Using linked selection removes alpha coloring

Description of expected behavior and the observed behavior

Using linked selections on between a 2D scatterplot and the derived histogram resets the alpha color options.

• Without link_selections (no brushing is possible, alpha is respected):
• With link_selections (brushing is possible, alpha is reset):

Complete, minimal, self-contained example code that reproduces the issue

import numpy as np
import holoviews as hv

hv.extension('bokeh')


color_map = {
    'p1': 'navy',
    'p2': 'green'
}

points_dct = {
    'p1': hv.Points(np.random.randn(1000, 2), kdims=['x', 'y'], label='p1').opts(color=color_map['p1'], alpha=0.2),
    'p2': hv.Points(np.random.randn(1000, 2) * 2 + 1, kdims=['x', 'y'], label='p2').opts(color=color_map['p2'], alpha=0.2)
}
nd_overlay = hv.NdOverlay(points_dct, kdims='cat')

hist_dct = {
    k: hv.operation.histogram(p, dimension='x').opts(color=color_map[k], alpha=0.5)
    for k, p in points_dct.items()
}
hist_nd_overlay = hv.NdOverlay(hist_dct, kdims='cat')

ls = hv.selection.link_selections.instance()
# plot = ls(nd_overlay) + ls(hist_nd_overlay)  # Removes "alpha"
plot = hv.selection.link_selections(nd_overlay + hist_nd_overlay)  # Removes "alpha"
# plot = nd_overlay + hist_nd_overlay  # No slection
plot

ALL software version info

Python version       : 3.9.6

bokeh    : 2.3.3
holoviews: 1.14.4
numpy    : 1.21.0

jupyter core     : 4.7.1
jupyter-notebook : 6.4.0
qtconsole        : not installed
ipython          : 7.25.0
ipykernel        : 6.0.1
jupyter client   : 6.1.12
jupyter lab      : 3.0.16
nbconvert        : 6.1.0
ipywidgets       : 7.6.3
nbformat         : 5.1.3
traitlets        : 5.0.5

[peterroelants]

@jbednar I spend some time looking at the code and I think I know what happens and how to change it.

Where the defaults come from:

• The unselected alpha default is set at selection.py#271
• The selected alpha default is set at selection.py#622

The unselected alpha can actually be passed as a custom unselected_alpha parameter to link_selections.instance() or link_selections constructor, which calls .instance(). .instance() has a **params argument that passes all keyword arguments along to the ParameterizedFunction base class.

If we want to multiply with the original alpha of the plot we probably want to extract the original value at OverlaySelectionDisplay._apply_style_callback(). However, I’m not sure myself how I can (and if it’s possible) extract the original alpha value from the element at this place. I had a look and couldn’t find any alpha parameter on the element.

If we want to support selected_alpha and unselected_alpha parameters then I think we also have to modify the _Styles stream, which seems to carry the unselected_alpha setting alpha=inst.unselected_alpha . This stream is further passed in as the selection_streams variable to OverlaySelectionDisplay.build_selection(), where this selection stream is further applied to the layer via OverlaySelectionDisplay._apply_style_callback(). It’s in this last method that the alpha value is being overwritten with alpha = 1 if the layer is not the first layer.

I might work on a PR updating _Styles sometime in the future.

[matthieugross-bib]

Hello, I’m facing the same “issue” (if we can call it like that). Any progress on how to handle the alpha ? Or at least a workaround?