scale_colour_manual html hex colour transparency doesn't appear in legend

[harrig12]

I'd like to set point transparency using html hex code

I don't want a separate alpha legend, just the colour one to be transparent. Although this works for the points themselves, the transparency doesn't seem to propagate to the legend when using set_colour_manual

In R, I might use override.aes but I haven't found the equivalent in plotnine

Any suggestions?

import numpy as np
import pandas as pd
from plotnine import * 

data = np.random.rand(20, 4)
df = pd.DataFrame(data)
df.columns = ['x', 'facet 1', 'facet 2', 'facet 3']
df['label'] = ['label 1']*10 + ['label 2']*5 + ['label 3']*5
df = pd.melt(df, id_vars=['x', 'label'])

g=(ggplot(df)
+ aes(x='x', y='value', color='label')
+ geom_point(size=10)
+ theme_seaborn('white')
+ scale_colour_manual({'label 1': 'orange', 'label 2': 'red', 'label 3': '#0000FF4D'})
)

[harrig12]

appears to also apply to scale_colour_identity

import numpy as np
import pandas as pd
from plotnine import * 

data = np.random.rand(20, 4)
df = pd.DataFrame(data)
df.columns = ['x', 'facet 1', 'facet 2', 'facet 3']
df['label'] = ['label 1']*10 + ['label 2']*5 + ['label 3']*5
df['colour'] = ['orange']*10 + ['red']*5 + ['#0000FF4D']*5
df = pd.melt(df, id_vars=['x', 'label', 'colour'])

g=(ggplot(df)
+ aes(x='x', y='value', color='colour')
+ geom_point(size = 10)
+ theme_seaborn('white')
+ scale_color_identity(guide='legend')
)

[harrig12]

I tracked this down!

in draw_legend, lines.Line2D() forces the alpha if provided, because it uses to_rgba internally

alpha defaults to 1 for geom_point

https://github.com/has2k1/plotnine/blob/main/plotnine/geoms/geom_point.py#L140-L149

Fix is easy enough - alpha should default to None. As far as I can tell, this doesn't break anything.
There are several geoms with default alpha = 1. I changed their defaults (except for geom_raster) with #666

[has2k1]

Fix is easy enough - alpha should default to None.

As plotnine is getting type annotations, it is more coherent to have variables initialised with values of the correct type.

The solution is to merge the alpha into fill/colour using plotnine.utils.to_rgba in the geom.draw_legend methods as we do in the geom.draw_group methods. geom_rect.draw_legend has it right, so the other cases are an oversight.

That means there should be no cases of alpha=data['alpha'] in the geom.draw_legend functions.

[harrig12]

👍 makes sense! thanks for the quick response