Re: Howto use pause and resume with latest IO

2011/5/23 Jan-Paul Bultmann <[email protected]>

>
>
>
> Hey MailingList, is there a Turorial on embedding Io?
>
>
> On May 23, 2011, at 3:01 PM, petri rokka wrote:
>
>
>
> I fail to see why application should always exit when yieldingCoros
> empties. It is a normal state because yieldingCoros is empty when i start io
> interpreter.
>
> Yes, but then there is already a Coroutine running. When you yield a
> coroutine you will append the one to yield to yieldingCoros, then remove the
> first in yieldingCoros and start it by calling resume on it.
> You see it may happen that yieldingCoros is empty, in this case the
> currently running coro will simply continue. But notice, at no point the
> coroutine that yields is stopped! This is all done by the resuming
> coroutine. This also means that there can never be more than one coroutine
> running at once because the newly started will always replace the running.
> And there can never be zero coroutines running (what would you replace the
> currently running with?).
> This is because coroutines are not about starting and stopping
> calculations, there are about context switching at the right time.
>

But there can be zero coroutines yielding. :) And that is what i want to
have by pausing the last yielding coro.

>
> I need to pause the last coro to give back control to the C++ app that
> called the io evaluation function. Io is running in a separate thread in the
> C++ app, but it needs to do c++ stuff to communicate with other threads.
>
> when you call doFile/String/Whatever out of your c++ code it should return
> there once the handling is done. I think.
>

Yes it does. But i have to have zero yielding coros.. :)


>
> I made overloaded implementation of pause that does not exit when
> yieldingCoros empties and it seems to work fine.
>
> Wanna see code^^, I suposse the coroutine of you simply continues to run,
> finishes its work and then exits.
>
>
I simply removed the exit


Coroutine do(
    pause := method(
        yieldingCoros remove(self)
        if(isCurrent,
            next := yieldingCoros removeFirst
            if(next,
                next resume
            )
        ,
            yieldingCoros remove(self)
        )
    )
)


>
> How can i get the currentCoro If I try-catch exception inside actor method.
>
> actor doSomething := method(
>     e := try(
>                 actor realCoro := ????
>         doFile("test_that_yields_many_times.io")
>     )
> )
>
> You could do the workaround from steve. Before pausing it you could save
> the currentCoroutine into a slot.
>

I can do that.. but does "actor pause" work after that try block?


>
> Cheers Jan
>
>
> ...
> actor doSomething
> ...
> actor pause
> ...
> actor realCoro resume
> ...
>
>
>
> 2011/5/23 Jan-Paul Bultmann <[email protected]>
>
>>
>>
>>
>> On May 18, 2011, at 3:58 PM, petri rokka wrote:
>>
>>
>>
>> Great that you found a fix :) . Can you post the fix here so i can try it
>> out. Should my posted test app work with it?
>>
>> Btw. I am also having problem with pause. I need to pause/resume  the only
>> existing coro in application to pause it and continue it later (io is
>> embedded in my c++ app) , but its impossible because pause forces
>> application system exit at that point. :/
>>
>> Could you make different pause that doesn't force the system exit?
>>
>>
>> Hey sorry I write that late, but I was on the Typo Berlin, so not much net
>> and even less time.
>>
>> I don't think that what you want to do is possible with coroutines.
>> The reason is that when pause is called, the coroutine wont actually
>> pause. Instead another coroutine will resume, taking care of replacing and
>> thus stopping the currently running.
>> I would instead recommend to run the embedded VM in it's own thread and
>> pause it externally, or use the thread Addon, though I don't know how to use
>> that.
>>
>> Hope it helps.
>> Cheers Jan
>>
>> 2011/5/17 Jan-Paul Bultmann <[email protected]>
>>
>>>
>>>
>>>
>>> On May 17, 2011, at 8:50 PM, Steve Dekorte wrote:
>>>
>>>
>>>
>>>
>>> On 2011-05-17 Tue, at 06:34 AM, Jan-Paul Bultmann wrote:
>>> > Why is a new coroutine generated anyways? And especially where? I would
>>> consider it unwanted behavior if you resume one Coroutine but your stuff
>>> will run in another.
>>>
>>> There are a few cases:
>>> 1) when you do a try() - exceptions handling is implemented with coros.
>>> 2) when a coro gets close to running out of stack space, it creates a new
>>> coro to continue executing in.
>>>
>>> Cool to know, thanks :)
>>>
>>>
>>> It's surprising this hasn't come up before, but your code exposes a bug
>>> in the current actor implementation. I'll add a bit of code to Coroutine to
>>> make sure resumes are sent to the proper child coroutine.
>>>
>>> - Steve
>>>
>>>
>>>
>>>
>>
>>
>>
>>
>
>
>  
>