SymLogNorm

[pordyna]

Hi!
I really love working with scipp and plopp. But if there is one thing that is really missing for me, then it would be a symmetric log norm. SymLogNorm centered around 0 and combined with a diverging color map is just great for visualizing datasets with positive and negative values, like e.g., vector field components.

Do you currently plan to work on such a feature? I could also imagine working on it myself, with just a bit of initial guidance from your side.

Also, is there maybe a relatively easy way of working with custom norms, that does not require switching completely to native matplotlib and doing everything oneself? Sth that I could use as a workaround.

Cheers
Paweł

[nvaytet]

Hi, adding a SymLogNorm would indeed be useful.
It would certainly be straighforward to support it as an option for the norm argument when making a plot.
The part I am not entirely sure about is what to do when we have interactive figures and one clicks the log button in the toolbar.
Currently, the interactive switching only supports linear or log as a 2-state button.

I don't know if we could make a 3-state button or something? Should the log option be a dropdown menu instead? Should the same also apply to the vertical axis when making a 1d plot?

Apart from all these questions, a workaround for now could be to set the normalizer on the colormapper, and re-create the colorbar.
Assuming you have a 2d figure named f, it would be something like:

from matplotlib.colors import SymLogNorm
from matplotlib.colorbar import ColorbarBase

cmapper = f.view.colormapper
cmapper.normalizer = SymLogNorm(linthresh=1.0e-4)
cmapper.colorbar = ColorbarBase(cmapper.cax, cmap=cmapper.cmap, norm=cmapper.normalizer)
f.view.render()
f

I have not extensively tested this, so I am not sure it all works properly. You'd have to double-check that the colors correspond to the expected values, but at first glance it seems to agree with what raw MPL produces.

[pordyna]

Ok, so the workaround is exactly what I was looking for, so thanks for that. About the widget, I'm not sure what can be done. I don't like the 3 state button, because that would mean re-rendering a figure twice sometimes instead of just once.