bug: text_input with caching

Summary

Using a cached value as text_input value, text doesn’t get updated unless focus changes. I use caching a lot to support more complicated scenarios.

Steps to reproduce

2- write something hit enter 3- write something else (without changing focus) hit enter

Code snippet:

import streamlit as st
from streamlit.report_thread import get_report_ctx
    
def query_cache(q=None):
    ctx = get_report_ctx()
    session_id = ctx.session_id
    session = st.server.server.Server.get_current()._get_session_info(session_id).session
    if not hasattr(session, "_query_state"):
        setattr(session, "_query_state", "default")
    if q is not None:
        session._query_state = q
    return session._query_state

# if st.button('reset'):
#     value = ""
#     query_cache("")
value = query_cache()    
my_input=st.text_input("Input", value)
c=query_cache(my_input)
st.write(c)

Expected behavior:

should work the same as text_input without caching.

Actual behavior: 1- type ‘poo’ enter 2- type ‘poop’ enter (doesn’t update result, loses focus) 3- after focus change write ‘poop’ again. It works

Is this a regression?

no

Debug info

• Streamlit version: 0.80.0
• Python version: 3.6.9
• OS version: ubuntu 18.04
• Browser version: chrome & firefox latest

Additional information

I’m caching the text_input values because I want to have a button clear them (commented out in code example) Solutions that don’t use caching and also allow me to reset with a button, while also being not buggy with text_input are also welcome.

[btjones-me]

Hi folks, even using the workaround with session state I find the same issue

        state = st.session_state
        if "_pm" not in st.session_state:
            state._pm = 1.0

        proximity_modifier = st.number_input('proximity_modifier', step=0.5, key='_pm')
        st.write(proximity_modifier)
        st.write(state._pm)

still results in the previous issue. Am I doing something wrong?

[kmcgrady]

Thanks for the report @cceyda ! I believe our solution with our implementation of Session State will help on this. I tagged this issue to the project so that when it’s released, we can follow up with a solution!