Fourth example from PEP 342

[Thomas Jollans]

On 19/01/18 22:50, Léo El Amri wrote:
> Hello list,
> 
> I am currently trying to learn co-routine/asynchronous mechanisms in
> Python. I read the PEP 342, but I stumble on the fourth example.
> I don't understand what the lines "data = yield nonblocking_read(sock)"
> in echo_handler() and "connected_socket = yield
> nonblocking_accept(sock)" in listen_on() are trying to do.
> 
> For example, can someone explain me how the returned value in the line
> "connected_socket = yield nonblocking_accept(sock)" can be used on the
> next line ("trampoline.add(handler(connected_socket))") ? To me, it
> looks like the returned value is lost in the Trampoline, when resume()
> gets the returned value of the yield expression.
> 
> Léo
> 

Let's see.

Upon the line

  connected_socket = yield nonblocking_accept(sock)

control is returned to t.resume. nonblocking_accept is supposed to be a
coroutine, so t.resume does this:

  if isinstance(value, types.GeneratorType):
      # Yielded to a specific coroutine, push the
      # current one on the stack, and call the new
      # one with no args
      self.schedule(value, (coroutine,stack))

Ergo, it schedules the nonblocking_accept coroutine (‘value’) to be
called on the next iteration, and keeps the running listen_on coroutine
(‘coroutine’) on the stack for safekeeping.

On a subsequent iteration (jump?) of the trampoline, nonblocking_accept
will yield a socket. Then, resume will

  elif stack:
      # Yielded a result, pop the stack and send the
      # value to the caller
      self.schedule(stack[0], stack[1], value)

plan to return control to listen_on (‘stack[0]’), and (the next time
around) send the value back in.

  if exc:
      value = coroutine.throw(value,*exc)
  else:
      value = coroutine.send(value)

Now, listen on is running again and has the value yielded by
nonblocking_accept.

I wish I knew how exactly nonblocking_accept was supposed to work here,
but that's beside the point as the world of Python coroutines and async
I/O has moved on since 2005.

Hope this helps
Thomas