Colorscale/colorbar fixes for Plotly backend

[jonmmease]

This PR includes various fixes related to colorscale/colorbar handling for the plotly backend

Commit summary:

• 8eb78b0: This addresses cmap for Surfaces not working with plotly #3414 by limiting the length of colorscales passed to plotly.js to a maximum of 255. Longer colorscales are subsampled. Plotly.js performs its own colorscale interpolation so I don't foresee this causing visual artifacts for any reasonably smooth colorscale.

import numpy as np
import holoviews as hv
from holoviews import dim
from holoviews import opts
hv.extension('plotly')

surface = hv.Surface(np.sin(np.linspace(0,100*np.pi*2,10000)).reshape(100,100))
surface.opts(cmap='fire', height=500, width=500, colorbar=True)

• 1ffdef1: Fix for the case where a named colorscale is known only to plotly.py. Previously this resulted in a malformed colorscale.

surface = hv.Surface(np.sin(np.linspace(0,100*np.pi*2,10000)).reshape(100,100))
surface.opts(cmap='Picnic', height=500, width=500, colorbar=True)

• 9cc745b: Fixes colorbar and edges_color handling. Both of these options are applied to separate traces (separate from the main mesh3d trace) which were being dropped. With these changes, an Element may be represented by a series of traces instead of only 1.

y,x = np.mgrid[-5:5, -5:5] * 0.1
heights = np.sin(x**2+y**2)
trisurface = hv.TriSurface((x.flat,y.flat,heights.flat))

trisurface.opts(height=500, width=500, cmap='plasma', colorbar=True, edges_color='cyan')

• 016a137: Add/fix colorbar/colorscale handling for the plotly bivariate element. Previously, the cmap and colorbar args were ignored.

normal = np.random.randn(1000, 2)
biv = hv.Bivariate(normal)
biv.opts(filled=True, cmap='cividis', showlines=True, colorbar=True)

normal = np.random.randn(1000, 2)
biv = hv.Bivariate(normal)
biv.opts(filled=False, cmap='plasma', showlines=True, colorbar=True)

• I tried out the colorbar/colorscale handling of other Elements and the rest of them worked fine.
• 8fb44d8: Adds a few additional colorbar test cases

[philippjfr]

This is great, thanks a lot! Subsampling the color scales sounds good to me as well, no reason to send the full palettes unlike in bokeh where interpolation is not implemented client-side.

[philippjfr]

Not sure what's up with the tests yet, seems unrelated.

[jonmmease]

Thanks! I'm taking a look at the tests as well.

[philippjfr]

Ah, I know what happened, the new param release dev release changed the formatting of warnings. I'll fix that in a separate PR shortly.

[jonmmease]

Ah, I know what happened, the new param release dev release changed the formatting of warnings. I'll fix that in a separate PR shortly.

Thanks for the detective work 🙂

[philippjfr]

Hopefully fixed in #3591

[philippjfr]

Thanks again @jonmmease.