GitLab

[doc] adapt CHANGELOG and README files



See merge request !199

[2pLiquidVapor] Add new model and test

* 2p1cni model to simulate steam injection

- [x] Find better name ?
- [x] Rename classes and files
- [x] Adapt tests
- [x] Clean-up
- [x] Proper documentation

See merge request !188

[postprocessing][pvpython] Fix use of input and output directory.



See merge request !177

[io] Add function to read and write container from and to file

Write some container to file (requires` begin()` and `end()`):
```c++
std::vector<double>(5, 0.0) v;
writeContainerToFile(v, "myvector.txt");
```

Read to a container from a file (requires `being()` `end()` and `push_back()`):
```c++
auto v = readFileToContainer<std::vector<double>>("myvector.txt");
```

I found this quite useful recently.

See merge request !194

[implict][fluxvars] Fix update function to make implicit problems compilable using Forchheimer



See merge request !195

[fluxvars] Remove deprecated constructors and methods in all fluxvariables



See merge request !180

[components] Replace std::pow were possible

I doubt it will bring a lot of performance, but it looks clearer to me.

See merge request !196

*Fix errors
*Add .bib entry for citation

Fix/extrusionfactor for outflow

While included in the cc localresidual, the stokes and box localresidual did not account for the extrusion factor when dealing with outflow boundary conditions,
leading to weird results.

See merge request !191

[mpfal3d] Bugfix for interaction volume container

Fix of flyspray task FS#313

See merge request !193

[h2oair] Use sqrt instead of pow to speedup viscosity evaluation

test_stokes2c with refined grid speeds up by 2.7%.

See merge request !192

test_stokes2c with refined grid speeds up by 2.7%.

*Flux leaving the domain needs to be scaled be extrusion factor

*Flux leaving the domain needs to be scaled by extrusion factor

Feature/scotch matrix reordering

Reordering the matrix leads to improved sparsity pattern. This is especially important for growing network grids.

Reordering results using the ReorderingDofMapper implemented by this MR with a growing network grid.

* Original unsorted FoamGrid
![original](/uploads/4f29ec169127ea67dbe814f6ee9daab5/original.png)
* Reordered dof indices
![reordered](/uploads/e2ade49c081353981b00cdf8ce7b2c51/reordered.png)


See merge request !186

The mapper computes a permutation of the index set using the external PT-SCOTCH library.
This is a simple replacement for element mapper (cc) / vertex mapper (box). It requires to
store on int per dof for the mapping (fast direct vector access).
This makes the matrix bandwidth small and reduces the fillins during LU factorization.
So it helps especially for direct solvers or iterative solver preconditions by ILU.
The reordering is particularly helpful for network problems with foamgrid to optimize
the matrix sparsity pattern.

This enables the use of Dune's single codim single geometry mapper
that otherwise exists failing a static assert

[2p2c] Improve Millington&Quirk and constant tau

A quick test (Clang 3.8, one run, no error bar) of test_box2p2c with the optim.ops from dumux and a 100 times finer grid:  
Millington&Quirk: 99 s  
Constant Tau: 95 s  
That's about 5%. But more measurements are required and the differences in the solution must be checked.

This merge request does not address all changes proposed in the task [FS#321](http://www.dumux.org/flyspray/index.php?do=details&task_id=321)

See merge request !181

* Reduce number of template arguments
* Rename property from tau to TauTortuosity

Use multiplications and a cubic root instead
of std::pow
* Estimated 5% less time in 2p2c computation
* Roughly inbetween old code and constant tau

This was wrongly altered in a previous commit. Fick's law depends on
the mole fraction gradient. The density has to be approximated at
the face.

Also fixes the diffusive fluxes that were sometimes wrongly added
to the total mass balance when using the replaceCompIdx to replace
one component balance by the total mass balance.

[fix][test] Fix the reference solution name for 2pbox fracture. We start counting from 0 now.



See merge request !190

* Add ref solution

*Clean-up
*Add cc test

* Remove unused code
* Adapt output of permeability