[Python] Streamlit renders plotly barplot incorrectly when running from global install

I have a barplot with stacked bars that I render in a streamlit app.

To add some context, I want to be able to pip install the wheel of my project on a different PC and then run it from Windows cmd and without creating a virtual environment every time.

For better understanding, I created a minimally reproducible example.

import streamlit as st
import plotly.express as px
import pandas as pd

# this is needed to write_html correctly
import plotly.io as pio
pio.templates.default = "plotly"

df = pd.DataFrame({
'Fruit': ['Apples', 'Apples', 'Oranges', 'Oranges'],
'Amount': [10, 15, 20, 25],
'City': ['SF', 'NY', 'SF', 'NY']
})

fig = px.bar(df, x='Fruit', y='Amount', color='City')
fig.update_layout(barmode='stack')

print("SAVE IMAGE")
fig.write_html("test.html")

st.plotly_chart(fig)


The saved html looks something like this:



That's also what it looks like when I use the streamlit installed in a virtual environment:

.\.venv\Scripts\streamlit.exe run .\streamlit_test.py

However, when I run it from the globally installed streamlit, the bars are not stacked.

C:\Users\my_username\AppData\Local\Programs\Python\Python312\Scripts\streamlit.exe run .\streamlit_test.py

The plot is not showing as stacked barplot, but the bars are next to each other.



I verified that the streamlit and python versions are the same for both cases (streamlit version 1.45.1 and python 3.12.

Is this a bug in streamlit or am I missing something?

Continue reading...