plot/palette proposal: remove rainbow, add palettes from http://www.kennethmoreland.com/color-advice/

[Chris Tessum]

The problems with the rainbow color palette are well documented (see, for example, here, here, and here). However, in plot/palette it is one of the two top-level, ready-to-use options, basically making it a default. To the extent that one goal of the plotting package is to encourage the use of best practices in visual displays, I think the rainbow color palette should be removed, or at least moved to a less conspicuous location.

The colors in plot/palette/brewer are great, but it could be argued that with a maximum of 11 colors, they are better suited for plotting categorical variables rather than continuous variables. For that reason, and also because it could be good to have color maps specifically designed for "pseudocoloring ... a scalar field", I also propose adding the color palettes described at http://www.kennethmoreland.com/color-advice/ and at the bottom of http://www.kennethmoreland.com/color-maps/ The rationale behind the chosen color palettes is described in manuscripts linked from the two pages.

Thoughts?

[Dan Kortschak]

Also https://www.youtube.com/watch?v=xAoljeRJ3lU is good.

The Heat, Rainbow and Radial are all R colour palettes, so they could
conceivably be put into something like .../palette/rpalettes where they
can sit quietly and think about what they've done.

When I was writing brewer originally, I did look into putting together a
curve-fitted equivalent for it, but absence of time and the adequacy of
brewer for my uses got in the way. If you want to write that, I think it
would be great.

The paper I was considering at the time is here
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.604.3427&rep=rep1&type=pdf but if the matplotlib approach is usable considering licensing, it might be worth looking at. (It looks like it should be https://github.com/matplotlib/viscm and https://github.com/njsmith/colorspacious).

Another thing to think about is that it's not clear if HSVA should be
exported.

Dan

On Sat, 2016-07-09 at 09:57 -0700, Chris Tessum wrote:
> The problems with the rainbow color palette are well documented (see, for
> example, here

> <https://cdn.mprog.nl/dataviz/excerpts/w4/Borland_Rainbow_Color_Map.pdf>,
> here
> <ftp://sidads.colorado.edu/pub/ppp/programmers/EOS_20051005_Light_Bartlein_End_of_Rainbow.pdf>,
> and here
> <http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=736450&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D736450>).