Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

Right now the boundarySegmentIndex is not a unique identifier to a certain facet.
The boundarySegmentIndex of a facet can change at growth/merge/removal.
This needs some thought. Unique boundaryIds might me more useful. This way the
user can write a map from boundaryIds (are sadly named boundarySegmentIndex in the intersection interface now),
to boundaryFlags/information on boundary conditions that always stays valid, and needs a new element
everytime a new boundary is created by growth or removal.

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

will vanish in the growth process (=grid shrinkage). In this case we might
have to take care about physical quantities on this element that might get lost
thorugh removal, e.g. mass. The return value true->for new elements is already
returned by the grow() method. Thus it was twice the same value returned

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

The grid growth interface is now completely independent of the adaptation interface.
The mark and getMark method are not overloaded anymore. Grid growth uses own mark functions
markForGrowth, markForRemoval, markForMerging that have easy readable names. The step by step
approach of preGrow, grow, postGrow remains the same. PreGrow returns true if elements are added.
grow returns true if elements or boundaries are added (I need to extrapolate the solution and/or
assign boundary conditions).

The internal state variables get more meaningful names. ADD_NEIGHBOR if a neighbor is added to
this element resulting in growth. VANISH if the entity vanishes. MERGE_WITH_NEIGHBOR if a new
element is created by connecting to existing facets.

This also implements the corresponding functions growSimplexElement, removeSimplexElement,
and mergeSimplexElement. The test is extended to show the new functionality and do some error checking.

Geometrical information needed for the growth is passed when marking and stored inside the element.

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

Newly branched facets get a new boundarySegmentIdx because they constitute a new domain boundary.
Added elements on boundary facets take over the index from the facet that is now inside the domain.

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

[bugfix] The foamgrid specific variables need to be called from a FoamGridEntityImp not a Dune Entitiy

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

The grid gets an additional preGrow, grow, and postGrow method that handle growth analogously to adaptivity.
A method spawnSimplexElement is introduced that spawns a new element from a previously marked element.
Needs more error checking. Does not implement the prune interface yet.

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

- the VANISH state is renamed to better PRUNE
- bool coarseningBlocked_ flag blocks the coarsening of maxlevel elements that have
a juntion facet that has no father due to growth.
- implements hasFather method for codim1 entities
- bool spawned flag for codim1 entities that can be set in order to avoid growth on
the same facet twice in one growth step.

Timo Koch authored 9 years ago and Timo Koch committed 9 years ago

SPAWN indicates that this element will spawn a new element resulting in grid growth.
VANISH indicates that the element will vanish resulting in a reduced grid size.

This is fulfilled automatically in regular FV methods, but in network situations
this may be violated.

The result looks better than previously -- not sure whether it is correct, though.

For the network case, this way to compute the flux should be 'more correct'
than the old one.  Not sure if the new version is correct yet, though.

The program runs, but the output doesn't look quite correct yet.

Refine it once

This is the initial commit of an example program solving the transport
equation using a finite volume method.  This first commit simply copies
an existing program using YaspGrid into the correct location.  Future
patches will replace the YaspGrid by a FoamGrid, and make it work
on an embedded network grid.

Otherwise they are not build/run when make test is called.

In particular, FoamGrid should work for any dimension of dimworld, not just 1,2,3.

The default parameter was introduced in false postion.
[bugfix] changes embarrassing absolute paths in tests to relative paths

Updated global refine test to unit test globalRefine(0) and globalRefine() (=globalRefine(1)).

1d elements have special constructors that did not initialize some private variables. They do that now.