Provide Jupyterlite as an option where possible

[wragge]

Wherever possible add Jupyterlite as another option for running notebooks. In the first instance focus on notebooks that will require minimal code changes, for example, the Getting Started notebook and the Trove API intro notebooks.

Also look at heavy-use notebooks such as Trove newspaper harvester and save newspaper images notebooks -- these will require more code changes but but make things a lot easier and quicker for users.

[wragge]

To minimise code changes use pyodide-http to patch requests. Eg will just need to include something like:

%pip install -q pyodide-http requests

import pyodide_http
pyodide_http.patch_all()  # Patch all libraries

Then requests, or libraries using requests can be imported and used as normal.

[wragge]

Note that the most recent version of pyodide-http (0.2.1) fixes a problem with Firefox in Jupyterlite. However, Pyodide currently packages an earlier version of pyodide-http (0.2.0). There doesn't seem to be any easy way of overriding this, so I'll need to wait for Pyodide to be updated to avoid problems with Firefox.

[wragge]

A quick investigation indicates that I'll probably have to make some changes to the trove-newspaper-harvester package. In particular, there's a problem with the way requests-cache creates new databases in the Jupyterlite filesystem. It looks like a way around this is to create the db in /tmp/. This is probably a good idea anyway.

[wragge]

See also Voici as a JupyterLite based replacement for Voila.

[wragge]

Requests is supported now, so the code changes should be minimal. The main problem now is going to be dealing with CORS issues. CORS implementation in v3 of the Trove API is currently broken. Even when (if) that gets fixed, I won't be able to scrape from web pages etc to get extra metadata. I really don't want to have to use a CORS proxy...