Python-扭曲的主循环内的while循环？

Question

I need to know if its possible to run a while loop inside a twisted websocket main loop

the while loop im referring to is the lib you see in this question: shout-python segmentation fault how can I fix this?

all I need it to do is send the new title once it updates, though I can handle that part. Its that while self.running: in the play() function. If you can help I'll surely appreciate it.

Answer 1

For Twisted's single-threaded, cooperative multitasking system to operate at its best, it's important that any particular piece of code running in the reactor thread not run for too long without giving control back to the reactor. As long as any one piece of code is running in that thread, no other code is running in that thread. In a single-threaded, cooperative multitasking system that means other events aren't being serviced.

Depending on your application, it may be fine for a single piece of code to run without giving up control for many milliseconds, many seconds, perhaps even minutes. It's entirely dependent on what events your application is responsible for handling and what level of responsiveness you want to get from it. When writing general purpose library code for a system like this, most people assume that it's only okay to run code for a single task for a handful of milliseconds or so before giving up control - to err on the side of being suitable for use in more applications rather than fewer (although people rarely consider the exact time limit, mostly operations are separated into "pretty quick" and everything else).

What's almost always unacceptable is to run a single piece of code indefinitely without giving control back to the reactor. The loop in the answer you linked to is effectively infinite and so it will hold control for an arbitrarily long period of time (perhaps for most of the runtime of the program). There are few if any applications that can tolerate this since the result is that other events will never be handled. If it's tolerable for your application to be unable to respond to any events while it spends its entire run time working on a single task then you may not need a multitasking system at all (ie, you may not need Twisted, you may be able to just use a while loop).

The loop in that answer is basically a "process some data as quickly as possible" loop. There are a few options for implementations of this kind of work in ways that are more multitasking-friendly.

One way is a literal translation of the loop into a pattern that's friendly to the reactor. You can do this with a generator:

from twisted.internet.task import cooperate

class Audio(object):
    def play(self):
        # ... setup stuff ...
        while self.running:
            # ... one chunk of work ...
            yield

def main():
    ...
    cooperate(Audio().play())

cooperate takes an iterator and iterates over it - but not all at once. It iterates it a few times and then gives up control to the reactor. Then it iterates it a few more times and then gives up control again. This continues until the iterator is exhausted (or the reactor is stopped).

Another slightly less literal translation is based on LoopingCall which takes over responsibility for the looping construct, leaving you only to supply the body of the loop:

from twisted.internet.task import LoopingCall

class Audio(object):
    def play(self):
        # ... setup stuff ...

        LoopingCall(self._play_iteration).start(0)

    def _play_iteration(self):
        # ... one chunk of work

This gives you control over the rate at which the loop iterates. The 0 passed to start in this example means "as fast as possible" (wait 0 seconds between iterations) - while remaining cooperative with the rest of the system. If you wanted one iteration per second, you would pass 1 , etc.

Another less literal option is to use a data flow abstraction - for example, Twisted's native producer/consumer system or the newer tubes library - to set up multitasking-friendly data processing pipelines that are further abstraction from the specific "read, process" loop in the linked answer.