How to replace values on a dataframe using pandas and streamlit?

Using Streamlit
1

I have a python that allow user to select a column from a given dataframe and change a value cell were he replace the old value with a new value using streamlit.

The problem is that when the user replace the old value with new one the original dataframe is not changing like so :

enter image description here

When I refresh the page or reselect the same column the old value still exist and not the new value.

Where is the error in my code??

code:

import pandas as pd
from pandas.api.types import is_numeric_dtype
import streamlit as st


df = pd.DataFrame({
            "source_number":
             [11199,11328,11287,32345,12342,1232,13456,123244,13456],
             "location":
             ["loc2","loc1","loc3","loc1","loc2","loc2","loc3","loc2","loc1"],
              "category":
             ["cat1","cat2","cat1","cat3","cat3","cat3","cat2","cat3","cat2"],
             })
     
if st.checkbox("replace"):
columns = st.selectbox("Select  column", df.columns)
old_values = st.multiselect("Current Values",list(df[columns].unique()),list(df[columns].unique()))
with st.form(key='my_form'):
     col1,col2 = st.beta_columns(2)
     st_input = st.number_input if is_numeric_dtype(df[columns]) else st.text_input
     with col1:
          old_val = st_input("old value")
     with col2:
          new_val = st_input("new value")
     if st.form_submit_button("Replace"):
           df[columns]=df[columns].replace(old_val,new_val)
           st.dataframe(df)
2

From what I know about streamlit, everytime the page refreshes it executes all code. So it is probably re-creating the df from scratch, and in the process losing your changes.

Try one of 2 things.
Create the data in a function and preface the function with @st.cache
Maybe that will help.
Save the df to a file and then reload that file in instead of creating the df.