Hi,
For certain reasons I can't paste code here, so I'll try to give a
pseudo-code algorithm that worked for me with Proton-C. Don't know how much
control you have with the Python client.
--- Step 1: init ---
- Initialize proton/main event loop up to the moment you get the FLOW
event --> yield
---- Step 2: program main loop ----
While(true)
{
Read from external
If (nothing read) then
{
if (AMQP echo must be sent) then
{
amqp_send(echo_message)
adjust the tracking of link credit
resume proactor main event loop (proactor_done + proactor_wait) and
handle events --> if FLOW event is seen then yield
continue
}
}
else if (message read)
{
amqp_send(message)
adjust the tracking of link credit
resume proactor main event loop (proactor_done +
proactor_wait) and handle events --> if FLOW event is seen --> yield
}
}
Others may have better ideas, I am just beggining to use Proton-C.
Hope this helps,
Adrian
On Thu, Jul 16, 2020, 8:36 AM Francesco Raviglione <
francescorav.es...@gmail.com> wrote:
> Dear Adrian,
> Thank you very much for your reply.
>
> I tried following your suggestion, by attempting to find a way to call an
> external function reading the data from an external source, and returning
> it when available.
>
> However, unfortunately, it seems to be slightly easier to control the
> behaviour of the event loop in C than in Pyhton.
> In particular, in Python, if I call any blocking function (even a
> time.sleep(0.1), sleeping for just 100 ms) inside "on_sendable", the
> messages are only buffered and never sent (do you know why? Is the
> "on_sendable" handler supposed to be executed as fast as possible without
> blocking to have the messages being sent immediately?).
> If, instead, I just do not block and send the data only if available (for
> instance by implementing a mechanism equivalent to a non-blocking read from
> a pipe, with poll(), letting "on_sendable" finish without sending anything,
> if there is no data available), only one "on_sendable" event is generated
> and the event loop does not seem to return any other event which I could
> use to send the data, which may be now available (for example, the
> equivalent of C's PN_DELIVERY does not seem to exist).
>
> Do you know if there are other ways to solve this issue?
> When you were "yielding" the control of the main loop to an external
> function, was it blocking in some way, waiting for data to be available?
>
> Thank you very much,
> Francesco
>
>
> Il giorno mer 15 lug 2020 alle ore 18:40 Adrian Florea <
> florea....@gmail.com>
> ha scritto:
>
> > Hi Francesco,
> >
> > I achieved similar behavior but using Qpid Proton-C (AMQP send-only
> > program).
> >
> > After initialiazing proton and getting the first FLOW event, I simply
> > "yield" main event loop control to an external/looping function that can
> > read from a different source and then send the same message to an AMQP
> > destination.
> >
> > It may seem hard in the beggining but once you get a hold on how to
> > integrate proton events loop into your program loop, it has good chances
> to
> > work.
> >
> > Adrian
> >
> >
> > On Wed, Jul 15, 2020, 11:22 AM Francesco Raviglione <
> > francescorav.es...@gmail.com> wrote:
> >
> > > Dear all,
> > > I'm trying to use the Python version of Qpid Proton to send data to an
> > AMQP
> > > 1.0 broker.
> > > I've seen that several examples are sending data which was
> > > available/defined before the AMQP event loop is started.
> > > However, I need to send data only when it becomes available from an
> > > external process (for example by waiting for it on a Pipe, with
> > > self.amqp_pipe_end.recv()), and then send it immediately to the broker.
> > > If I try to wait on a pipe inside "on_sendable", no message is actually
> > > transferred to the broker, as the event loop is "blocked" waiting for
> new
> > > data and cannot manage properly the AMQP 1.0 connection (is this
> > correct?).
> > > How could I achieve the desired result? How can I make the loop wait
> for
> > > external data and then immediately send it?
> > >
> > > Thank you very much in advance,
> > > Francesco Raviglione
> > >
>
