`SerafinHeader.from_triangulation` performance about pyteltools HOT 3 CLOSED

Dear Keith,

Thank you for reporting this performance issue, which might occur as I do not have as many boundaries (or islands) in my meshes.
The cost of build_ipobo is probably also related to the complex implementation of iter_on_boundaries (maybe we can do better).
Do you have in ADCIRC mesh format the information to build the IPOBO table? Or would you like to set the table to 0 and compute it afterwards (in BlueKenue for example)?
Concerning the proposal of adding an argument to provide the IPOBO table, it is a good idea and I think it should be in from_triangulation, as some verifications can be done (at least shape of the array).

Best Regards,
Luc

from pyteltools.

Hey Luc,

Thanks for your timely response.

I can calculate the IPOBO table in Python directly from the IKLE table (I call it entities in my codes). I have some code that perhaps could be inherited here.

https://github.com/CHLNDDEV/oceanmesh/blob/master/oceanmesh/edges.py#L114

This function computes the edges of the boundary of a mesh in winding order and can handle meshes with many boundaries. Then it should be walk the edges, increment a counter, and load this into the right position in the IPOBO table.

I can draft up PR shortly.

from pyteltools.

I close this issue, because I merged your pull request (thank you!).

SerafinHeader.from_triangulation has now an option to provide an IPOBO table (and avoid the call to SerafinHeader.build_ipobo). Moreover the IPOBO table seems to useless (not read) for Telemac...

from pyteltools.