Automatic test notebooks

[caspervdw]

As of the automatic doc build in trackpy (soft-matter/trackpy#362), there is an opportunity to run the notebooks on each trackpy build, using:

jupyter nbconvert --to=html --execute notebook.ipynb

In view of upcoming API changes, this might be a good idea to ensure the working of the tutorial notebooks.

However, we have to think about dependencies of the notebooks, and maybe cull down the steps that take a long time to run (for instance, batch locate in the walkthrough)

[danielballan]

I think the long running time could become a drag, and culling the steps sounds complex. As long as the notebooks get run at least once before a tag, I don't think it's critical to run them every time. If our test suit eisn't good enough to catch API breaks that would break the notebooks, we should fix the tests.

[caspervdw]

Ok agreed.

Another thing: why don't we just link to the tutorials via nbviewer? (like this ). Seems a lot easier to me .. We could include the ipynb on the versioned doc repo, and link the dev version directly to the github versions.

[danielballan]

Ha. Good point. I think our current approach predates nbviewer.

I'm actually meeting with some Jupyter devs this week to discuss better approaches to notebook+sphinx integration, including potential new features like interactivity. There are many excited but half-baked projects in that space, and it sounds they want to settle on a standard.

Some of this is relevant to day-job work. I'll report back if it could apply to trackpy too.