xpublish-file-metadata: A simple dataset router plugin to serve underlying non-zarr file metadata (help wanted)

[xaviernogueira]

Hi all.

I quickly spun up this repo: https://github.com/xaviernogueira/xpublish-file-metadata which stores a dataset router plugin for serving non-zarr metadata. See the README for a full description.

Some key points:

• One can hide attributes based on file type, or globally for all files. I took care to implement this safely (i.e., making sure the cache can't be exploited), but wouldn't mind a second glance.
• So far I have implemented Netcdf and GeoTIFF. Next, I am implementing HDF5. Finally, I will do GRIB, however, the cfgrib library has non-windows binary dependencies making the installation a bit more convoluted.
• I used a sub-plugin design such that one can download only the dependencies one needs for the file types they intend to serve.
• Since many data catalogs contain a variety of file types, I didn't follow the datasets/{dataset}/file-metadata/{format}/... pattern but rather dropped the "format" path parameter in exchange for a unified response type regardless of what the underlying file format is. One can see what the underlying format is via file-metadata/format. This is different than the Zarr plugin implementation, but I believe it makes the most sense rather than documenting 4 endpoints (one for each file type) where only one would work.

I have a CI/CD yaml ready to publish this to PyPi as soon as I make a release. Before I do I was hoping to get some feedback. I don't have too much time to spend on this, so feedback that requires major re-write I will ask that one makes a PR instead.

I am also happy to transfer ownership to xpublish-community once things are stable.

Thanks! Next up is the xpublish-stac provider plugin...