Skip to content

pitfalls.md

Latest commit

 

History

History
54 lines (37 loc) · 2.08 KB

pitfalls.md

File metadata and controls

54 lines (37 loc) · 2.08 KB
Raw

Chapter 21 — Pitfalls: Asynchronous Programming

Pitfall 1: Forgetting to await

Because async def functions return Coroutine objects rather than executing immediately, forgetting the await keyword results in completely silent failures.

async def fetch_database():
    print("Fetching data...")
    return {"status": "ok"}

async def main():
    # BUG: We forgot 'await'!
    # This evaluates to a Coroutine object, not the dictionary.
    data = fetch_database()
    
    # BUG: This will crash because 'data' is a coroutine object, not a dict.
    if data["status"] == "ok":
        print("Success")

Fix: You must always await fetch_database(). If you don't, Python will eventually print a warning: RuntimeWarning: coroutine 'fetch_database' was never awaited, but your code will still have failed.

Pitfall 2: The Unbounded Concurrency Trap (DDoS'ing yourself)

If you use concurrent.futures.ThreadPoolExecutor(max_workers=10), your concurrency is strictly capped at 10. asyncio.gather has no such cap.

async def scrape(url):
    # Simulated network request
    pass

async def massive_scrape(urls):
    # BUG: If `urls` has 50,000 items, asyncio will instantly attempt
    # to open 50,000 TCP sockets. Your OS will crash with "Too Many Open Files"
    # or the target server will ban your IP address for a DDoS attack.
    tasks = [scrape(url) for url in urls]
    await asyncio.gather(*tasks)

Fix: You must introduce an asyncio.Semaphore to throttle your concurrency. See mini_project.py for the correct implementation.

Pitfall 3: Using requests instead of httpx

Many developers start writing async APIs (like FastAPI) and continue using the legendary requests library.

Why it fails: requests.get() is a synchronous, blocking function. Because asyncio operates on a single thread, calling requests.get() will literally freeze every single other concurrent user on your server until that HTTP request finishes. Fix: You must use an asynchronous HTTP client like httpx.AsyncClient() or aiohttp.ClientSession().