Revisions for current intake-esm

[ktyle]

This PR updates the notebooks/foundations/enhanced-catalog notebook to resolve various errors related to changes in the underlying intake and intake-esm libraries. Also took an opportunity to refresh the content a bit, based on current examples in https://intake-esm.readthedocs.io/en/stable/index.html .

[github-actions]

👋 Thanks for opening this PR! The Cookbook will be automatically built with GitHub Actions. To see the status of your deployment, click below.
🔍 Git commit SHA: 4e58baf
✅ Deployment Preview URL: In Progress

[ktyle]

The build is failing but that is due to as-yet unresolved issues with the example-workflows/key-figures notebook.

[brian-rose]

@ktyle thanks for opening this! I'll take a look at the notebook tonight.

It seems that your PR commits the output of the enhanced-catalog.ipynb notebook. That's handy for now so I can easily look at it, but is it violating the spirit of our push for reproducible notebooks? Should we always strip output before committing and let the CI system do the execution (as we enforce for Foundations)?

I'm not sure we ever came up with a clear policy about this for the Cookbooks.

[ktyle]

Yes, I should have cleared the output before committing. Although I think we should have an exception for #3 ...

[brian-rose]

@ktyle I pushed a new commit that cleared the output.

Before doing this, I verified that the notebook runs cleanly on my own system. I also see that it ran cleanly on our CI system.

I really like the modifications to this notebook! Everything reads well and is clearer than the original.

I suggest we merge this update, and then open a new PR to deal with #3

[brian-rose]

I think that if we merge this, the update will not get published because of the build failure related to #3.

A short-term solution, as discussed in #3, is to commit a copy of the key-figures notebook that contains locally-rendered output.

But we will also have to change the execution settings in the config file to a format that will not try to execute that notebook.

e.g. change execute_notebooks: binder to execute_notebooks: auto
(which will not try to execute a notebook that does not have missing output)

[ktyle]

hmmm I wonder if there is a more granular way in jupyterbook to "opt out" certain notebooks from being executed.

But in this case, since there are just two notebooks in the cookbook, setting the flag to "auto" seems reasonable. I propose we merge this PR, and then create a new PR that contains the fully-executed key-figures notebook and also switches the build flag to auto.

[ktyle]

(and having fully read your comment, @brian-rose , we're on the same page so I will go ahead and merge this and then embark on a PR that deals with #3 (tomorrow morning!).

[brian-rose]

@ktyle I think that auto is the granular way, in the sense that JupyterBook will skip execution of a notebook if and only if it doesn't have any missing output.