AsyncIO part 2: Threads, green threads, coroutines and The Event Loop.
Taras Voinarovskyi
Taras Voinarovskyi
Threads
- Minimum 1 MB per Thread (8MB default)
- Context switching is expensive
- Processing bound to 1 CPU
- Other GIL related issues
- etc.
Why can be bad in some cases:
Green threads
PyPy's
greenlet
Stackless tasklet
coroutines
tasks
Green threads
Coroutines
PEP 492 -- Coroutines with async and await syntax
async def my_coroutine():
res = await io_operation()
async with my_async_context():
print("In context")
async for x in my_asyn_iter():
print("Yielded", x)
yield "We can even do async generators" Coroutines
import time
class Future:
def __await__(self):
sleep_time = 0
while True:
sleep_time = yield f"Sleeping for {sleep_time}s"
time.sleep(sleep_time)
async def coro():
await Future()
c = coro()Coroutines
Questions?
AsyncIO part 2: Threads, green threads, couroutins and The Event Loop.
By Taras Voinarovskyi
AsyncIO part 2: Threads, green threads, couroutins and The Event Loop.
Threads are bad, we can do it in a single thread and still get the same result. Well, yes, but why? What makes a thread fail in comparison to a coroutine and when? Socket handling in a loop is good for a simple echo server, but will not help us do a normal project. What features are needed for a normal socket server, that can also connect to a database, do complex processing, do background processing, read files and still retain 1 thread to do that all? Lets talk about delayed calls, transports, protocols and other parts that make up an event loop.
