Re: GSoC 2017 - pdepe/bvp4c solvers

[Bill Greene]

I started writing the function I call bvp1d
(https://github.com/wgreene310/bvp1d)
to become familiar with the algorithm in the standard bvp4c paper and then
just kept adding to it. I believe it is a relatively complete implementation
of what is in the paper. I subsequently became aware that the bvp4c
implementation in MATLAB also supports a capability they call
"multipoint boundary value problems." The user interface to this functionality
seems very awkward and error prone to me; I had/have no interest in implementing
it. Even more fundamentally, I began to question the whole value of the
nodal collocation approach in bvp4c compared with a different approach based on
the weak form of the BVP, like FEM.

My real interest is in 1D pde solvers like pdepe and I have an implementation
that I call pde1d (https://github.com/wgreene310/pde1d). It uses a
linear approximation function compared with the higher-order approximation
in pdepe. But, beyond that, I think it is mostly compatible with pdepe.
I had intended to make the capability available through Octave forge but
the difficulty of configuring C++ packages there discouraged me. The
inclusion of Sundials/IDA in core Octave might make this easier but I have
not revisited the issue.

On Thu, Mar 2, 2017 at 2:51 PM, Sebastian Schöps <address@hidden> wrote:
> vlefkopo wrote
>> Hello,
>> I am Vasilis Lefkopoulos, and I'm interested in contributing to a project
>> as part of GSoC 2017.
>>
>> I've studied the list of suggested projects [1] for GSoc 2017. I've also
>> read the more general list of projects for Octave [2]. Specifically, I
>> would be interested working on the "solver for initial-boundary value
>> problems for parabolic-elliptic PDEs in 1D" [3] or/and the "solver for 1D
>> nonlinear boundary value problems" [4].
>>
>> I would like to know:
>> 1. What is the current status of these projects? I know that they're not
>> as of yet implemented into Octave, and I also saw that they were
>> implemented in the past but were taken down due to copyright issues with
>> Matlab.
>> 2. Are these projects suitable for my GSoC application (and potential
>> 3-month full-time work)? If yes, is there a relevant mentor that could
>> support me?
>> 3. Is there any relevant short project that I can work on, to familiarize
>> myself with these topics and also prove that I'm capable of committing to
>> them? I've contributed to Octave Forge before, so I am familiar with the
>> way contributions are made.
>
> Dear Vasilis,
>
> 1. I am aware of several programmers who asked about these projects on the
> mailing list but to my knowledge nothing has been finished. The last one was
> laserluke. The thread can be found in the mailing list archive with subject
> "Working on bvp4c". Does anyone knows what happened to this attempt?
>
> 2. Depends who is relevant for you :) I guess there are many developers
> interested in that including Carlo, Nick, Bill and myself.
>
> 3. you could start by compiling https://github.com/wgreene310/bvp1d and
> investigate whether this is a viable path. I would personally favour a
> m-file implementation but we already have a dependency on Sundials.
>
> Kind regards,
> Sebastian
>
>
>
> --
> View this message in context: 
> http://octave.1599824.n4.nabble.com/GSoC-2017-pdepe-bvp4c-solvers-tp4682117p4682131.html
> Sent from the Octave - Maintainers mailing list archive at Nabble.com.
>