[MOAB-dev] Changing adjacency behavior

[Jason Kraftcheck]

On 10/22/2010 03:04 PM, James Porter wrote:
> On Fri, 2010-10-22 at 14:57 -0500, Jason Kraftcheck wrote:
>> Hold on a minute.  I understood the plan to be that we change only the
>> iMesh_getEntArrAdj behavior to conform to the spec.  I'm not so sure about
>> changing the underlying MBCore::get_adjacency functionality.
> 
> Hm, well I guess I could do that, though it seems strange that two
> functions that claim to do the same thing would have different behavior.
> 

1) Changing the MOAB API is likely to break many things
2) Changes to the iMesh API are justified (standards conformance),
    where we have no real reason to change the MOAB API other than
    to be consistent with iMesh
3) The iMesh API is impractical in the way it differs from MOAB for this
    particular behavior (or to put it another way: the MOAB API is the
    way it is because we like it that way.)
4) The iMesh function differs from the MOAB one in many other ways
    (e.g. MOAB's cannot be used to get connectivity arrays because
    it returns a unique, sorted list of entities, while the iMesh
    API is the only method to get connectivity.)
5) Changing the MOAB API alone isn't sufficient to effect the desired
    change in the iMesh API.  Due to other differences the iMesh API
    does not use get_adjacencies if the target dimension is zero.
    This is where the inconsistent behavior comes from (not
    Core::get_adjacencies).