viridis as default colormap for density plots?

[JS3xton]

The colormaps here: https://bids.github.io/colormap/ seem well thought out. Looks like they already exist in matplotlib 1.5 and will be the default in matplotlib 2.0, so using them might eliminate a dependency on palettable.

Looks like we're currently using palettable.colorbrewer.diverging.Spectral_8_r.mpl_colormap, which may achieve the same things that the ones above do, I don't know. I'm also not sure how complicated it is to support the colormaps mentioned above during this transition period. May be better to wait until they're the matplotlib default, and then update FlowCal to embrace them?

[castillohair]

I think the first step would be to generate a density plot and a couple of clustering plots, to see if we like the result and we want to go further with this. I personally really like the current clustering plots. The different colors indicate different discrete groups, but there is also an order associated with them. So I wouldn't care about perceptual uniformity here. But I can see the benefit of changing the 2D density plots.

There is also the issue of the standard curve plots, which would revert to the plain blue, red, and green used by matplotlib by default. This means that we would still need palettable for this plot. They are changing the default styles in matplotlib 2.0 (http://matplotlib.org/style_changes.html), but I can't find specifics. You might wanna spend some time in the github repo (https://github.com/matplotlib/matplotlib) and see if you can find something. I'd like to have this information and see a sample plot before proceeding any further.

[castillohair]

On the standard curve plots, it turns out matplotlib includes styles besides the classic one. This includes some styles from seaborn and ggplot. Some of the seaborn styles are similar to what is currently done for the standard curves. See http://matplotlib.org/users/style_sheets.html.

[castillohair]

Now that some time has passed and I have some experience with viridis, I have to say that I'm warming up to this idea. Defaults in matplotlib > 2.0.0 are pretty nice and could justify ditching palettable. If anyone wants to do this, I'd be interested. I probably won't do it, though.