如何解决Plotly旭日图内层颜色与数值总和不匹配问题？

问题描述 #

我制作了一个Plotly旭日图，外层颜色显示正确（对应数值为32k）。鼠标悬停时，内层显示所有外层数值的总和（约64k），但颜色却对应16k的数值。修改示例后，最内层“a”的数值总和为10，应显示橙色，但实际颜色不符。

可复现代码 #

from pandas import DataFrame
import plotly.express as px

A = ["a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c"]
B = ["b11", "b12", "b21", "b22", "b31", "b32", "b11", "b12", "b21", "b22", "b31", "b32"]
C = ["c111", "c112", "c121", "c122", "c211", "c212", "c221", "c222", "c311", "c312", "c321", "c322"]
C_values = [1, 2, 3, 4, 2, 3, 1, 2, 3, 4, 2, 3]
df = DataFrame(dict(A=A, B=B, C=C, C_values=C_values))

fig = px.sunburst(df, path=["A", "B", "C"], values="C_values", color="C_values", color_continuous_scale=[(0, "green"), (0.5, "yellow"), (1, "red")], range_color=(0, 15))
fig.show()

问题原因 #

• 默认情况下，px.sunburst的color参数直接调用原始数据列的数值，不会自动计算父节点的数值总和
• 内层节点的颜色实际是基于子节点C_values的平均值（默认聚合方式），而非总和，导致颜色与悬停显示的总和数值不匹配

修复方案 #

需要手动计算各层级节点的数值总和，将总和作为颜色映射的依据：

修改后的代码 #

from pandas import DataFrame
import plotly.express as px

A = ["a", "b", "c", "a", "b", "c", "a", "b", "c", "a", "b", "c"]
B = ["b11", "b12", "b21", "b22", "b31", "b32", "b11", "b12", "b21", "b22", "b31", "b32"]
C = ["c111", "c112", "c121", "c122", "c211", "c212", "c221", "c222", "c311", "c312", "c321", "c322"]
C_values = [1, 2, 3, 4, 2, 3, 1, 2, 3, 4, 2, 3]
df = DataFrame(dict(A=A, B=B, C=C, C_values=C_values))

# 计算各层级的总和
# 计算B层级的总和
b_totals = df.groupby(['A', 'B'])['C_values'].sum().reset_index(name='b_total')
df = df.merge(b_totals, on=['A', 'B'])
# 计算A层级的总和
a_totals = df.groupby(['A'])['C_values'].sum().reset_index(name='a_total')
df = df.merge(a_totals, on=['A'])

# 为每个层级节点分配对应总和值
df['color_value'] = df['C_values']
# 给B层级节点分配B组总和
df.loc[df.duplicated(subset=['A', 'B'], keep='first'), 'color_value'] = df['b_total']
# 给A层级节点分配A组总和
df.loc[df.duplicated(subset=['A'], keep='first'), 'color_value'] = df['a_total']

fig = px.sunburst(df, path=["A", "B", "C"], values="C_values", color="color_value", 
                  color_continuous_scale=[(0, "green"), (0.5, "yellow"), (1, "red")], 
                  range_color=(0, 15))
fig.show()

效果说明 #

修改后：

• 最外层C节点的颜色基于自身C_values值，保持正确
• 中间层B节点的颜色基于对应B组的C_values总和
• 最内层A节点的颜色基于对应A组的C_values总和
• 示例中“a”节点的总和为10，会匹配range_color=(0,15)的中间色调，符合预期

内容的提问来源于stack exchange，提问作者Sebastian Tramposch