How to create a consistent plotly color map with dynamic input?

Question

I have some dataframes I want to plot with a variable number of Fruits. When plotting, each dataframe will always have the same Fruit types, even though some may have a value of 0 in their respective row.

df1
Day  Value Fruit
1    5     Banana
1    3     Pear
2    4     Banana
2    2     Pear
3    3     Banana
3    3     Pear

df2  # note all pears 0'd
Day  Value Fruit
1    5     Banana
1    0     Pear
2    4     Banana
2    0     Pear
3    3     Banana
3    0     Pear

# AKA
df1.Fruit.unique() == df2.Fruit.unique()

Plotting code:

import dash_bootstrap_components as dbc
from plotly import express as px

bar_charts = []
for df in dfs:
  # to get a stacked bar chart
  fruit_data = df.groupby(["Day", "Fruit"]).Value.sum().to_frame().reset_index()
  bar_charts.append(
    dcc.Graph(
      figure=px.bar(
        fruit_data,
        x="Day",
        y="Value",
        color="Fruit",
      ),
    )
  )
return dbc.Col(bar_charts)

Plotly seems to be assigning differnt colors to the Fruit between dataframes based on the value of the Fruit. I want the Fruits to always be the same color across graphs regardless of value (e.g. Banana always yellow, Pear always green). HOWEVER, I don't know what Fruits are available until runtime, so I can't just hard code a color map ahead of time.

How do I tell plotly to always color the Fruit the same?

Im sure this is some silly little option im missing.

Answer 1

To make sure that the same colors are assigned to the same Fruit, you can make a color map using fruit_data['Fruit'].unique() and any list of colors like ['yellow', 'green'] or a longer list px.colors.qualitative.Alphabet for color_discrete_map in px.bar like this:

colordict = {f:px.colors.qualitative.Alphabet[i] for i, f in enumerate(fruit_data.Fruit.unique())}
colordict = {f:['yellow', 'green'][i] for i, f in enumerate(fruit_data.Fruit.unique())}

Plot:

And if you're working with a large number of variables and worry about running out of colors, just incorporate the approach described in Plotly: How to increase the number of colors to assure unique colors for all lines like this:

n_colors = len(fruit_data.Fruit.unique())
colorscale = colors = px.colors.sample_colorscale("viridis", [n/(n_colors -1) for n in range(n_colors)])
colordict = {f:colorscale[i] for i, f in enumerate(fruit_data.Fruit.unique())}

Complete code:

import io
import plotly.express as px
import pandas as pd

# data
fruit_data = pd.read_csv(io.StringIO("""Day  Value Fruit
                                1    5     Banana
                                1    3     Pear
                                2    4     Banana
                                2    2     Pear
                                3    3     Banana
                                3    3     Pear"""),sep= '\\s+')

# color assignment:
# colordict = {f:px.colors.qualitative.Alphabet[i] for i, f in enumerate(fruit_data.Fruit.unique())}
colordict = {f:['yellow', 'green'][i] for i, f in enumerate(fruit_data.Fruit.unique())}

# color assignment for a large number of variables:
n_colors = len(fruit_data.Fruit.unique())
colorscale = colors = px.colors.sample_colorscale("viridis", [n/(n_colors -1) for n in range(n_colors)])
colordict = {f:colorscale[i] for i, f in enumerate(fruit_data.Fruit.unique())}

# plotly figure
fig = px.bar(
        fruit_data,
        x="Day",
        y="Value",
        color = 'Fruit',
        color_discrete_map = colordict
      )
fig.show()