Suppress dataframe inference warnings

[anthoant]

Summary

In my app I create a pandas pivot-table in a human-readable grid. Below the columns headers I have included a row of descriptive text while the remaining row data are all numeric floats. After converting from streamlit 1.17.0 to 1.20.0, I am getting small yellow triangle exclamation points and a warning saying: “The value cannot be interpreted as a number.” How can I suppress or disable these warnings? The pivot table will always have non-numeric text.

Steps to reproduce

Code snippet:

import streamlit as st
import pandas as pd

d = pd.DataFrame({
    'Mix number':[1,2,3],
    'Mix Name':['Control','Test1','Test2'],
    'Ingredient number':[4,5,6],
    'Ingredient name':['A','B','C'],
    'Level':[33,33,34]
    })

pivot_table = pd.pivot_table(
    data=d,
    values='Level',
    index=['Ingredient number','Ingredient name'],
    columns=['Mix number','Mix Name']
)

st.dataframe(pivot_table)

If applicable, please provide the steps we should take to reproduce the error or specified behavior.

Expected behavior:

Display data without data warnings.

Actual behavior:

Shows little yellow triangles with data warning.

Debug info

• Streamlit version: 1.20.0
• Python version: 3.10
• Using Conda? PipEnv? PyEnv? Pex?
• OS version: Windows 10
• Browser version: Chrome

[lukealexdavis]

I’m having a similar issue so would love to hear more.

[anthoant]

I’ve tried converting the numeric column header “Mix number” into string but still same issue.

[anthoant]

Bump…still having this issue. Any ideas? I haven’t been able to find any documentation of this or similar issues anywhere.

[mathcatsand]

Perhaps not the most elegant, but here are a couple options:

import streamlit as st
import pandas as pd

d = pd.DataFrame({
    'Mix number':[1,2,3],
    'Mix Name':['Control','Test1','Test2'],
    'Ingredient number':[4,5,6],
    'Ingredient name':['A','B','C'],
    'Level':[33,33,34]
    })

pivot_table = pd.pivot_table(
    data=d,
    values='Level',
    index=['Ingredient number','Ingredient name'],
    columns=['Mix number','Mix Name']
)

pivot1 = pivot_table.fillna('')

pivot2 = pivot_table.copy()
for col in pivot_table.columns:
    pivot2[col] = pivot2[col].astype(str)

st.dataframe(pivot_table)
st.dataframe(pivot1)
st.dataframe(pivot2)

I suspect this is related to some underlying work on the data editor. Ping @lukasmasuch

[lukasmasuch]

Unfortunately, we do not officially support dataframes with multiple header rows at the moment There are some workarounds, as shown by @mathcatsand by manually converting to string. We are tracking this issue here on Github: Enable `st.dataframe` to handle dataframes with hierarchical headers · Issue #6319 · streamlit/streamlit · GitHub You can help to prioritize this issue by upvoting it on Github.