The metadata extracted from the files is at the moment unstructured, meaning that every Extractor behaves slightly differently.
It might be more straightforward and consistent to dump all metadata read and parsed from files, as is, without changes (except mapping from int/bool flags to strings) into an original_metadata element of each Datatree.attrs. This would then follow the convention also used in HyperSpy (https://hyperspy.org/hyperspy-doc/current/user_guide/io.html#metadata).
The metadata extracted from the files is at the moment unstructured, meaning that every Extractor behaves slightly differently.
It might be more straightforward and consistent to dump all metadata read and parsed from files, as is, without changes (except mapping from int/bool flags to strings) into an
original_metadata
element of eachDatatree.attrs
. This would then follow the convention also used in HyperSpy (https://hyperspy.org/hyperspy-doc/current/user_guide/io.html#metadata).Processing of those elements of
original_metadata
into (ideally semantically annotated)metadata
might then be the next step, see also https://github.com/hyperspy/rosettasciio/issues/89.The
original_metadata
element would then serialise as a json string in the exported files.