Is it possible to use custom strings for legend labels?

Is it possible to pass a custom list of strings for the legend labels? In a chart, I want to show missing values as a part of the total so I use a stacked barchart where the segments/pieces(?) come from two columns: Missing and Difference , combined together they add up to the total. The legend shows the two variables as Missing and Difference and I would like to be able to change Difference -> Total somehow.

df = pd.DataFrame(
  {'Year': [2014, 2015, 2016, 2017, 2018, 2019],
   'Total': [376, 1198, 1871, 4386, 4920, 2417],
   'Missing': [299, 758, 1063, 233, 32, 11],
   'Difference': [77, 440, 808, 4153, 4888, 2406]}
)

melted = pd.melt(df, id_vars=['Year'], value_vars=['Difference', 'Missing'])


alt.Chart(
    data=melted,
    background='white'          
).mark_bar(
).encode(
    alt.X('Year:O', title=''),
    alt.Y('value:Q', title=''),
    alt.Color('variable:N', 
              scale=alt.Scale(domain=['Difference', 'Missing'],
                              range=[sand, red])),
    order=alt.Order('variable', sort='descending')
)

In the plot below I have the chart I want but the Legend says Difference and I would rather it say Total. I know I can change the original column names or the values within the melted dataframe, but I was hoping for a cleaner way to change the legend labels.

[jakevdp]

You can use the legend values setting to specify a subset of the range to show in the legend; you can’t use it to change the scale range to contain different values. Here’s an example of how it’s intended to be used:

import altair as alt
import pandas as pd

df = pd.DataFrame({'Year': [2014, 2015, 2016, 2017, 2018, 2019],
 'Total': [376, 1198, 1871, 4386, 4920, 2417],
 'Missing': [299, 758, 1063, 233, 32, 11],
 'Difference': [77, 440, 808, 4153, 4888, 2406]}
)

alt.Chart(df).transform_fold(
    ['Total', 'Missing'],
    as_=['variable', 'value']
).mark_bar().encode(
    alt.X('Year:O', title=''),
    alt.Y('value:Q', title='', stack=None),
    alt.Color('variable:N', 
              scale=alt.Scale(domain=['Total', 'Missing'],
                              range=['#ddccbb', 'red']),
              legend=alt.Legend(values=['Total'])),
    order=alt.Order('variable:N', sort='descending')
).properties(
    background='white'
)

[jakevdp]

Another more slick option would be to actually use your Total entry and set stack=None to create an unstacked bar chart (here I use a fold transform in the chart rather than a pd.melt external to the chart)

import altair as alt
import pandas as pd

df = pd.DataFrame({'Year': [2014, 2015, 2016, 2017, 2018, 2019],
 'Total': [376, 1198, 1871, 4386, 4920, 2417],
 'Missing': [299, 758, 1063, 233, 32, 11],
 'Difference': [77, 440, 808, 4153, 4888, 2406]}
)

alt.Chart(df).transform_fold(
    ['Total', 'Missing'],
    as_=['variable', 'value']
).mark_bar().encode(
    alt.X('Year:O', title=''),
    alt.Y('value:Q', title='', stack=None),
    alt.Color('variable:N', 
              scale=alt.Scale(domain=['Total', 'Missing'],
                              range=['#ddccbb', 'red'])),
    order=alt.Order('variable:N', sort='descending')
).properties(
    background='white'
)