[Dune-devel] [GSoC 2016]

[Christian Engwer]

There is one thing one should consider and it is worthwile
investigating this as part of the junior-job.

The boost::python code is a horrible mess, as it still uses C++03. And
as far as I know it is basically unmaintained.

There is a new "version" called pybind11. It uses C++11, which
dramatically simplifies the code and it is header-only, which I also
consider a big benefit.

Ciao
Christian

On Mon, Mar 14, 2016 at 09:20:46AM +0100, Martin Nolte wrote:
> Hi Govind,
> 
> as promised, we have added a junior job to get in touch with writing
> Python bindings for DUNE, namely writing example bindings for our
> FieldVector class in boost::python.
> 
> We think boost::python Bindings will be easier to maintain over the
> years (after all, we are C++ programmers). The downside is that we have
> to port or rewrite the work done in the BEM++ project. In our opinion,
> it's worth, though, because the bindings can be slimmer. Notice,
> however, that this is not yet a final decision and you are highly
> welcome to make another proposal by yourself.
> 
> Best,
> 
> Martin
> 
> On 03/11/2016 08:25 PM, Andreas Dedner wrote:
> > Hi Govind,
> > Welcome to Dune to thanks for your interest in our GSoC projects.
> > 
> > To answer your question w.r.t. boost::python and Cython. I would think that the first step would be to 
> > evaluate the work done within the bem++ project which is based on cython.  For dune-fem we are providing bindings for 
> > quite high level features of our packages while the bindings in bem++ export a big part of the grid interface. So the existing
> > cython bindings are much more low level then what we are currently working on for dune-fem and would complement what we are doing and also be potentially of great interest to the whole dune community.
> > 
> > We will add a junior job to the list over the weekend which is more focused towards the dune-python project. Until then 
> > checking out bempp and dune and seeing how the cython bindings work and can be accessed within python would be a great starting point.
> > 
> > Best
> > Andreas 
> > 
> > 
> > 
> > 
> > On 11/03/16 12:31, Govind Sahai wrote:
> >> Hi to all,
> >> I am Govind Sahai, currently a 3rd year engineering student in Indian
> >> Institute of Technology, BHU Varanasi, India pursuing Integrated Dual
> >> Degree in Mathematics And Computing. While searching for project ideas
> >> on your ideas
> >> <http://users.dune-project.org/projects/gsoc2016/wiki/Project_ideas#Project-4-Python-bindings-for-the-DUNE-grid-interface> page,
> >> I found interest in "Project 4: Python bindings for the DUNE grid
> >> interface". I have prior experience in C++ and Python, and also have
> >> some experience with boost::python. Also since some part of DUNE is
> >> already been exported to Python, using Cython, and dune-fem is using
> >> boost::python, will the new DUNE module use Cython or boost::python
> >> (or both)?
> >> Also, the link of *BEM++* given in *Project Description* of *Project
> >> 4* is giving *404* on my system.
> >>
> >> -- 
> >> ------------------------------------------------------------------------------------
> >>
> >>
> >> Govind Sahai
> >> Indian Institute of Technology
> >> Mathematical Sciences
> >> Part-III
> >> Varanasi-221005
> >>
> >>
> >>
> >> _______________________________________________
> >> Dune-devel mailing list
> >> Dune-devel at dune-project.org
> >> http://lists.dune-project.org/mailman/listinfo/dune-devel
> > 
> > 
> > 
> > _______________________________________________
> > Dune-devel mailing list
> > Dune-devel at dune-project.org
> > http://lists.dune-project.org/mailman/listinfo/dune-devel
> > 
> 
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
> 

-- 
Prof. Dr. Christian Engwer 
Institut für Numerische und Angewandte Mathematik
Fachbereich Mathematik und Informatik der Universität Münster
Einsteinstrasse 62
48149 Münster

E-Mail  christian.engwer at uni-muenster.de
Telefon +49 251 83-35067
FAX     +49 251 83-32729