[styleguide] Colour choices for Plotting + Figures

[mmcky]

Reading QuantEcon lectures is so rewarding in part because they are high quality and they adopt a style that is consistent across lectures. This discussion is to capture some ideas on how we can make colour choices in figures and plotting more consistent across lectures.

Option 1: Select colours like matplotlib selects colours from default colour cycler

One option to improve consistency is to adopt colours for charts in the same way that matplotlib chooses colours from the default colour cycler. This would mean that (in many cases) we can leave figure generation as simple as possible and let matplotlib choose colours. However in some cases this is not always possible due to the ordering of content, so following this selection cycle and setting these colours manually in this case would still improve consistency.

The main issues are:

import matplotlib.pyplot as plt
print(plt.rcParams['axes.prop_cycle'].by_key()['color'])

produces the following colours

['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728', '#9467bd', '#8c564b', '#e377c2', '#7f7f7f', '#bcbd22', '#17becf']

they are not simple colour names so using them will complicate the code

ax.plot(data, colour='#1f77b4')

These are the colours

Idea: We could define a colour map for QuantEcon (that is similar to these colours using simple colour tables such as green, blue as close as possible to the matplotlib colours)

---

Option 2:

We could specify a configuration dict at the top of the lecture to centralise parameters such as these (to make it easier to harmonise across lectures)

cfg = {
    'supply' : 'green',
    'demand' : 'orange,
}

then it would be used as

ax.plot(data, colour=cfg['demand'])

but that is not as simple to read as

ax.plot(data, colour='green')

[jstac]

Thanks @mmcky , very helpful!

I like the idea of adding a colormap to QuantEcon. To clarify your suggestion, we would assign suitable names to the colors in

['#1f77b4', '#ff7f0e', '#2ca02c', '#d62728', '#9467bd', '#8c564b', '#e377c2', '#7f7f7f', '#bcbd22', '#17becf']

that are close to the actual colors in that sequence and also easy to remember, such as orange, blue, etc., and then use them via

ax.plot(x, y, color=qe.blue)
ax.plot(x, z, color=qe.orange)

and so on. Is that the idea?

[mmcky]

thanks @jstac that is the idea I was thinking about.
I don't think we would ever have a module that needed to be called a colour so that should work at the top level.

Otherwise we could use

qe.fig.blue

but the simplicity of qe.blue is appealing. Would need to check that @oyamad would be OK with this re: package.

[mmcky]

@jstac I am also currently working on a list of "simple name" colours that may be close enough to the matplotlib colours that we can just document in the manual. If that is close enough then that might be the best approach. I will update you once I know if this is feasible.