Dennis Gläser authored Sep 01, 2021 and Timo Koch committed Sep 22, 2021

The element was never used, but solely passed to have a conforming
interface to the TPFA filler. However, the interfaces are not the same
anyway, since e.g. IVDataStorage is passed to the mpfa filler.

Simon Emmert authored Mar 26, 2020 and Timo Koch committed Apr 01, 2020

Place no deprecation macros where GridView will be substituted by Grid::LeafGridView after release 3.2.

Dennis Gläser authored Mar 30, 2020 and Timo Koch committed Mar 30, 2020

[mpfa][fluxvarscachefiller] detect zero diff coeffs also for single phase for compatibility with tracer model

Timo Koch authored Mar 26, 2020 and Ned Coltman committed Mar 26, 2020

FluidSystem::getMainComponent returns the main componet index of the phase.
This doesn't actually make sense for the tracer models which why this commit also
* removes getMainComponent for tracer fluid systems (Changelog amended)
* guards all occurences of getMainComponent in code that works for tracer too
with constexpr if

Dennis Gläser authored Mar 04, 2020 and Ned Coltman committed Mar 26, 2020

[mpfa] throw deprecation warning if vol vars do not implement an effectiveDiffusionCoefficient() function

Katharina Heck authored Sep 05, 2019 and Kilian Weishaupt committed Oct 02, 2019

in Ficks law to mass reference system

Timo Koch authored Apr 06, 2019 and Kilian Weishaupt committed Apr 24, 2019

The default flux variable cache filler helper classes for mpfa and tpfa
are actually specific for porous medium flow models, just like the cache
itself. This change moves them to a new header
porousmediumflow/fluxvariablescachefiller.hh
and deprecates the old classes and headers in the discretization folder.

Dennis Gläser authored Nov 26, 2018 and Timo Koch committed Dec 02, 2018

The interaction volumes and data handles are now stored within an
extra container structure. Access to interaction volumes from the
grid flux variables cache or its local view is now only possible
for a given scvf, for which then the respective iv it is embedded
in is returned.

This is more in agreement to the local view classes.

That makes it possible to overload the local assembler class while
reusing an interaction volume.

instead of storing pointers to the transmissibilities and pressures
etc, we now only store a pointer to the data handle and reconstruct
the fluxes in the DarcysLaw implementation. This dramatically reduces
the number of lines needed for the cache implementation.

Instead of deriving from the process-related handle classes the
handle class contains these sub-handles and provides access to it.
That allows process-independent algorithms on the sub-handle classes
as soon as the return functions get more generic names.

Resizing of the matrices place in the local assembler directly wherr
needed. This has a small negative impact on performance but saves a lot of code.