Dataframe columns of lists of strings render the strings as ASCII numbers instead of strings

[shrianshChari]

When I attempt to render a dataframe that contains a column that is a list of strings (in this case my dataset is called spl and the column is titled team), the strings will be displayed correctly in Streamlit (run locally on my machine) but not when it gets translated using stlite (run using stlite sharing):

spl['team'].iloc[0:5]

Output in Streamlit:

Output in stlite:

When I take the first row of the output from stlite and convert each number into its corresponding ASCII character value,

>>> s = '91,34,83,110,111,114,108,97,120,34,44,34,71,111,108,101,109,34,44,34,71,101,110,103,97,114,34,44,34,90,97,112,100,111,115,34,44,34,70,111,114,114,101,116,114,101,115,115,34,44,34,83,116,97,114,109,105,101,34,93'
>>> s = s.split(',')
>>> c = list(map(lambda x: chr(int(x)), s))
>>> ''.join(c)
'["Snorlax","Golem","Gengar","Zapdos","Forretress","Starmie"]'

It seems that it is able to recognize that spl['team'] is a column that contains a list of strings, as when I run:

spl['team'].iloc[0]

I get the same output for both Streamlit and stlite:

[whitphx]

Thank you for reporting this!

[whitphx]

df.to_parquet() here is done without any error. https://github.com/whitphx/streamlit/blob/stlite-1.35.0/lib/streamlit/type_util.py#L1131

Maybe the problem is from fastparquet and/or parquet-wasm? -> Looks like fastparquet sets the column metadata "pandas_type": "mixed" in this case where pyarrow does "pandas_type": "list[unicode]". The code is here? https://github.com/dask/fastparquet/blob/1891a4a55fbe2ac23b29c064258c9c2eba480d28/fastparquet/util.py#L407-L408

---

Sample code:

import streamlit as st
import pandas as pd

df = pd.DataFrame({
    "names": [["foo", "bar"], ["baz", "quz"]]
})

st.dataframe(df)