Writing class data to disk

[phockett]

Quick-fix is Pickle, but quite brittle to package version changes (e.g. recently broke after updating XR & Pandas versions).

Otherwise need to wrap something per-dataset/dict/dataframe/dataarray...? This is probably OK for aggregate cases, but a bit unwieldy for batch fit cases.

Basics at https://docs.xarray.dev/en/stable/user-guide/io.html

For more on this (for IO in ePSproc) see phockett/ePSproc#8

[phockett]

Some basics in places as of 47b3541

• Pickle for all self.data
• Wrapped Pandas (>HDF5) and Xarray (>netCDF) routines tested for aggregate 'fits' data.
• Basic Xarray > dict > JSON also test, but not yet wrapped to a function.

See also phockett/ePSproc#8 (comment)

Dev notes: code-share/jupyter-shared/PEMtk_dev_2022/io_dev/pemtk_dataIO_tests-dev_200522_StimpyDocker.ipynb

TODO:

• Finish testing & docs.
• Generalise methods, currently mainly tested with aggregate fit data. Should be OK for other data types too.
• Finish identifying issues - currently problems with nested Xarrays and tuple coord data (Euler dim) for HDF5 or JSON serialization. May be able to avoid and/or convert such cases?
• Custom routine with h5py? Might be best way to handle some tricky cases.
• Fix ep.readXarray restacking, should include dim checking routines to avoid failure.

[phockett]

Some of the above now fixed by ePSproc updates (see phockett/ePSproc#8 (comment)), see also docs: https://epsproc.readthedocs.io/en/dev/dataStructures/ePSproc_dataStructures_IO_demo_280622.html

Needs testing with and/or porting to PEMtk.