px.density_contour() ignores color labels for negative histfunc results

Consider the following example:

import plotly.express as px
import pandas as pd
import numpy as np
import plotly


np.random.seed(0)

n_samples = 1000
df = pd.DataFrame({
    'x': np.random.randint(low=1, high=5, size=n_samples),
    'y': np.random.randint(low=1, high=4, size=n_samples),
    'z': np.random.normal(loc=-40, scale=20, size=n_samples),
})

df['z'] = df['z'] * df['x']

# uncomment to make some values above 0
# df['z'] = df['z'] + 80

# uncomment to make all values above 0
# df['z'] = df['z'].abs()

fig = px.density_contour(df, x='x', y='y', z='z', histfunc='avg', height=800, width=1200, title=f'Plotly version: {plotly.__version__}')
fig.update_traces(contours_coloring="fill", contours_showlabels=True)
fig.show()

It calculates average of mostly negative values, which results in the following plot:

Color bar has correct color range, but all data points use same color - yellow, which is not particularly useful. Similar situation happens when only part of data is negative (uncomment first commented line in the example):

Left (positive) section looks fine, while right (negative) is colored with only one color.

There are workarounds, that involve changing of input data, but this makes plot reading much harder.

Comece executando o exemplo fornecido de px.density_contour() com resultados negativos e mistos de histfunc; em seguida, inspecione o ponto de entrada density_contour e como a coloração dos contornos lida com esses valores. Está concluído quando os resultados negativos e mistos usam todo o intervalo de cores em vez de colapsarem para uma única cor, enquanto o caso positivo existente continua correto.