Use _ensure_str_array_size in FlattenedStorage.extend

[pmrv]

Should fix #1407.

@Leimeroth Can you give this a try and add your example as a test case?

[Leimeroth]

It works, but I am actually confused why. I am adding the EAM example for the publication notebook and the trainingcontainer has a byte encoded str (dtype.char S). I guess it is accessed in a way that decodes it when reading it from the hdf5 container. However, I am not sure whether this would always be the case, as right now the problem is only the 'identifier' which is defined as a property also.

[pmrv]

HDF5 wants to store unicode as UTF-8, but numpy does UCS-32 (dtype 'U') internally, so the to_hdf/from_hdf convert raw UTF-8 encoded bytestrings (dtype 'S') for the file.

[jan-janssen]

Looks good to me

[jan-janssen]

Does it make sense to submit this fix to h5io?

[jan-janssen]

@pmrv Can this be merged? I would like to release a new pyiron_base version and would like to include this pull request.

[pmrv]

Does it make sense to submit this fix to h5io?

Not sure what you mean, this way of storing unicode string arrays in HDF in general should be in h5io?