Python asyncio gather: A Practical Guide
Python asyncio gather runs coroutines concurrently instead of one-by-one. Practical patterns: fan-out, error handling, create_task, TaskGroup, and not blocking the loop.
If you come from Dart, you already know this story: await in a loop runs things one at a time, and Future.wait runs them together. Python's asyncio has the same shape — and asyncio.gather is its Future.wait . The python asyncio gather pattern is how you fetch ten URLs, query five services, or fan out any I/O-bound work without threads and without waiting for each task to finish before starting the next. This guide is relentlessly practical: runnable Python 3.12+ code for every pattern, the exact errors beginners hit, and a Dart-to-Python translation table so the concepts click instantly. What you'll learn Coroutines and the event loop in 60 seconds (the Dart-bridged version) Why sequential await is slow, measured with a timer asyncio.gather : the fan-out workhorse, with a real example Surviving failures: return_exceptions=True asyncio.create_task : fire-and-forget done right gather vs create_task vs TaskGroup : which to use when Never block the loop: asyncio.to_thread and asyncio.sleep vs time.sleep Bounded concurrency with Semaphore (the production pattern) Step 1 — The problem: sequential awaits are slow Three fake network calls, one second each. Awaited in a loop, they take three seconds — each waits for the previous to finish, even though they're independent: import asyncio import time async def fetch(url: str) -> str: await asyncio.sleep(1) # stands in for a real network request return f"data from {url}" async def sequential() -> None: start = time.perf_counter() results = [] for u in ["a", "b", "c"]: results.append(await fetch(u)) # waits a full second each time print(f"sequential: {time.perf_counter() - start:.1f}s -> {results}") async def parallel() -> None: start = time.perf_counter() results = await asyncio.gather(*(fetch(u) for u in ["a", "b",…