Automatic selection of different colormaps for anomaly & intensity data

George_Nurser1 · July 2, 2015, 3:48pm

Hi,
I’ve been following the discussions about the new default colormaps.

I think it might be really helpful if the default behaviour were that matplotlib simply examined your data (Z, say), and if

(1) Z.max() & Z.min() had the same sign then used a sequential colormap

whereas if

(2) Z.max() & Z.min() had opposite signs then it used a diverging colormap, centered on zero.

Keywords e.g. Anomaly=True & centre = 10. could be extra arguments to override this behaviour.

I realise that

a) this really only requires a user to write a simple helper application.

But 99% of users will never do this.

b) it may not always be desired; but again 99% of the time it probably is.

This would enable people just starting to use matplotlib perhaps to see that it can give ‘better’ plots than matlab

Anyway, just a thought.

George Nurser.

tacaswell · July 2, 2015, 5:18pm

This has been discussed and the consensus was that getting the heuristics right is probably impossible (I work with dark-current subtracted image data so it can have negative values, but using a diverging color map is very wrong, taking 0 as the center only makes sense some of the time). Defaulting to a linear norm and sequential color map is in the worst case not helpful, where as using the wrong center (or incorrectly using a diverging color map) can be misleading. We are prioritizing not being misleading over being slightly more convenient in the core library.

A function with a call signature like you describe should definitely be in the examples.

Also see PR https://github.com/matplotlib/matplotlib/pull/3858 which is adding a new normalizer to make setting up non-symmetric diverging color maps easier.

Tom

···

On Thu, Jul 2, 2015 at 11:49 AM George Nurser <gnurser@…287…> wrote: