[Dune-devel] Reduce number of core modules?

[Christoph Grüninger]

Hi Oliver,
thanks for sharing your thoughts!

> Modules were introduced to enforce a clear
> design and structure onto the Dune code, and I think that goal is as
> relevant today as it was when we introduced the modules.

I never have seen it that way. And I think having istl and geometry in
subfolders could be enough separation.
One counter example are vector and matrix structures which are
distributed between common and istl for historical reasons.

>> On the other hand it increases work: checking out, building, fixing
>> build-system issues gets more tedious with increasing number of modules.
> 
> I don't think so:
> a) CI has automated lots of that testing, and it currently scales well
> b) Our build system is overly complicated, AIUI it a lot of stability
>    and speed can be gained by simplifying and modernizing it.
> c) Random users should really use distribution packages, and not build
>    each and every module themselves.

Without doubt, improving the build system is a shared wish.
But you missed my point. Building the distribution packages, downloading
the core modules and so one is cumbersome because we have so many of
them. To set up PDELab or ACFem you can easily reach a dozen of dune
modules. Then the modularity creates quite some effort.

> The only point that I would agree with is not on your list (merge dune-uggrid
> into dune-grid). And that one has been officially voted down before.

Because it would oppose the modularity as I see it. I don't want to
built UGGrid and get all its warnings, if I don't want to use it. This
is not the case for geometry or ISTL, most of the disretization packages
use them anyway.

I would like to hear more opinions.

Bye
Christoph

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.dune-project.org/pipermail/dune-devel/attachments/20190613/824981c7/attachment.sig>