Await Is Not a Context Switch: Understanding Python's Coroutines vs. Tasks
2 hours ago
- #Python
- #Concurrency
- #Async Programming
- Python's async model differs from JavaScript, C#, and Java, where awaiting a coroutine doesn't yield to the event loop by default.
- Concurrency in Python is introduced by tasks, not coroutines, which is a key distinction often misunderstood by developers.
- Awaiting a coroutine in Python is synchronous unless the awaited object isn't ready, unlike in other languages where every await is a potential context switch.
- Tasks in Python are the unit of concurrency, and only awaiting a task yields control back to the event loop.
- The misconception that all async boundaries are suspension points leads to unnecessary locking and complexity in Python code.
- Python's async design stems from generators, offering fine-grained control over concurrency and explicit task creation.
- Practical guidelines include auditing task creation points, scanning critical sections for suspension points, and teaching the coroutine vs. task distinction.
- Understanding where code can actually interleave is more effective for ensuring correctness than applying blanket locking rules in async Python code.