/la/solve.cpp dolfin/la/uBLASVector.cpp dolfin/la/uBLASVector.h 
> > > dolfin/nls/NewtonSolver.cpp dolfin/ode/MonoAdaptiveNewtonSolver.cpp 
> > > dolfin/ode/MultiAdaptiveNewtonSolver.cpp 
> > > site-packages/dolfin/__init__.py site-packages/dolfin/errornorm.py 
> > > description: Some enum -> std::string for la norms.
> >
> > The nice thing with enums was that I could easily get hold of the 
> > possible values. I just took a look in enum_la.h.
>
> And a decent editor could do auto-completion as well.
>
> > Now I have to look into the
> > different .cpp files and scan the 'if' statements.

This still needs to be documented in the header files.

> > Could we predefine some variables for this purpose? Something like this 
> > for preconditioners:
> >
> >  std::string no_prec("no_prec");  // Instead of "none" due to PyDOLFIN?
> >  std::string jacobi("jacobi");
> >  std::string sor("sor");
> >  std::string ilu("ilu");
> >  std::string icc("icc");
> >  std::string amg_hypre("amg_hypre");
> >  std::string amg_ml("amg_ml");
> >  std::string default("default");
>
> Then what's the point of using strings instead of enums?
> Could just as well use enums and make
> enum-to-string-to-enum utilities for file handling and printing.

- enums are a bit clumsy to use - enums are not wrapped automatically for 
the Python interfface - we can use the string "none" in the Python 
interface (?) - the diversity of linear algebra backends; the enum list can 
become very long with solvers, precondintioners, etc which may only be 
supported by one backend.


> > If I misspell the string I will not get an error but rather the default
> > version. Predefined variables will help for this.

Only in the Epetra backend as far as I know (which I didn't write). The 
others throw an error.

Garth

> > I parameter system with "allowed" values would ofcourse solve this.
>
> Yes, warnings or errors when using invalid strings is simply a must have.
>
> Martin
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev