On 05/18/2012 09:38 AM, endolith wrote:

I'm learning about all the colormaps to update the
documentation. and it seems some
were copied from ColorBrewer and converted into continuous
smoothly-varying maps. But ColorBrewer has 3 types of schemes:

1. Sequential palettes are suited to ordered data that progress
   from low to high. Lightness steps dominate the look of these
   schemes, with light colors for low data values to dark colors for
   high data values. 2. Diverging palettes put equal emphasis on
   mid-range critical values and extremes at both ends of the data
   range. The critical class or break in the middle of the legend is
   emphasized with light colors and low and high extremes are
   emphasized with dark colors that have contrasting hues. 3.
   Qualitative palettes do not imply magnitude differences between
   legend classes, and hues are used to create the primary visual
   differences between classes. Qualitative schemes are best suited to
   representing nominal or categorical data.

So the "qualitative" maps should not be smoothly-varying.
http://rgm2.lab.nig.ac.jp/RGM2/func.php?rd_id=RColorBrewer:ColorBrewer

It's possible to convert these to segmented maps:
https://gist.github.com/2719900#file_accent.py

Should I go through them and do that and submit a patch? I'm not
sure if the matplotlib colormaps are even meant to be used for
categorical data like this.

As I read it, segmented (qualitative) maps should be generated with a
number of segments matching the (small) number of categories, so it
looks like dumping them in as if they were continuous was a mistake--but
it would equally be a mistake to convert them to segmented.
ListedColormap is more appropriate for working with a few discrete colors.

I think we are stuck with the silly things the way they are.

Eric

--- Reply to this email directly or view it on GitHub:
#881

Ok, I wasn't aware of ListedColormap. Is there a way to use these colormaps with it? All the examples I see just use an explicit list, like cmap = mpl.colors.ListedColormap(['r', 'g', 'b', 'c'])

I wonder if the binary colormap belongs in the same group. It's produces an identical gradient to gray_r, but the example included with it only uses the most extreme values of pure white or pure black. I'm not sure why it exists.

I think this issue can be closed, based on #921 having been merged.

#921 isn't related to this, but the docs change in #889 to discourage their use is. (endolith@cce4441#L0R1789)

Thank you for the correction.