Re: default colormap

[Carnë Draug]

On 4 November 2015 at 12:16, Ben Abbott <address@hidden> wrote:
>
> Mathworks recenlty changed a lot of things when it introduced its
> new graphics system.
>
> http://blogs.mathworks.com/loren/2014/11/05/matlab-r2014b-graphics-part-3-compatibility-considerations-in-the-new-graphics-system/
>
> I’m not sure what “compatible” means at the moment. The pre-2014b
> Matlab and post-2014b Matlab are not compatible with each other.
> [...]

I agree with you.  I feel that since they broke backwards compatibility,
it gave us some leeway to have a different default.

On the specific case of colormaps, even if only targeting Matlab, a
user will either have to be specific on the colormap or be willing
to accept anything.  Assuming this, we are free to choose our default.
The only thing needed for Matlab compatibility being having parula
implemented.

And parula doesn't look that different from viridis.  I'll guess the
typical user would be happy with any.  But listening to the experts
(biased, since they're the authors of viridis), there seems to be good
reason to favour viridis.  They gave a talk about viridis and why it was
choosen for matplotlib default.  It's quite interesting

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

Basically, both viridis and parula come up with a colormap that
maintains a monotonically increasing perception of intensity
(similar in idea to cubehelix which we already have in Octave).
The main difference between parula and viridis is that parula
follows an outdated model for the color intensity perception (Lab)
while viridis uses the most up to date model (CAM02-UCS).
(see 10:47--12:07 and 14:00-14:30 of the youtube video above).

Carnë