[SciPy-Dev] Delay Differential Equation (DDE) solver

[Ralf Gommers]

On Wed, May 27, 2020 at 11:47 PM Jansen Jérôme <jrme.jansen at gmail.com>
wrote:

> Hi Ralf,
> I was very happy to read your answer and see that my piece of code could
> interest the SciPy community!
>
> About maintenance of the solver, it won't be a problem for me to help with
> that.
>
> Indeed, this issue 818 was opened a long time ago and did not interest
> many people ... I note that already in 2013 the references of L.F. Shampine
> and S. Thompson were already cited.
>
> As you pointed out, DDEs are very specific and a big problem will be
> finding a charitable soul to review the code. Being at my first attempt to
> contribute, I imagine that I have to wait for a reviewer(s) ..?
>

Being new doesn't imply you go to the back of the line - we're very happy
to see new contributors. That said, we do have a perpetual backlog, so it
depends on finding someone else with both interest and some expertise.  It
doesn't have to be a SciPy maintainer. What has worked well in the past for
such larger features is finding another domain expert to
test/review/collaborate. With two people saying "this solves my problem and
the algorithm looks correct" it becomes a lot easier to review, getting it
to meet SciPy's code and test requirements is usually a lot less work than
figuring out whether the algorithmic part is right.

Cheers,
Ralf



> Best regards,
> Jérôme.
>
> Le mer. 20 mai 2020 à 23:02, Ralf Gommers <ralf.gommers at gmail.com> a
> écrit :
>
>>
>>
>> On Sun, May 17, 2020 at 1:41 PM Jansen Jérôme <jrme.jansen at gmail.com>
>> wrote:
>>
>>> My dear SciPy community,
>>>
>>> In my PhD, I work on biomechanical problematics which are (sometime)
>>> modeled with delay differential equations.
>>> Being an avid user of python and SciPy, I looked for DDEs solver but ...
>>> I was surprised not to find it in SciPy. On the contrary, I discovered lots
>>> of more or less old and very different projects about DDEs in Python
>>> (JiTCDDE, ddeint, pydelay, diffeqpy). While in other programming languages,
>>> robust code exist (matlab: dde23, julia: DelayDiffEq.jl, fortran: DKLAG6
>>> ...).
>>> So I decided to start implementing a solver using the architecture of
>>> scipy.integrate.solve_ivp.
>>>
>>> The fork us here :
>>> https://github.com/jrmejansen/scipy/tree/ddeSolver
>>>
>>> I added the folder "*scipy/integrate/_dde*" and modified files from
>>> "scipy/integrate/_ivp" to resolve DDEs. You would find also some benchmarks
>>> but not yet unit tests.
>>>
>>> At this time, the solver, *solve_dde*, handles DDEs with constant
>>> delays for none stiff equations.
>>>
>>> Does my code interest the SciPy community ?
>>> If it does, I can add extensive unit tests and also docstring.
>>>
>>
>> Hi Jérôme, thanks for proposing this. Having a good DDE solver would be
>> quite nice. I've certainly missed it in the past. The difficulty here is
>> that it is fairly niche, and we'd need to find someone to review the PR.
>> scipy.integrate is already struggling a bit with lack of maintainers. Also,
>> we need to think about future maintenance - if you would be willing to
>> continue to maintain it, that would certainly help.
>>
>> We did have a feature request open for a long time, but there wasn't too
>> much interest it looked like: https://github.com/scipy/scipy/issues/818
>>
>> So first question would be, is there someone else - either a SciPy
>> maintainer or a domain expert - who is interested in helping to get this
>> merged?
>>
>> Best,
>> Ralf
>>
>>
>>> I am open to any discussion and do not hesitate to give me feedbacks.
>>> Best regards,
>>> Jérôme JANSEN.
>>> _______________________________________________
>>> SciPy-Dev mailing list
>>> SciPy-Dev at python.org
>>> https://mail.python.org/mailman/listinfo/scipy-dev
>>>
>> _______________________________________________
>> SciPy-Dev mailing list
>> SciPy-Dev at python.org
>> https://mail.python.org/mailman/listinfo/scipy-dev
>>
> _______________________________________________
> SciPy-Dev mailing list
> SciPy-Dev at python.org
> https://mail.python.org/mailman/listinfo/scipy-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scipy-dev/attachments/20200528/e902d2d8/attachment-0001.html>