Setting name of scale_color_continuous legend

[jwhendy]

Greetings,

I was puzzled by this on a recent plot. Here's a reproducible example:

import pandas as pd
import random
from plotnine import *

random.seed(1)

df = pd.DataFrame({'x': [random.randrange(100) for _ in range(100)],
                   'y': [random.randrange(100) for _ in range(100)],
                   'c': [random.randrange(5) for _ in range(100)]})
p = ggplot(df, aes(x='x', y='y', color='c')) + geom_point()
p

By default, we get a viridis continous color scale:

I wanted to change the name, so I added this line:

p = ggplot(df, aes(x='x', y='y', color='c')) + geom_point()
p = p + scale_color_continuous(name='foo')
p

I got this error instead:

ValueError: 'foo' is not a valid value for name; supported values are [list of color palette names]

For some silly reason I thought to try the european R ggplot spelling instead, which works but changes the color scale...

p = ggplot(df, aes(x='x', y='y', color='c')) + geom_point()
p = p + scale_colour_continuous(name='foo')
p

How can I have both a viridis palette and a custom scale name? I'm sure this is something trivial I'm overlooking, but I haven't figured it out! Many thanks.

[has2k1]

You have hit upon two issues.

1. scale_color_continuous and scale_colour_continuous do not refer to the same scale.
2. The name parameter for scale_color_continuous which is an alias of scale_color_cmap refers to the name of the colormap and not the name of the scale.

You can use this work around.

import pandas as pd
import random
from plotnine import *

random.seed(1)

df = pd.DataFrame({'x': [random.randrange(100) for _ in range(100)],
                   'y': [random.randrange(100) for _ in range(100)],
                   'c': [random.randrange(5) for _ in range(100)]})
p = ggplot(df, aes(x='x', y='y', color='c')) + geom_point()
p + guides(color=guide_colorbar(title='foo'))  # <- Title of the colorbar

[jwhendy]

Thanks for clarifying. Any particular reason this is like that? I'm guessing it's due to passing on name to matplotlib? I ask as I'm curious if there's any interest in standardizing, for example having name always refer to the legend title as with other scales and, for example, colormap becomes the new arg name for scale_color_cmap instead if name.

Obviously this is a breaking change, but just throwing it out there. It would seem that 1:1 parity with R's ggplot would be great, and I can't recall any analogs like this off the top of my head (where two arguments with the same name mean different things for different scales).

Thanks for the workaround. My current solution was just to rename the column directly, but this is easier.

[has2k1]

Any particular reason this is like that?

No, just an oversight which came about because when making a scale for the colormap palette cmap_pal, I used the same parameter names.

The think new parameter for the scale will become cmap_name.