Improve usability

[constantinpape]

• add conda and pip environment files for necessary dependencies (probably just zarr and skimage)
• download and unpack the example data automatically if it's not found

[sbesson]

Feel free to kick this discussion into a separate issue. As this raises questions around packaging and distribution, it made me wonder about the future of this repository and its relationship compared to https://github.com/ome/ome-zarr-py.

Trying to summarize my understanding of the similarities and differences:

• both code bases are Python-based, use similar dependencies, have support for writing multiple versions of OME-NGFF and include utilities for generating sample OME-NGFF datasets.
• ome-zarr-py aims to be a reference library with a stable writing and reading API while this repo is more lightweight and aimed at fast prototyping of new specification samples
• the synthetic examples in ome_zarr.data are based on skimage.data while the examples of this repository are derived from datasets associated with published manuscripts (similarly to the OME-NGFF examples derived from IDR datasets).

From the ome-zarr-py perspective, I would consider it valuable to extend the ome_zarr create command and support the generation of a suite of sample datasets. Interesting to brainstorm on how to re-use functionalities.

[constantinpape]

Thanks for bringing this up @sbesson. But the use-case of this repository is different from ome-zarr-py: the purpose is to be able to quickly prototype a writer for a new version without the need to integrate it with the rather complex class structure in ome-zarr-py (which I found pretty hard to grasp when first working with it). It is not meant as a reference implementation or implement general purpose writers or any reader functionality. So I would vote for just leaving this repository as it is (with some minor improvements to usability so that others can continue creating prototypes for new versions with it.)

• the synthetic examples in ome_zarr.data are based on skimage.data while the examples of this repository are derived from datasets associated with published manuscripts (similarly to the OME-NGFF examples derived from IDR datasets).

From the ome-zarr-py perspective, I would consider it valuable to extend the ome_zarr create command and support the generation of a suite of sample datasets. Interesting to brainstorm on how to re-use functionalities.

I agree that it would be nice to have more example data functionality in ome-zarr-py, e.g. in ome_zarr.data. Maybe you could just re-use some of the exmaple data from here or from IDR? Feel free to host the data I am using here somewhere else if that's necessary for it.