[feature request] add read support for n5/ome-zarr datasource

[xgui3783]

In siibra-python, we use neuroglancer-scripts to read neuroglancer precomputed formatted multiresolution volumetric data.

We recently encountered an external data source in n5 compressed by blosc, and would like to support it.

Ideally, we would like to extend neuroglancer-scripts so that it is able to read n5 datasource. This is so that, we can keep siibra-python's access pattern to multiresolution data source similar. I imagine it would also benefit neuroglancer-script, as it grows in capability.

We considered using an existing python library, such as cloud-volume , but we wanted to keep the module that reads multiresolution volume as light as possible. CloudVolume appears to pull in more dependencies than we would like. Since the author of cloud volume ended up rolling their own implementation, I can only imagine that there probably exists no other libraries in python that does this.

We want to gauge your opinion on this proposal, as we are prepared to dedicate resource to implement this feature.

edit: added ome-zarr (ngff) spec to file format to support

[xgui3783]

ping @ylep

[ylep]

Dear @xgui3783, I can only see advantages to supporting a new input format to neuroglancer-scripts! If you can dedicate resources on your side for implenting it, all the better :-) Given your previous contributions, I fully trust that the implementation will meet high standards, but I can still dedicate some time to reviewing the code, maybe participate in design decisions if you would like me to.

[xgui3783]

Good to hear! I have a PR in motion, but I am not yet 100 percent happy with it yet.

I will ping you for code review, when I feel they are ready.

[xgui3783]

do we still need to support python 3.6?

I would like to add dataclass to better type the dictionaries. dataclass support is backported to python3.6, but it meant adding an additional dependency, which I would prefer not add.

ping @ylep

[ylep]

Hi @xgui3783 , looking at DistroWatch it seems to me that no major still-supported Linux distribution is shipping with Python < 3.7, so I am fine with dropping support for Python 3.6!