How to implement correctly type from NXdetector in hdf?

[RGBYCP]

Dear Nexus community,
Could I ask here for advice, please?

NXdetector has an entry called type
https://manual.nexusformat.org/classes/base_classes/NXdetector.html

"Description of type such as He3 gas cylinder, He3 PSD, scintillator, fission chamber, proportion counter, ion chamber, ccd, pixel, image plate, CMOS, …"

How do I implement this with h5py correctly?
The manual for h5py does not give an example for string datasets. https://docs.h5py.org/en/stable/strings.html#storing-strings

My approach:

uv_spectra=grp_uv.create_group("uv_spectra")
uv_spectra.attrs["NX_class"] = 'NXdetector'
string_dt = h5py.special_dtype(vlen=str)
uv_spectra_type=uv_spectra.create_dataset('type', data='ccd', dtype=string_dt)

But I don't know if this is the correct way.
Thank you for your help.

[rayosborn]

That is the correct way of storing variable-length strings, but I think it's the h5py default, so you probably don't have to specify a dtype at all. You could also store the string as fixed-length byte strings, because NeXus doesn't specify how strings are stored. However, variable-length strings would allow you to change the value more easily if you need to and are more flexible in handling Unicode.

I would also like to suggest you have a look at the nexusformat package, which would help to ensure that you write NeXus-compliant files. It's uses h5py, but handles things like setting the class variables automatically and is more succinct. Your four lines above could be written as:

uv_grp['uv_spectra'] = NXdetector(type='ccd')

There is more documentation here.

[RGBYCP]

Dear @rayosborn

Thank you very much for your quick help.
Thanks for your confirmation of how to do this with h5py. I tried it without the dtype, it works as well. Thank you for the explanation about the two string types. I will then opt for the moment for the variable-length string.

Thanks for pointing out to the nexusformat package. I saw it before, but I know already h5py and it was suggested to me that I use h5py. I will make a suggestion to use the nexusformat.

Again, thank you for your help.

[rayosborn]

The nexusformat package is a thin layer on top of h5py that ensures that certain NeXus rules are obeyed, understands NeXus classes, and how to construct NXdata groups, so I personally think it makes sense to use it, particularly if you are new to NeXus, but then I am biased. You can try it out in an online Jupyter notebook if you want to see what it does.