Hacker News Reader

Comment by geoduck14 | Hacker News Reader

I was using matplotlib yesterday - it is pretty rough. The style sheets will be helpful to make the colors better - but I also struggled with generating line thick enough to be meaningful, showing out enough to see the chart, and labeling data

View on HN

It has its issues, but as someone who has used a Lot of MATLAB I will say the built-in colormaps of matplotlib are far superior. For personal stuff it's fine to rely on file exchange colormaps, but I feel less great about using external code in my academic and government work environments.

My quick and dirty trick is to set an alpha=0.8 to soften the default colors. It always looks better.

I don't mind the default colormap (parula). If I'm using the color for any sort of dimension then I start to care about the colormap choices. Perceptually uniform and colorblind agnostic choices are important for any sort of publication, and arguably a starter for any sort of accessibility considerations.

Take a look at colorcet. https://colorcet.holoviz.org/

The four uniform colormaps that the page references are the value that I'm referring to. Viridis, plasma, inferno, and magma. I don't want a huge selection of colormaps, just a few good ones.

Here is where to get them on the file exchange. This really ought to just be folded into MATLAB proper releases somehow.

https://www.mathworks.com/matlabcentral/fileexchange/51986-p...

Indeed that does make charts look better. I do this for grid line:

ax.set_axisbelow(True) ax.grid(True, alpha=0.3)

Reply on news.ycombinator.com