Floating point error in plotting.bokeh.colormap_generator

[palmhjell]

I came across a subtle bug in the way plotting.bokeh.colormap_generator normalizes colormaps with the default arguments, such that something like hv.Palette(f'Category{N}', samlples=N) will not return every color in the colormap. Example here for N=10:

Two pinks (#e377c2) are returned and gray (#7f7f7f) is missing.

This can be fixed by adding a very small term to the upper value of the range argument (e.g., (0, 1+1e-10)), but is mainly related to floating point error in what plotting.bokeh.colormap_generator returns:

Clearly that value is supposed to be 7 but int(6.9999...) is 6.

I played around with other types of creating that integer (e.g., round), but the best method I found that always keeps the same expected behavior for all colormaps and also fixes these float errors is to just add a small term in the colormap_generator function:

def colormap_generator(palette):
    # Account for potential floating point error
    error = 1e-10
    return lambda value: palette[int(value*(len(palette)-1)+error)]

This fixes the issue and returns the expected behavior:

I've got a fork with this fixed and can open a PR, but it's honestly a 1/2-line fix, so I figured an actual dev could just make that change easily enough.