Run nbval on documentation notebooks to check documentation is up to date

[fangohr]

Might be as simple as making sure the target test-ipynb https://github.com/computationalmodelling/fidimag/blob/master/Makefile#L91 is run as part of the continuous integration on travis.

[This should always have happened.]

A bug was reported that could have been avoided given the feedback from nbval (#131)

[rpep]

I turned this on, but because the notebooks take a lot of time to run, Travis times out. Do you know how to stop this?

[fangohr]

Thanks!

You can run the nbval tests on your own machine and get py.test to display the n longest running jobs (in reverse execution time order). Typically there are only a few notebooks or even cells that take a long time. We can exclude cells from execution using the nbvalskip tag.

That may be a simple solution - not perfect but might be easy to achieve and keep most of the benefit.

[rpep]

Working through these now:

• 1d_domain_wall.ipynb
• FMR-stdprob.ipynb
• grain-structure.ipynb
• current-driven-domain-wall.ipynb
• isolated_skyrmion.ipynb
• skyrmions-applied-field-gradient.ipynb
• spin-polarised-current-driven-skyrmion.ipynb
• spin-waves-in-periodic-system.ipynb
• standard_problem_4.ipynb
• steepest-descent-atomistic.ipynb
• tutorial-basics.ipynb
• tutorial-docker-container

[rpep]

This is now running all of the notebooks using a shell script workaround that prints an output every 9 minutes to keep happy. There are various issues with the notebooks themselves. I've fixed a couple but not everywhere, because some notebooks take a long time to run so I've been doing this in the background while working on other things.

It seems that the Matplotlib checking in Nbval must have changed form, because all of the
cells with matplotlib figures fail, despite the text appearing to have been sanitised.

[fangohr]

Thanks, Ryan.

With Matplotlib many such tests haven broken with the change from version 2 to 3. If that is the reason, we should rund and save the reference notebook with version 3, and that should fix it.

[rpep]

#153 should make unnecessary