David Sagan

> It looks like pytables and h5py both use 'r' and 'i' as names for the compound dataset. Yes using 'r' and 'i' to be compatible with pytables and h5py...

Chris has been using h5py. I have not.

@PrometheusPi > How would this new standard definition effect ℂ3 datasets as used in PIConGPU? Here there are three complex numbers (x, y, z). PIConGPU uses a single level to...

@ax3l @ChristopherMayes Since the beam physics extension is actively being used, and since the question of complex storage has come up, I added a note to the extension on how...

@ChristopherMayes So right now the note I wrote mandates "r" and "i" to be consistent with h5py.

@ax3l So how would you word this in the standard? What happens if someone is using a format that does not support complex nor compound types? Also with your example...

@ax3l > We intentionally do not try to re-invent file formats and a complex is literally just a value[2] or equivalently struct{ T r; T i}, so one has quite...

@ax3l Actually it does not look like the h5py way is broadly supported. For example see: https://mathematica.stackexchange.com/questions/63983/hdf5-and-complex-numbers > HDF5 does not directly support complex numbers. Programs that do seem to...

@ax3l @ChristopherMayes Another issue I thought of is that in the case where, say, the complex component is zero, If you follow the h5py way then you double the storage...

@ax3l @ChristopherMayes OK but let me write something first since I have a new proposal...