use latest pangeo docker image

[rabernat]

Attempting to resolve the catalog build error in #58.

[rabernat]

@davidbrochart & @martindurant - I tried updating the docker image we are using to validate our intake catalog, and I am now getting some errors with the hydrosheds catalogs.

Any chance you can look into this and figure out why they are failing?

[davidbrochart]

Trying to create a fresh conda environment, rasterio doesn't seem to install anymore.

[davidbrochart]

Might be related to conda-forge switching to Python 3.8.

[davidbrochart]

With the following environment:

conda install python=3.7 ipython intake intake-xarray rasterio gcsfs

That works:

import intake
cat_url = 'https://raw.githubusercontent.com/pangeo-data/pangeo-datastore/master/intake-catalogs/master.yaml' 
cat = intake.Catalog(cat_url)
cat.hydro.hydrosheds_acc.to_dask()

[davidbrochart]

Looks like a (temporary) Google authentication problem to me.

[jhamman]

I restarted the build here (and updated to the latest image). Neither worked to resolve this. I'm leaning against the temporary auth problem but I don't know where else to look.

[davidbrochart]

The tests that are failing are related to the rasterio driver. Back then we worked around the issue by caching the data locally: intake/intake-xarray#31 (comment)
Maybe this doesn't work any more?

[martindurant]

There is no particular reason caching by intake doesn't won't work, although you may find things simpler when using the whole-file caching interface of fsspec directly "filecache".

[davidbrochart]

It would be nice to have the pinned version of packages to be able to reproduce.