Skip to content

coroutines

Chris Vine edited this page Jan 17, 2020 · 2 revisions

The (a-sync coroutines) module provides the following procedures:


(make-iterator proc . args)

This procedure takes a generator procedure, namely a procedure which has a 'yield' parameter for its first or only argument, followed by such other arguments (other than the one for the 'yield' parameter) as the generator procedure requires, and constructs an iterator from them. When the iterator (which optionally takes no or one argument) is invoked, it will begin executing the procedure unless and until the argument comprising the yield procedure is called, which will cause the iterator to suspend computation and instead return the value passed to yield, if any (yield is a procedure taking no or one argument). If invoked again, the iterator will resume computation at the point where it last left off (the yield procedure returning the value, if any, passed to the iterator on resuming), following which yield may be applied again, and so on. When the generator procedure has executed to the end, the iterator returns 'stop-iteration. This procedure has some resemblance to call/ec, except that (i) instead of executing the passed procedure immediately, it returns an iterator which will do so, (ii) it is resumable, and (iii) the procedure to be executed can receive starting arguments in addition to the yield/break argument, to provide an alternative to binding them with a lambda closure. It is similar to ECMAScript generators and python generators.


(make-coroutine proc . args)

This procedure takes a generator procedure, namely a procedure which has a 'yield' parameter for its first or only argument, followed by such other arguments (other than the one for the 'yield' parameter) as the generator procedure requires, and constructs a coroutine. It is similar to make-iterator, in that it takes a generator procedure and returns a lambda object (a coroutine) which when called will begin executing the generator procedure unless and until the argument comprising the yield procedure is called, which will cause computation to be suspended. However unlike make-iterator, the resumption continuation generated on yielding is returned by the coroutine when yielding rather than being stored internally in an iterator, so there is no explicit retained mutable state. The return value of the coroutine comprises two values: first the resumption continuation, and second the value (if any) passed to 'yield' when called. If the returned resumption continuation is subsequently called again, computation will be resumed at the point where it last left off (the yield procedure returning the value, if any, passed to the continuation on resuming) until it completes or it again calls the yield procedure.

Upon the generator procedure finally completing, the value to which it evaluates is returned by the resumption continuation together with a continuation value of #f. This differs from the behaviour of make-iterator, which returns 'stop-iteration when the generator procedure finishes to completion and ignores its return value.


(a-sync waitable . args)

a-sync takes a waitable procedure (namely a procedure which takes 'await' as its first parameter, which is a yield procedure obtained by a call to make-iterator, and 'resume' as its second parameter, which is an iterator constructed by make-iterator), followed by such other arguments (if any) as the waitable procedure requires to be passed on to it. The 'resume' argument must only be called by an asynchronous callback, and the 'await' argument must only be called by the waitable procedure in order to block until the callback is ready to let it resume. When it unblocks, the 'await' argument returns the value (if any) passed to 'resume' by the callback. This async procedure must be called in the same thread as that in which the event loop runs (as must 'await' and 'resume').

None of the code in the waitable procedure should block on other things in the program, except by calls to await (which do not in fact block, even though they appear to do so).

The way it works is that the call to a-sync will begin executing the waitable procedure and will return as soon as the first (or only) call to 'await' is made by that procedure, or as soon as the waitable procedure returns if it makes no calls to 'await'. Any subsequent resumptions of the waitable procedure will take place in the event loop concerned as delimited continuations, via the calls to 'resume' made by the callbacks. The effect of the waitable procedure subsequently ending, or of further calls to 'await' being made within the same waitable procedure, is to return control to the event loop by recommencing execution at the point where the most recent previous call to 'resume' was made by the last callback to execute.

An exception thrown in a waitable procedure before the first call to 'await' to be made by it which is not handled locally will propagate out of the a-sync procedure where it may be caught normally. However, if so caught but a callback established by the same waitable procedure call still runs and invokes 'resume', the waitable procedure will begin running again in the callback. If the same exception is thrown again in consequence, it will propagate out of 'resume' in that callback, and then out of the event loop - if the event loop in the event-loop module is used, this means that it will propagate out of the call to event-loop-run!. It is therefore best if such exceptions are handled locally within the waitable procedure. Any exception thrown in the waitable procedure after the first call to 'await' which is not handled locally will propagate into the previously called callback at the point where 'resume' was last called. If this is handled in the callback, then control will be returned to the event loop and the remainder of the waitable procedure will not execute. If that exception is not handled locally in the callback, or if the callback throws an exception of its own, then it will propagate out of the event loop - if the event loop in the event-loop module is used, this means that it will propagate out of the call to event-loop-run!. If an exception propagates out of event-loop-run! for that or some other reason, then the event loop will be left in a valid state and it will be as if event-loop-quit! had been called on it, but it is then up to the user to catch that exception once it is out of event-loop-run! if she does not want the program to terminate.

After the call to 'resume', the callback should normally just return (with a #t or #f value in the case of a file watch or a timeout on an event-loop object from the (a-sync event-loop) module). If a-sync is used with a file watch or timeout on an event-loop object constructed by make-event-loop, the watch callback or timeout callback should normally, when the call to 'resume' returns, either always return #f (so the callback only fires once) or always return #t (so it is responsibility of the waitable procedure to terminate the watch or timeout repetitions). That way, there can never be a case where the callback has been removed from the event loop by returning false but the waitable procedure still thinks it has a call to 'await' to be made. The (a-sync event-loop) module has await-task!, await-task-in-thread!, await-task-in-event-loop!, await-yield!, await-generator!, await-generator-in-thread!, await-generator-in-event-loop!, await-timeout! and await-sleep! convenience procedures, the (a-sync await-ports) module has await-read-suspendable!, await-write-suspendable!, await-getline!, await-geteveryline!, await-getsomelines!, await-getblock!, await-geteveryblock!, await-getsomeblocks!, await-put-bytevector! and await-put-string! procedures, and the (a-sync thread-pool) module has await-task-in-thread-pool! and await-generator-in-thread-pool! procedures, which will correctly set this up for you automatically. If those convenience procedures are used, exceptions should always be handled locally in the waitable procedure (and if the callback might throw, in the callback also) if it is undesirable that uncaught exceptions propagate out of event-loop-run!. In the case of await-task-in-thread!, await-task-in-thread-pool!, await-generator-in-thread! and await-generator-in-thread-pool!, those procedures also take an optional handler argument which will handle any exceptions thrown by the task or generator: otherwise a throwing thread would terminate the program if not caught.

There can be as many calls to 'await' and asynchronous callbacks in any one waitable procedure as wanted, to enable composition of asynchronous operations. However, you cannot run two or more asynchronous tasks at the same time with the same await-resume pair without an intervening call to await except by doing extra work, because the first call to 'await' will match the first callback which happens to call 'resume', and so on. In such cases, 'resume' would need to return something like a key-value pair so that the result can be correctly identified. Accordingly this practice is discouraged. Instead, when composing asynchronous tasks within any one waitable procedure, operate on a 'start-task --> await-on-result --> start-task --> await-on-result ...' basis, and make calls to a-sync on separate waitable procedures for tasks which are to run independently (as soon as any code calls a-sync's 'await' procedure in any a-sync block, a-sync will return and whatever follows it will begin executing, including any following a-sync block). The convenience procedures mentioned above make this easy for many use cases (see the examples below). A waitable procedure can itself call a-sync to construct another await-resume pair for the purpose of starting other asynchronous events.

So, don't do this:

(set-default-event-loop!) ;; if none has yet been set
(a-sync (lambda (await resume)

          (timeout-post! 100
                         (lambda ()
                           (resume "expired")
                           #f))
          (event-post! (lambda ()
                         (resume (+ 1 1))))

          (simple-format #t "Timeout ~A\n" (await))
          (simple-format #t "1 + 1 is ~A\n" (await))))
(event-loop-run!)

Instead do this, where the two asynchronous operations are to run consecutively:

(set-default-event-loop!) ;; if none has yet been set
(a-sync (lambda (await resume)

          (timeout-post! 100
                         (lambda ()
                           (resume "expired")
                           #f))
          (simple-format #t "Timeout ~A\n" (await))
          (event-post! (lambda ()
                         (resume (+ 1 1))))

          (simple-format #t "1 + 1 is ~A\n" (await))))
(event-loop-run!)

or this, where the two asynchronous operations are to run concurrently (the summing of 1 and 1 will finish before the timeout):

(set-default-event-loop!) ;; if none has yet been set
(a-sync (lambda (await resume)
          (timeout-post! 100
                         (lambda ()
                           (resume "expired")
                           #f))
          (simple-format #t "Timeout ~A\n" (await))))
(a-sync (lambda (await resume)
          (event-post! (lambda ()
                         (resume (+ 1 1))))
          (simple-format #t "1 + 1 is ~A\n" (await))))
          
(event-loop-run!)

or this, to loop asynchronously with a series of results:

(set-default-event-loop!) ;; if none has yet been set
(event-loop-block! #t) ;; worker thread in operation
(a-sync (lambda (await resume)
          ;; executing task in event loop thread
          (let loop ((count 0))
            (if (< count 5)
                (begin
                  (event-post! (lambda () (resume (* 2 count))))
                  (loop (1+ count)))
                (event-post! (lambda () (resume 'stop-iteration)))))
          (let loop ()
            (let ((res (await)))
              (when (not (eq? 'stop-iteration res))
                (display res)(newline)
                (loop))))

          ;; executing task in a worker thread
          (call-with-new-thread
           (lambda ()
             (let loop ((count 5))
               (if (< count 10)
                   (begin
                     (event-post! (lambda () (resume (* 2 count))))
                     (loop (1+ count)))
                   (event-post! (lambda () (resume 'stop-iteration)))))))
          (let loop ()
            (let ((res (await)))
              (when (not (eq? 'stop-iteration res))
                (display res)(newline)
                (loop))))
          (event-loop-block! #f)))
(event-loop-run!)

Or better, for many use cases when composing on the event loop from the (a-sync event-loop) module, you can use the await helper procedures, as follows (in this example the two asynchronous operations run consecutively):

(set-default-event-loop!) ;; if none has yet been set
(a-sync (lambda (await resume)
          (simple-format #t
                         "Timeout ~A\n"
                         (await-timeout! await resume
                                         100
                                         (lambda ()
                                           "expired")))
          (simple-format #t "1 + 1 is ~A\n"
                         (await-task! await resume
                                      (lambda ()
                                        (+ 1 1))))))
(event-loop-run!)

or this, where the two asynchronous operations run concurrently (the summing of 1 and 1 will finish before the timeout)

(set-default-event-loop!) ;; if none has yet been set
(a-sync (lambda (await resume)
          (simple-format #t
                         "Timeout ~A\n"
                         (await-timeout! await resume
                                         100
                                         (lambda ()
                                           "expired")))))
(a-sync (lambda (await resume)
          (simple-format #t "1 + 1 is ~A\n"
                         (await-task! await resume
                                      (lambda ()
                                        (+ 1 1))))))
(event-loop-run!)

or this, to loop asynchronously with a series of results:

(set-default-event-loop!) ;; if none has yet been set
(event-loop-block! #t) ;; worker thread in operation
(a-sync (lambda (await resume)
          (await-generator! await resume
                            (lambda (yield)
                              (let loop ((count 0))
                                (when (< count 5)
                                  (yield (* 2 count))
                                  (loop (1+ count)))))
                            (lambda (val)
                              (display val)
                              (newline)))
          (await-generator-in-thread! await resume
                                      (lambda (yield)
                                        (let loop ((count 5))
                                          (when (< count 10)
                                            (yield (* 2 count))
                                            (loop (1+ count)))))
                                      (lambda (val)
                                        (display val)
                                        (newline)))
          (event-loop-block! #f)))
(event-loop-run!)

Or the compose-a-sync macro in the (a-sync compose) module can be used so that these examples are constructed as follows (in this first case the two asynchronous operations run consecutively)

(set-default-event-loop!) ;; if none has yet been set
(compose-a-sync ((ret1 (await-timeout! 100
                                       (lambda ()
                                         "expired")))
                 (ignore ((no-await (simple-format #t "Timeout ~A\n" ret1))))
                 (ret2 (await-task! (lambda ()
                                      (+ 1 1)))))
           ((no-await (simple-format #t "1 + 1 is ~A\n" ret2))))
(event-loop-run!)

and this, where the two asynchronous operations run concurrently (the summing of 1 and 1 will finish before the timeout)

(set-default-event-loop!) ;; if none has yet been set
(compose-a-sync ((ret (await-timeout! 100
                                      (lambda ()
                                        "expired"))))
           ((no-await (simple-format #t "Timeout ~A\n" ret))))
(compose-a-sync ((ret (await-task! (lambda ()
                                     (+ 1 1)))))
           ((no-await (simple-format #t "1 + 1 is ~A\n" ret))))
(event-loop-run!)

or this, to loop asynchronously with a series of results:

(set-default-event-loop!) ;; if none has yet been set
(event-loop-block! #t) ;; worker thread in operation
(compose-a-sync ()
                (await-generator! (lambda (yield)
                                    (let loop ((count 0))
                                      (when (< count 5)
                                        (yield (* 2 count))
                                        (loop (1+ count)))))
                                  (lambda (val)
                                    (display val)
                                    (newline)))
                (await-generator-in-thread! (lambda (yield)
                                              (let loop ((count 5))
                                                (when (< count 10)
                                                  (yield (* 2 count))
                                                  (loop (1+ count)))))
                                            (lambda (val)
                                              (display val)
                                              (newline)))
                ((no-await (event-loop-block! #f))))
(event-loop-run!)

The a-sync procedure can be used with any event loop, including the glib main loop provided by g-golf and guile-gi and so with gtk+ callbacks, and with the event loop in the (a-sync event-loop) module.

For a further overview of the operation of 'a-sync', see the general overview