Apprendre Running Blocking Code with run_in_executor()

Glissez pour afficher le menu

Not every library supports asyncio. requests, pandas, file system calls, and CPU-heavy functions are all blocking – calling them directly inside a coroutine freezes the entire event loop. run_in_executor() moves blocking code to a thread pool, freeing the event loop to keep running.

The Problem

Calling a blocking function inside an async context stalls all other coroutines until it returns:


              123456789101112131415161718192021
            
import asyncio
import time
import nest_asyncio

nest_asyncio.apply()

# Blocking call inside a coroutine – freezes the event loop
async def process_data(label):
    print(f"Starting {label}")
    time.sleep(1)  # Blocking – no other coroutine runs during this
    print(f"Done {label}")

async def main():
    start_time = time.time()
    await asyncio.gather(
        process_data("task_A"),
        process_data("task_B"),
    )
    print(f"Total time: {time.time() - start_time:.2f}s")  # ~2s – sequential

asyncio.run(main())

Using `run_in_executor()`

loop.run_in_executor(executor, func, *args) runs func in a thread pool and returns an awaitable. Pass None as the executor to use the default ThreadPoolExecutor:


              123456789101112131415161718192021222324
            
import asyncio
import time
import nest_asyncio

nest_asyncio.apply()

# Moving a blocking call to a thread pool
def blocking_process(label):
    print(f"Starting {label}")
    time.sleep(1)  # Runs in a thread – event loop stays free
    print(f"Done {label}")

async def main():
    loop = asyncio.get_event_loop()
    start_time = time.time()

    await asyncio.gather(
        loop.run_in_executor(None, blocking_process, "task_A"),
        loop.run_in_executor(None, blocking_process, "task_B"),
    )

    print(f"Total time: {time.time() - start_time:.2f}s")  # ~1s – concurrent

asyncio.run(main())

A Realistic Example: Mixing `requests` with asyncio

When you must use a synchronous HTTP library alongside async code:


              1234567891011121314151617181920212223242526272829
            
import asyncio
import requests
import nest_asyncio

nest_asyncio.apply()

# Wrapping a blocking requests call for use in asyncio
def fetch_post_sync(post_id):
    response = requests.get(
        f"https://jsonplaceholder.typicode.com/posts/{post_id}"
    )
    return response.json()["title"]

async def fetch_post_async(post_id):
    loop = asyncio.get_event_loop()
    title = await loop.run_in_executor(None, fetch_post_sync, post_id)
    return title

async def main():
    titles = await asyncio.gather(
        fetch_post_async(1),
        fetch_post_async(2),
        fetch_post_async(3),
    )

    for title in titles:
        print(title)

asyncio.run(main())

Using a Custom Executor

Pass a ThreadPoolExecutor to control the number of threads, or a ProcessPoolExecutor for CPU-bound work:


              1234567891011121314151617181920212223242526
            
import asyncio
import time
from concurrent.futures import ThreadPoolExecutor
import nest_asyncio

nest_asyncio.apply()

# Using a custom thread pool with a fixed size
def heavy_computation(value):
    time.sleep(0.5)  # Simulating CPU work
    return value ** 2

async def main():
    loop = asyncio.get_event_loop()

    with ThreadPoolExecutor(max_workers=4) as executor:
        results = await asyncio.gather(
            *[
                loop.run_in_executor(executor, heavy_computation, value)
                for value in range(1, 6)
            ]
        )

    print(results)

asyncio.run(main())

When to Use `run_in_executor()`

Use it when:

A third-party library does not support asyncio (e.g., requests, legacy database drivers);
File I/O uses standard open() instead of an async file library;
A CPU-bound function would block the loop for more than a few milliseconds.

Do not use it as a general workaround for writing async code – always prefer native async libraries when they exist.

Tout était clair ?

Merci pour vos commentaires !

Section 4. Chapitre 1

Demandez à l'IA

Posez n'importe quelle question ou essayez l'une des questions suggérées pour commencer notre discussion

Section 4. Chapitre 1

Running Blocking Code with run_in_executor()

The Problem

Using run_in_executor()

A Realistic Example: Mixing requests with asyncio

Using a Custom Executor

When to Use run_in_executor()

Using `run_in_executor()`

A Realistic Example: Mixing `requests` with asyncio

When to Use `run_in_executor()`