You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The same thing happens with the custom norm at the end of the tutorial. I'm not sure about TwoSlopeNorm, but at least for the custom one, the 0 is definitely not supposed to be in the middle. This half broke when merging #20054, with the 0 tick at the old place but the 0 colour in the middle. Then with d1c5a6a, the 0 tick also moved to the middle.
The tick moving to the middle is expected - colorbars are supposed to space equally in colourspace (i.e. consider log norm).
The -1000 is more problematic, as that is simply incorrect I believe. Part of the issue here is that two-slope norm doesn't have a scale, and hence doesn't have good ticking.
Activity
jklymak commentedon Jul 6, 2021
The tick moving to the middle is expected - colorbars are supposed to space equally in colourspace (i.e. consider log norm).
The -1000 is more problematic, as that is simply incorrect I believe. Part of the issue here is that two-slope norm doesn't have a scale, and hence doesn't have good ticking.
jklymak commentedon Jul 6, 2021
... the Custom norm is in even worse shape ;-)
jklymak commentedon Jul 6, 2021
OK, the -1000 is because TwoSlopeNorm didn't extrapolate at all, and so -1000 == -500. Adding extrapolation fixes that.
The
MidpointNormalizeexample was broken becauseMidpointNormalizedid not define an inverse. Defining an inverse also fixes this.