[Dune-devel] Alberta: State of the union?

[Simon Praetorius]

Hi Christoph,

For me, the Alberta grid and grid reader was some years ago a reason why 
to move to dune as FEM basis. We, in our group, have developed long ago 
a framework on top of an old Alberta fork and have developed many tools 
for working with this grid and the Alberta grid files. This we didn't 
want to give up when developing a new framework or switching to 
something else. So, the Alberta GridManager was an argument for us to 
consider DUNE. Nowadays, I would also prefer some other grid 
implementations instead. But, for example, there is no other grid 
implementation in dune-grid directly supporting (unstructured) surface 
grids, or unstructured periodic grids. So you have to switch to an 
external grid module.

There is also the AlbertaReader supporting the alberta grid format, 
including the ability to directly read boundary IDs from the files and 
store its directly in the grid. (this also works with gmsh format but 
needs more manual steps to read in that information)

So, in summary: In my opinion the Alberta Grid/GridReader are still 
used. If we would move it out of dune-grid into a separate module (that 
would not be maintained either) it would simply die, since then maybe 
no-one feels responsible for fixing the small (cmake) bugs, or 
compatibility issues. Now, when it is in dune-grid, it needs to be 
working, it is in the tests and is available in all the local setups. 
So, if something breaks, an issues will be opened and mostly small bugs 
are fixed. A cmake cleanup is on its way and simplifies the 
configuration slightly. (But still it is required to link 
dimension-specific libraries.)

Maybe we could update the documentation, mark it as "not maintained", or 
even as deprecated. But if deprecated, we have to list alternatives. In 
dune-grid there is no alternative. ALUGrid is an external module. UGGrid 
does not support some of the features I would choose AlbertaGrid for.

Best,
Simon

Am 08.02.21 um 07:37 schrieb Christoph Grüninger:
> Hi Dune!
> What's the current state of Albert? It has been in some kind of extended
> maintenance mode for the last couple of years, but I am not sure, what
> to think about it at the current state:
>
> - The Dune page describing Alberta's installation is outdated: Link to
> "recent nightly builds" is broken and gets redirected to the generic
> IANS homepage. OpenSuse packages were removed as they failed to build
> years ago. A link to IANS' GitLab is missing. I'll update this page
> (!371), but still, it is a sign that there are not many people who care
> about Alberta.
> - Our build systems has issues with Alberta, but there are no vocal
> complaints. Alberta won't adapt, we have to get it working on CMake's side.
> - Who is maintaining Alberta? Who is advocating Alberta's concerns
> within Dune? Claus did most with regards to my first question, not sure
> about the second one.
> - If you don't find the release tarballs on IANS' GitLab page, it's
> really tricky to build Alberta. You have to look into MAINTENANCE,
> ignore all the outdated stuff about Solaris, Subversion, and Alberta 2.0
> and find that using the script fromsvnreconf.sh is still the way to go.
> - Debian package is still 3.0.1, not 3.0.3.
>
> Please don't take this as a rant. I appreciate all work and all the
> effort made over the past years!
> We have three major grids we advertise (e.g. bindings are/were part of
> dune-grid), as they were the historic foundations of Dune: ALUGrid, UG,
> and Alberta. I am not sure, whether Alberta should still remain first
> tier and would like to hear your perception.
>
> Bye
> Christoph
>
-- 
Dr. Simon Praetorius
Technische Universität Dresden
Institute of Scientific Computing
phone: +49 351 463-34432
mail: simon.praetorius at tu-dresden.de
web: https://tu-dresden.de/mn/math/wir/das-institut/beschaeftigte/simon-praetorius