On widget variable change, a previous function not using the variable is re-run and I receive Warning

Using Streamlit
1

I have 5 functions setup, and the first 3 establish the connection to databases, and collect data, and do some manipulations. The next two manipulate the dataset based on variables passed into them. Whenever any variable changes via a widget, I’m returned WARNING: Cached Object Mutated, and the 3rd function (that should create a static dataset) gets re-run. Here’s some dummy code.

@st.cache(allow_output_mutation=True)
def connect_engine(url, conn_args=None):       
    if conn_args != None:
        engine = sqlalchemy.create_engine(url, connect_args=conn_args)
    else:
        engine = sqlalchemy.create_engine(url)
    return engine

@st.cache(hash_funcs={sqlalchemy.engine.base.Engine: id}, suppress_st_warning=True)
def get_data_1(engine):
    sql = "SELECT * FROM table"
    df = pd.read_sql(sql, engine)
    return df

@st.cache(hash_funcs={sqlalchemy.engine.base.Engine: id}, suppress_st_warning=True)
def get_data_2(dataframe, engine):
    sql = "SELECT * FROM table"
    df = pd.read_sql(sql, engine)
    dataframe['col_2'] = df['col_1'].add(3)
    return df

@st.cache(allow_output_mutation=True)
def add_more(dataframe, var=True):
    if var:
        dataframe['col_2'] = dataframe['col_2'].add(3)
    else:
        pass
    return df

db1 = connect_engine(connection_url1)
db2 = connect_engine(connection_url2)
df = get_data_1(db1)
df = get_data_2(df, db2)

if st.checkbox('Use variable'):
    variable=True
else:
    variable=False
df = add_more(df, variable)
st.write(df.head()
2

Hi @ccoggins -

It’s hard to say without me replicating this locally, but in each case, you are changing the binding of df, so it seems that you are mutating the object. As a debugging method, have you tried using unique names for each instance of df?

Best,
Randy