Change how upper limit is taken for color_levels

[WesleyTheGeolien]

Fixes #4897

I believe that the offset is due to the N parameter in the plotting.utils.color_intervals code and how it is used if there is a clip.

Because of the int(round(...)) we can end up with more values in the camp than the N parameter. So when a slice is taken we should use the Len(cmap) rather than N which introduces an offset if len(camp) != N

[philippjfr]

Thanks @WesleyTheGeolien! Will have to update the related tests:

FAIL: test_color_intervals_clipped (holoviews.tests.plotting.testplotutils.TestBokehPaletteUtils)
FAIL: test_path_continuously_varying_color_op (holoviews.tests.plotting.bokeh.testpathplot.TestPathPlot)

[philippjfr]

Also not totally following the correctness of this fix, wouldn't it be more correct not to round and to use floor instead?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.47%. Comparing base (f39e0a0) to head (0a00be8).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4898      +/-   ##
==========================================
- Coverage   88.48%   88.47%   -0.01%     
==========================================
  Files         323      323              
  Lines       67640    67639       -1     
==========================================
- Hits        59848    59846       -2     
- Misses       7792     7793       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[hoxbro]

With this change the reported problem now looks like this:

Respecting the limit, and also better colorbar axis.