[MOAB-dev] r3299 - MOAB/trunk

[James Porter]

This seems to break iMesh_getEntAdj and iMesh_getEntArrAdj when passing
in iBase_ALL_TYPES for the requested type. Attached is a test that shows
the problem for iMesh_getEntAdj. Using specific entity types (e.g.
iBase_EDGE) works as expected, however.

- Jim

On Fri, 2009-11-06 at 16:29 -0600, kraftche at cae.wisc.edu wrote:
> Author: kraftche
> Date: 2009-11-06 16:29:04 -0600 (Fri, 06 Nov 2009)
> New Revision: 3299
> 
> Modified:
>    MOAB/trunk/MBCore.cpp
> Log:
> o Fix very broken vector-vector get_connectivity:
>   assumed that all entities in array had the same type as the first entity.
> o Change vector-range get_connectivity such that it is worst-case
>   O(n ln n) instead of O(n^2)
> o Rewrite vector-vector get_adjacencies with specialized code instead
>   of call to range-range get_adjacencies
> o Templatize implementation of range-range get_adjacencies such that
>   it can be shared by range-range get_adjacencies and vetor-range
>   get_adjacencies w/out converting the input list.
> o Fix bug in some forms of get_adjancies when input list contains
>   only one entity and target type is vertex: still need to handle
>   INTERSECT/UNION flag if 'output' container is not empty.
> o Fix inconsistent behavior for different versions and special
>   cases of get_adjacencies: all should return input entities if
>   asked for adjacent entities of the same dimension (not returning
>   such entities would be more consistent with iMesh, but that broke
>   other code in MOAB.)
> 
> Time required to skin a mesh using vertex->element adjacencies
> was reduced from 9.32 seconds to 7.88 seconds with these changes.
> 
> Passes all tests.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: test2.c
Type: text/x-csrc
Size: 1419 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/moab-dev/attachments/20091106/0cc31906/attachment.c>