Re: [Vala] Documenting 'async' / WAS: Further speculations on couroutines, generators and threads : Emulating Go's goroutines and channels in Vala

[Serge Hulne <serge hulne gmail com>]

I am right in assuming that:


2. Async methods have virtually no use outside the scope of event-driven
GTK+ applications (except perhaps as a way to implement additional Vala
features like Luca Bruno's generator, but at the cost of
runtime-performance).

3. There is always a way to write an equivalent program without async
methods (with just ordinary function calls), except for applications with a
graphical interface (GTK+ applications).

4. That a program avoiding async methods will always :
        - be more readable (simpler).
        - perform faster.

5. That async methods are not about runtime performance, but exclusively
about:
        - callbacks for graphical interfaces.
        - convenience (as an alternative to an ordinary function call, when
it seems convenient to do so).

6. That async methods are not at all meant to provide a way to run ancillary
tasks in the background, in order to gain runtime performance.

Serge.



On Fri, Jul 15, 2011 at 12:11 AM, Luca Bruno <lethalman88 gmail com> wrote:

> On Thu, Jul 14, 2011 at 04:52:52PM -0500, Jim Peters wrote:
> > There seems to be special stuff associated with an async method which
> > I haven't found documented well anywhere.  For async method 'method',
> > there are all these ways to use it:
> >
> > - Call 'method' from non-async code, starts it running until its first
> >   'yield', at which point it returns to the caller.  (Correct?)
> >
> >     method(args);
> >
> > - Get callback (from within method itself).  This is the callback to
> >   resume execution after the following 'yield'.  (Correct?)
> >
> >     SourceFunc callback = method.callback;
> >
> > - Use the resume callback from elsewhere.  'method' resumes and takes
> >   control again, running until its next 'yield', at which point it
> >   returns from this call.  (Correct?)
> >
> >     callback();
> >
> > - Give up control and return to the caller.  This doesn't guarantee in
> >   any way that the method will be resumed, i.e. callback() must be
> >   called somewhere else.  (Correct?)
> >
> >     yield;
> >
> > - Give up control but arrange for a resume callback when idle.  This
> >   requires the main loop to be running.  (Correct?)
> >
> >     Idle.add(method.callback);
> >     yield;
> >
> > - Call async method 'other_method' from 'method'.  This automatically
> >   sets up a callback for 'method' to resume itself and collect the
> >   return value when 'other_method' completes.  (Correct?)
> >
> >     yield other_method(args);
>
> All correct.
>
> >   Question: At the C level, I guess this first calls forwards to
> >   'other_method' to start it before returning to the caller due to the
> >   'yield'.  If the 'other_method' also yields, then there is no
> >   problem, but if 'other_method' finishes without yielding (e.g. if it
> >   can return the result right away without doing any asynchronous
> >   work), then the 'method' callback would be called again, only a few
> >   stack frames lower.  In theory if 'method' called 'other_method'
> >   repeatedly like this, the stack could overflow.  Is this correct?
> >   If so maybe that needs documenting.
>
> Variables in async methods are allocated on the heap. By the way I think I
> didn't understand well the problem you're raising.
>
> > - Call with .begin().  Question: Is this just like the 'method()' call
> >   but adding a callback request for when method() finishes?  If so
> >   will it also just run until the first 'yield' and then return?
> >
> >     method.begin(args, callback);
>
> Yes.
>
> > - Call with .end().  For use within the 'begin' callback, to get
> >   method return value and clean up.  Question: Is an .end() call
> >   required for plain 'method()' style invocation?
> >
> >     method.end(out_args);
>
> Can you rephrase?
>
> > Also, questions:
> >
> > - Are there any other special features associated with async methods?
> >   (Any other .<identifier> features, for example.)
>
> No.
>
> > - When it is necessary to have a main loop running?  Some of the
> >   generated C code requests an idle callback.  I guess this won't work
> >   without a main loop running.  But the Generator, for example, seems
> >   to run through fine without a main loop.
>
> Mainloop is required only for async methods that never use yield.
>
> > - The docs say "The .callback call is used to implicitly register a
> >   _finish method for the async method".  What does this mean?  Also:
> >   "end() is a syntax for the *_finish method".  I'm confused.
>
> Try looking at the generated C code.
>
> --
> http://www.debian.org - The Universal Operating System