Describe your changes

This PR attempts to fix the behavior of st.spinner so that it does not cause stale elements. Instead of hiding the spinner in the frontend via dg.empty(), this PR uses dg.container(). An empty container gets ignored on the frontend; in comparison, an dg.empty() placeholder is treated as an empty element in the front end, resulting in unwanted stale elements in some situations. The stale issue is caused by a scenario where in the first run, we are using a spinner, but in the second run, we are just showing the data, for example:

import streamlit as st
import pandas as pd
import time

if "df" not in st.session_state:
    with st.spinner("Load data"):
        time.sleep(5)
        df = pd.DataFrame({"A": [1, 2, 3]})
        st.session_state.df = df

st.dataframe(st.session_state.df)

# Do something else that takes some time:
time.sleep(5)

The first run uses two elements, while the second one uses only one. This will cause a duplicated greyed-out version of the dataframe until the script has fully finished.

Alternatively, we could introduce some kind of dg.clear() or dg.empty(clear=True) command that removes the current element from the frontend.

Testing Plan

Updated unit tests related to st.spinner. I also had to deactivate the spinners in some caching tests since the testing utility does not work yet with st.container as mentioned here:

Contribution License Agreement

By submitting this pull request you agree that all contributions to this project are made under the Apache 2.0 license.