File_uploader() - ValueError: I/O operation on closed file.

Summary

I upgraded Streamlit and file_uploader() haven’t the expected behavior. I’m trying to load a file, then use a dropdown menu to select the right column to analyse. I adapted the proposed solution here , but I still have “ValueError: I/O operation on closed file.” when I select a value in the dropdown menu.

Steps to reproduce

What are the steps we should take to reproduce the bug:

1. Select a file to upload
2. Select a value in the dropdown menu

Screencast

Expected behavior:

Explain what you expect to happen when you go through the steps above, assuming there were no bugs.

Is this a regression?

That is, did this use to work the way you expected in the past? yes

Debug info

streamlit==0.70.0 streamlit-nightly==0.70.1.dev20201101

Additional information

Here is my code :

def upload_file():
    file = st.sidebar.file_uploader("Choisir un fichier", type=["csv"])
    if file is None:
        return None,None,None,None,None,None
    file.seek(0)
    df = pd.read_csv(file, sep=';',index_col=None, encoding = "utf-8")
    menu_value = st.sidebar.selectbox('Quelle colonne correspond au texte à analyser ?',df.columns)
    language,remove_hashtags,remove_usernames,ncomponents=sidebar()
    return(df,menu_value,language,remove_hashtags,remove_usernames,ncomponents)

def sidebar():
    language = st.sidebar.selectbox('Quelle est la langue du corpus ?',['french','english'])
    remove_hashtags=st.sidebar.checkbox('Supprimer les hashtags',value=False)
    remove_usernames=st.sidebar.checkbox('Supprimer les mentions',value=False)
    ncomponents=st.sidebar.slider("Combien de topics à détecter", min_value=0, max_value=100, value=20, step=1, format=None, key=None)
    return language,remove_hashtags,remove_usernames,ncomponents
    
def main():
    st.set_page_config(
        page_title="Topic Modeling",
        page_icon="🧊",
        layout="wide",
        initial_sidebar_state="expanded",
    )
    st.set_option('deprecation.showfileUploaderEncoding', False)

    st.sidebar.title('Paramètres')
    st.title('Topic Modelling')
    projet=st.sidebar.text_input("Donner un nom de projet", value='monProjet', max_chars=None, key=None, type='default')
    
    
    df,menu_value,language,remove_hashtags,remove_usernames,ncomponents=upload_file()
   
    if st.sidebar.button('Valider'):
        analyse(projet,language,remove_hashtags,remove_usernames,ncomponents,df_input,menu_value)
        list_results(path="./"+projet+"/")

[karriebear]

Thanks for sending this along. The fix was recently merged and will be available in version 0.73 that will be coming out this week. If you would like to get the latest, you can use our streamlit-nightly package that is updated with the latest changes every day.

[mmcguffi]

I think this is the best thread to leave this – I am encountering this error when using the file_uploader in combination with st.radio(), and uploaded_file.seek(0) does not solve the problem. Ill admit, I honestly do not understand what seek(0) is fundamentally doing, so maybe I am missing something obvious. Here is a minimal reproducible example:

import streamlit as st
import io

option = st.radio(
    'Choose method of submitting sequence:',
    ["Upload a file (.txt)", "Enter a sequence"])

txt = None

if option == "Upload a file (.txt)":

    uploaded_file = st.file_uploader("Choose a file:", type=['txt'])

    if uploaded_file is not None:
        uploaded_file.seek(0) #does not do anything
        text_io = io.TextIOWrapper(uploaded_file,encoding='UTF-8')
        txt=list(text_io.readlines())

elif option == "Enter a sequence":
    txt = st.text_area('Input sequence here:')

if txt:
    st.write(txt)

streamlit 0.71.0 and 0.72.0

When users upload a file, switch to manually entering text, and then try to upload again, the ValueError: I/O operation on closed file is thrown. This is a regression, as I did not encounter this error with a much older version of streamlit

any help would be much appreciated!