search

ome / ome_zarr_test_suite

Test suite of reader and writer implementations
4 stars 6 forks source link

Try bfio #14

Open joshmoore opened 3 years ago

joshmoore commented 3 years ago

see: https://pypi.org/project/bfio/

Nicholas-Schaub commented 2 years ago

Hey, I am the lead developer of bfio. I would be happy to contribute in any way. We are in the process of upgrading our ome zarr spec. It's not a complete implementation, but we are updating the way we handle metadata since there wasn't a clear approach to doing it when we first created it.

joshmoore commented 2 years ago

Hey @Nicholas-Schaub! Thanks for stopping by. A first question: do you have reader & writer implementations?

Nicholas-Schaub commented 2 years ago

Yes, but they are from v0.1. We are in the process of updating now. When that spec came out, the Bioformats raw converter was just dumping the ome xml into the same directory, which did not work for us. We ended up storing the metadata as zarr attributes, but I see that you are now storing the file inside zarr, which I think is an excellent solution.

Also, our implementation is not a complete implementation. We don't support things like the hcs structure or multiple images. That's primarily because we don't have a need for them on our platform. I think based on the files you have here, I think our implementation (when finished updating) should pass.

In the future we will probably do a full implementation since we generally deal with HCS data, but we are in the process of converting from ome tiff to ome zarr, and are maintaining a one image per file standard for now.

Nicholas-Schaub commented 2 years ago

@joshmoore Hey. I'm starting to integrate this as a submodule into bfio, develop unit tests, etc. Should v0.3 and v0.4 data have OME XML? It appears as though v0.1 and v0.2 both have OME XML, but these are missing for the other versions.

Is OME XML missing from the OME NGFF spec? Is this something that is going to be added? Is there some convention you plan on using even if not formally defined in the spec?

joshmoore commented 2 years ago

Should v0.3 and v0.4 data have OME XML?

Hi @Nicholas-Schaub. bioformats2raw has produced the OME-XML for most if not all versions but it's not considered part of the specification to date. https://github.com/ome/ngff/issues/104 is looking to add that support. I'd assume that an implementation SHOULD (per RFC 2119) ignore the OME-XML. The result of that is that an implementation then treats individually numbered subdirectory as a standalone OME-Zarr.

Is OME XML missing from the OME NGFF spec? Is this something that is going to be added? Is there some convention you plan on using even if not formally defined in the spec?

Yes. Yes. Yes. 🙂 per 104.