Pretty way to change a function’s parameters via st.dropdown if these parameters are not in app.py?

🎈 Using Streamlit
1

Hi guys,

Not sure if there’d be a pretty way to change a function’s parameters via st.dropdown if these parameters are actually located in a secondary Python file? Please see screenshot below for context:

image
image1668×621 99.2 KB

At the moment, I’ve having to paste all these secondary files in app.py… and that monlithic bloc is neither handy nor pretty! :grinning_face_with_smiling_eyes:

Thanks,
Charly

2

How about using session state to hold kwargs selected in the dropdown? Can’t you pass them directly as parameters?

Alternatively, put the components that are serving the app behind a FastAPI microservice running on a separate thread.

HOST = 'localhost'
PORT = 8000

def thread_runner(app, host, port):
    uvicorn.run(app, host=host, port=port)

app = FastAPI()
@app.get('/prepare_query')
def prepare_query(**kwargs):
    for name, val in kwargs.items():
        # build your query string here, do the work, and return response
        pass

thread = threading.Thread(name='QueryServer', target=thread_runner, args=(app, HOST, PORT))
thread.start()
2 Likes
3

Hey @Charly_Wargnier,

So what you want to do is change the behavior of that suggests library, and more specifically add a custom parameter to change the language part of the following hardcoded string:

The cleanest solution would be to fork and patch the library to add a new parameter, create your PyPI and build your streamlit app on top of it.

Another solution would be to patch the library’s function directly from your app.

import streamlit as st
from suggests import suggests

def custom_get_google_url():
    # Retrieve language from session state, or set lang to 'en' by default
    lang = st.session_state.get("google_url_language", "en")

    return f"https://www.google.com/complete/search?sclient=psy-ab&hl={lang}&q="

# Select your language, and store it in session state as 'google_url_language'
st.selectbox("Language", ["en", "fr", "es"], key="google_url_language")

# Here we replace suggests's function by our own
suggests.get_google_url = custom_get_google_url

Two things to note here:

  • You must do an import suggests instead of from suggests import get_google_url to patch the function, like done above. suggests.get_google_url = ... redefines the function used by the library, whereas get_google_url = ... is just a regular assignment.

  • Patching a library’s function is done globally, so to make sure each user session uses its own language parameter, I retrieve the latter through a very basic session state implementation which will keep track of the language selected by each user. Without session state, you might run into race condition issues where one user might use the language parameter of another user connected to your app.

2 Likes