Also, it's kind of a shame that matplotlib is still so deeply ingrained in the Python data ecosystem.

Don't get me wrong, the general concept is great. Shareable, interactive code snippets are awesome. But for a demo article like this I want a single python file that I can run and immediately see the results. I don't want to have to spin up a jupyter or pylab instance (I've never used pylab so not sure if it works the same EDIT: pylab is analogous to matplotlib, not jupyter. lesson learned). I just want to run the damn code.

On a bit of a tangent here but I also hate the way Jupyter makes git diffs absolutely unusable.

Now, the above being said, the fact that sites like Github have native Jupyter functionality is awesome. It'd be a lot less painful if they'd (the author) have linked to a repo that we (I) could run.

--

For example, first I copied the initial code snippet that defines the xkcdify function. Then I ran it with python3, and realized it needed python 2 due to the urllib2 dependency (this is not a big deal since the article is from 2012). Then I ran it again with python2, and realized I don't have numpy/scipy etc installed for python2, so I pip installed those. Then I copied in the following code snippet that generates the plot. I then ran it again, and it still didn't work. Finally after a brief google search, I realized I needed to put pylab.show() at the end since I didn't have the %pylab inline or whatever that command is since I'm running it with "pure python".

Honestly, it really wasn't _that_ much effort, but I vastly would prefer to have a demo like this given as a single python file, with the dependencies clearly specified in the blog post.

Last year I found the template that makes everything look hand-drawn. It changed the way my brain saw the diagram, going from “this thing must be absolutely pixel perfect” to “this is something I literally sketched on the back of a napkin”.

This allowed my brain to create the thing, without worrying about whether everything was lined up just so. Then, at the end, I can choose to flip the theme back to straight edges, and line everything up if I choose.

Actually, many times I just left it as it was. Others seemed to like the hand-drawn look, which came as a surprise.

In a same vain: when producing content using latex ( in the eighties ) I had a problem that papers under construction looked way to good ( lie as if was printed) ; in order to counter the good looks of the printed paper I choose to use a typewriter like blurred ink dropped font to ensure that what was printed was a draft and still under construction; producing graphs that show intent but not precision is very useful.

Now it is built-in matplotlib: https://matplotlib.org/api/_as_gen/matplotlib.pyplot.xkcd.ht...

https://yuml.me/diagram/scruffy/class/draw

http://jakevdp.github.io/blog/2013/07/10/XKCD-plots-in-matpl...