Fix stacked barplot missing groups issue (#5428)

[AJSTYLE-lab]

Pull Request Description for #5428

Title:
Fix stacked barplot missing groups issue (#5428)

Description:
This PR addresses the bug where certain groups (commanders) with non-missing values are not displayed in stacked bar plots when using plotly.express.bar.

Problem:

When plotting a stacked bar chart of commander_perc_of_card vs. savedate_week, some commanders with values for certain weeks were missing in the overall plot.

This happens because plotly.express.bar only plots groups that exist for each specific x value (savedate_week). Missing combinations are ignored, even if other weeks have values.

Solution:

Generated all possible combinations of savedate_week and commanders using itertools.product.

Merged these combinations with the original dataset.

Filled missing commander_perc_of_card values with 0 to ensure that every commander is displayed for every week.

This guarantees that stacked barplots now show all groups, maintaining accurate visual representation.

Files Modified:

reproduce_bug.py — added a reproducible example to demonstrate the fix.

Before

After

Code:

import pandas as pd
import plotly.express as px
from itertools import product

# Load the data
plot_data = pd.read_csv("example_plot_data.csv")
plot_data['savedate_week'] = pd.to_datetime(plot_data['savedate_week'])

# Get all unique weeks and commanders
all_weeks = plot_data['savedate_week'].unique()
all_commanders = plot_data['commanders'].unique()

# Generate full combinations of weeks and commanders
full_index = pd.DataFrame(list(product(all_weeks, all_commanders)), columns=['savedate_week','commanders'])

# Merge with original data and fill missing values with 0
plot_data_full = full_index.merge(plot_data, on=['savedate_week','commanders'], how='left')
plot_data_full['commander_perc_of_card'] = plot_data_full['commander_perc_of_card'].fillna(0)

# Create the stacked barplot
fig = px.bar(
    plot_data_full,
    x="savedate_week",
    y="commander_perc_of_card",
    color="commanders"
)

fig.show()

[camdecoster]

Thanks for the PR! One of our team members will review and provide feedback.

[camdecoster]

It looks like your PR isn't updating the library, just providing a workaround for the original issue. Did you intend to update the library with a fix?