sync_to_async.md

Run synchronous functions asynchronously

Running your bound, synchronous functions asynchronously can be an easy way to make your application responsive and scale to more users.

Asyncify

This example will show how to make your app responsive by running a sync, cpu bound function asynchronously. We will be using asyncer by Tiangolo. You can install the package via pip install asyncer.

import numpy as np
import pandas as pd

from asyncer import asyncify
import panel as pn

widget = pn.widgets.IntSlider(value=5, start=0, end=10)

def do_sync_work(it, n):
    return sum(pd.DataFrame(np.random.rand(n,n)).sum().sum() for _ in range(it))

async def create_result():
    yield pn.indicators.LoadingSpinner(value=True, width=25, height=25)
    result = await asyncify(do_sync_work)(it=5, n=10000)
    yield f"Wow. That was slow.\n\nThe sum is **{result:.2f}**"

pn.Column(widget.rx() + 1, create_result).servable()

Your browser does not support the video tag.

Without asyncify the app would have been unresponsive for 5-10 seconds while loading.

asyncify works well for IO bound functions as well as for CPU bound functions that releases the GIL.

Dask

If you run many cpu bound functions you may consider offloading your functions asynchronously to an external compute engine like Dask. See our Dask how-to Guide.