Quick work-around; unused max cells parameter

[agzimmerman]

This is a work-around for issue #216 .

The 'max number of cells' parameter in the refinement section does not work for my simple code that uses the Poisson problem interface. I have not found the root cause. This commit is a quick work-around that instead:

1. adds a new parameter max_cells to the pi-DoMUS interface.
2. checks the number of active cells against this parameter in should_solver_restart

Such a parameter seems at least as relevant as already existing parameters such as 'kelly threshold' and 'max iterations', so maybe this is actually what we want. This 'work-around' commit violates the concept of 'do not repeat yourself', so I would call this an open issue.

[asartori86]

Hi @alexanderzimmerman, thanks for your contribution.

However, the ParsedGridRefinement already features the max_cells parameter. Could you try if this works (without your patch)? By default the section for grid refinement is as follows

subsection Refinement
  set Bottom fraction                        = 0.100000
  set Maximum number of cells (if available) = 0
  set Order (optimize)                       = 2
  set Refinement strategy                    = fraction
  set Top fraction                           = 0.300000
end

You should change the Refinement strategy to number and specify the Maximum number of cells

[luca-heltai]

Can you Move this workaround in deal2lkit instead of pi-domus? I think deal2lkit already has a similar logic behind its max_cells argument. There may well be a bug there if that does not work, but it is supposed to... If it doesn't, the right way is not to fix pi-domus, but to fix deal2lkit...

[luca-heltai]

And, by the way, thanks!