Save region-decomposed information?

[johnomotani]

It is a pity that hypnotoad cannot save more-or-less the information in the MeshRegion objects to grid files, and have BOUT++ read that information directly, rather than translating everything to a logically-rectangular global mesh.
Even without being useful for BOUT++, it might be nice to have that capability, although optionally so normal grid files are not made bigger unnecessarily. This information (e.g. connections between different regions) could be useful for post-processing tools. With a little bit more effort, I guess it should be possible to reload hypnotoad objects (MeshRegion, Mesh, EquilibriumRegion, etc.) from a grid file, in case it is useful to have access to those objects directly (e.g. post-processing calculations) without having to re-run the full grid generation process from inputs.
Saving more information could also make regression tests a bit more comprehensive - I think they are just comparing the output grid files for differences, so if more information (e.g. connectivity) was saved, that could also be regression-tested.

This issue is just to float the idea in case it seems useful enough to anyone for them to want to spend the time to implement it.

[bendudson]

Thanks @johnomotani ! I agree this would be a better way to manage the multiblock mesh information, rather than the 'jyseps' indices. I think the blocks should still map to regions of a global array, but the global array can have gaps. This is what ADIOS2 does, and would be compatible with NetCDF. The scheme needs to handle some unusual features like twist-shift conditions, but having a way to connect regions together in different ways and label boundary regions would be very useful.