如何在Holoviews热力图中为每个点指定自定义颜色（含透明度）

I wanted to generate a HeatMap in Holoviews where each point has custom transparency levels. Initially, I tried overlaying two HeatMap objects (one with reduced opacity, another filtered to show fully opaque points), but this resulted in an Overlay instead of a single HeatMap instance:

import pandas as pd
import holoviews as hv
hv.extension('bokeh')

data = pd.DataFrame([(i, 97+j, i*j) for i in range(5) for j in range(5)], columns=['x', 'y', 'val'])
data_filtered = data[(data.x < 3) & (data.y < 100)]
hm_opts = dict(kdims=['x', 'y'], vdims=['val'])
hm = hv.HeatMap(data, **hm_opts).opts(alpha=0.5)
hm_filtered = hv.HeatMap(data_filtered, **hm_opts).opts()
hm * hm_filtered

I also tried mapping each (x,y) coordinate to a hex color with built-in transparency and adding a color column to my DataFrame, but ran into issues: passing the color list to cmap treated it as a continuous color range, passing the column name threw errors, and using the color parameter made the chart fail to render entirely.

I did manage to get this working directly with the Bokeh backend, but wanted to use Holoviews for building interactive applications:

from bokeh.plotting import figure, show
from bokeh.models import ColumnDataSource
from bokeh.models.ranges import FactorRange

source = ColumnDataSource(data=data)
x_unique = data['x'].unique()
y_unique = data['y'].unique()
min_width = 110
min_height = 80
width = min_width + 25 * len(x_unique)
height = min_height + 25 * len(y_unique)
x_rect_width = 0.90
y_rect_width = 0.90
plot = figure(
    plot_width=width, plot_height=height, title='',
    x_range=FactorRange(*x_unique), y_range=FactorRange(*y_unique),
    x_axis_label='x', y_axis_label='y',
)
plot.rect('x', 'y', height=y_rect_width, width=x_rect_width, source=source, color='color')
plot.xgrid.grid_line_color = None
plot.ygrid.grid_line_color = None
show(plot)

Working Holoviews Solution #

With help from @thomas-pedot, I found the proper way to achieve this with a single Holoviews HeatMap by binding the alpha value directly to a column in the DataFrame:

import pandas as pd
import holoviews as hv
hv.extension('bokeh')

data = pd.DataFrame([(i, 97+j, i*j) for i in range(5) for j in range(5)], columns=['x', 'y', 'val'])
# Add an alpha column with custom transparency values
data = data.assign(alpha=((data.x < 3) & (data.y < 100)).replace({True: 1.0, False: 3/8}))

# Define custom color levels and corresponding colors
red = '#FF0000'
yellow = '#FFFF00'
green = '#00FF00'
blue_violet = '#8A2BE2'
max_count = data['val'].max()
levels = [0, 1, max_count / 2, max_count - 1, max_count]
colors = [green, yellow, red, blue_violet]

# Include 'alpha' in the value dimensions
hm_opts = dict(kdims=['x', 'y'], vdims=['val', 'alpha'])
hm = hv.HeatMap(data, **hm_opts).opts(
    alpha=hv.dim('alpha'),  # Bind alpha to the DataFrame column
    color_levels=levels,
    cmap=colors
)
hm

Key Details: #

• Add an alpha column to your DataFrame with values between 0 (fully transparent) and 1 (fully opaque)
• Include alpha in the vdims parameter so Holoviews recognizes it as a value dimension
• Use hv.dim('alpha') in the options to dynamically map each point's transparency to the column value
• Define custom color levels and a corresponding colormap to control the fill colors independently of transparency

内容的提问来源于stack exchange，提问作者iRave