Wow I have to agree with this (even though the post is using plotly). I don't know where the culture difference comes from between the R and Python ecosystems that makes the latter so much worse than the former when it comes to graphics. Python seems to be able to do so many things so well, but then at the final output stage it all falls apart and I find myself back in R. On that point, I even wonder if with all the Julia bruhaha, the actual real competitor in data science might actually be Javascript! Much faster than Python, and a community that truly understands graphics quality (I'm looking primarily at D3 but things like Mathbox.js and even Hicharts are pretty decent). I'm thinking that SVG and increasingly Webgl will incentivize linear algebra skills in the Javascript ecosystem, and those same skills will suddenly be transferrable to data.

MPL's documentation is a catastrophe, and the API design is indeed abysmal. The plots, however, are only as ugly as you like them to be. Here's a great auxiliary package with much more sensible defaults:

https://github.com/mwaskom/seaborn

Try bokeh http://bokeh.pydata.org/en/latest/