save custom attributes in archived h5py data files

[fanmingyu212]

ARTIQ Feature Request

Problem this request addresses

Currently users can only save HDF5 datasets in the datasets group of a file. This does not provide a clean way to save custom metadata, which include parameters not fetched as experiment arguments and other experiment parameters (wavemeter frequencies, ion trap voltages, etc.).

Describe the solution you'd like

Some solutions are:

1. Provide access to HDF5 attributes of datasets and groups
2. Provide functions to create new groups, and save custom attributes as datasets in custom groups.

Comments/ideas from others would be appreciated. It would be helpful to know how the more advanced ARTIQ users save their metadata in the data file.

[airwoodix]

This is linked to #1545 and references therein (in particular #1345).

[sbourdeauducq]

This does not provide a clean way to save custom metadata, which include parameters not fetched as experiment arguments and other experiment parameters (wavemeter frequencies, ion trap voltages, etc.).

Why not just save those as output datasets? Taking laser wavelength and ion trap voltage measurements is pretty much part of a regular Experiment.

[fanmingyu212]

Why not just save those as output datasets? Taking laser wavelength and ion trap voltage measurements is pretty much part of a regular Experiment.

I would prefer to save these variables in a different hdf5 group than "datasets" to organize data better. We also save data such as AOM frequency shift direction, single/double pass, etc. These configurations seem more like attributes than datasets to me, though I agree we could save them as datasets too.

Currently we use a homebuilt server to save hdf5 files in custom format, so I can close this issue if you would like.