Use local tree id instead of global tree id in the geometry interface

[Davknapp]

We usually iterate over local elements, and request process local information. Apart from functions that are use to create a cmesh (set_vertices etc.) we should always use a local tree id in my opinion. Otherwise it results in multiple calls of t8_cmesh_get_global/local_id. Is there another good reason to use global ids in function that are not used to create cmeshes?

[holke]

My initial line of thought was that the geometry is quite independent from any other information and that a caller may only have the global id available.

Looking at it from a different angle, your points are valid. Also

• we need the cmesh to call the functions and we can always use the global_to_local computation should we only have the global id
• geometry information will probably be partitioned, so it does not make sense to call it for non-local trees

Maybe indices of ghost trees might be an issue...

I would suggest trying to change it to local and see if it breaks.

[Davknapp]

Will do that, thanks for sharing your thaughts on this. I will try it and keep you up to date ;)