[dune-pdelab] Why are constraints in finiteelementmap/

[Christoph Grüninger]

Hi!

> So before starting on that
> could someone please clearly state what should be done and why?

Why:
Having a folder "constraints" which contains helper classes for
constrains while the implemented constrains are located in the
folder "finiteelemenmap" seem not logical to me.
The finiteelemenmap is crowded with headers that are clearly no
FE maps and I would expect the constraints in the constrains folder.

What:
- Move stuff from constraints/ to constraints/common/
- Move constraints implementations from finiteelementmap/ to constraints/
(obviously: adapting other headers and the how-to, create transition headers)

> Anyway, the connection of constraints implementation
> to the finite element spaces must remain clear. And I would certainly vote
> against some generic folder two levels down.

I don't think constraints/ is a generic folder. To me, this is by far more
specific than finiteelementmap/.
I am not sure what you mean by "connection .. must remain clear". This 
is or should be documented.
I quickly checked the includes, neither the constraints nor the FE maps seem
to include one another or share a significant part of code. The connection is
by its meaning. But I may be wrong.

Best wishes,
Christoph

-- 
Die meisten Gebäude stammen aus einer Zeit, in der man auch 
Türklinken aus Beton gemacht hätte, wenn das denn ginge.
[Spiegel Online, 25. Februar 2013]
********************************************
CMWR 2014: 9th - 13th June 2014 in Stuttgart
         Please visit www.cmwr14.de
********************************************