[KRLango]

This is a quick and dirty fix for
#87
While we await the computation work that will allow a better fix

[lisham2000]

Very much quick and dirty but as an interim fix it looks fine. Might be worth making an issue to review and resolve this post computation work

[Ion-e]

LGTM

[Tiomat85]

LGTM

[cmeyer]

This approach is fine as a short term solution. But the format we use for transferring metadata during computations results in the source metadata being in a sub-group:

Can you make it match that?

Here is the related general issue for this which would standardize how we do this:

• nion-software/nionswift#1622

Also, are there other places in both eels-analysis and experimental that have this issue? This will be resolved completely with upcoming computations PR, but as long as we're implementing a stop-gap solution, should we implement it comprehensively or just for this particular case?

[cmeyer]

Here's how it's done in computations. It will transfer source_metadata if it already exists so that stacked computations get the original metadata; and it only transfers acquisition related metadata.

                metadata = data_item.metadata
                if "source_metadata" in metadata:
                    metadata = metadata["source_metadata"]
                copy_keys = ("hardware_source", "instrument", "scan")
                data_item_d = dict[str, typing.Any]()
                for key in copy_keys:
                    if key in metadata:
                        data_item_d[key] = metadata[key]
                return data_item_d

[cmeyer]

For reference, this is the code in nionswift which does this.