Dealing with versioning

[tischi]

@joshmoore @sbesson

I assume you have a lot of experience with dealing with multiple versions.

Do you already envision a way of representing the multiple version of OME-Zarr on the Java side? Taking the multiscales as an example?!

I could also lay out different ways that I could imagine, but, as I have no experience with this, I think it makes more sense if you present your view first.

[sbesson]

I assume you have a lot of experience with dealing with multiple versions.

The answer is probably yes and no.

If I understand the intent of this library, the existing repository with the closest intent is https://github.com/ome/ome-model and the ome-xml Java component which contains code-generated classes for all elements defined in the OME-XSD schema - see https://javadoc.io/doc/org.openmicroscopy/ome-xml/latest/index.html

Importantly, this component does not support multiple versions of the OME schema but tracks the latest version. This is why the release of a new version of the OME-XSD has such impact on the OME stack as breaking version of this component and all the downstream ones are required.

The ome-zarr-py explored another paradigm to support reading/writing from multiple versions of the OME-NGFF. There are likely various patterns that could be reviewed and reused like the usage of versioned classes https://github.com/ome/ome-zarr-py/blob/master/ome_zarr/format.py i.e. MultiscalesV01, MultiscalesV02...

[tischi]

i.e. MultiscalesV01, MultiscalesV02...

That's the pattern that also @tpietzsch suggested. If you made good experiences with that, why not?

1. Combined with a JSONDeserializer that picks the right one based on the version string in the JSON?!
2. And down-stream code only support that latest, assuming that older ones are converted into the latest?

[joshmoore]

I'm not sure another library will be able to do the automatic detection of version, but I assume our library will be able to encapsulate that. And yes, I'd imagine some form of "upgrade chain" will let clients get to the "latest version that I support".

[tischi]

Ok!

How do we proceed practically?

Shall I put my current code into a branch and we iterate on that?

Or do you have code already?

[joshmoore]

I think putting your code here is a fine place to start (since I don't have any other code). The version detection I would "steal" from ome-zarr-py. The code generation that I was discussing will take me a bit longer.