On June 21, 2015 6:53:12 AM EDT, Oliver Kowalke wrote:

2015-06-21 12:19 GMT+02:00 Aaron Levy :

...

Are boost::coroutines::symmetric_coroutine<> and boost::coroutines::asymmetric_coroutine<> part of the coroutine library or coroutine2 library. Besides being C++14-only, in what ways does coroutine2 differ from coroutine?

Coroutine2 provides only boost::coroutines2::asymmetric_coroutine<> (boost::coroutines2::coroutine<> is a typedef of the previous one). Symmetric_coroutine<> (introduced for boost.fiber) is not implemented because its functionality/semantic is better represented by class boost::context::execution_context (part of boost.context). Class execution_context is used as building block (implementation) of boost::coroutines2::asymmetric_coroutine<> - the code becomes cleaner.

If there's only one kind of coroutine in coroutine2, why isn't it just named "coroutine"? ___ Rob (Sent from my portable computation engine)

On June 22, 2015 4:41:05 AM EDT, Oliver Kowalke wrote:

2015-06-22 10:29 GMT+02:00 Rob Stewart :

...

If there's only one kind of coroutine in coroutine2, why isn't it just named "coroutine"?

asymmetric_coroutine<> describes the coroutine type (assymetric transfer of execution control),

With only one kind, the documentation already describes how it behaves. Having asymmetric_coroutine suggests other variations are expected in the future.

but coroutine.hpp contains:

template< typename T > using coroutine = asymmetric_coroutine< T >;

so you could use coroutine<> if desired.

If you think asymmetric_coroutine is necessary, and clearly I do not, you should reverse those. That is, asymmetric_coroutine should be the alias. ___ Rob (Sent from my portable computation engine)

On June 22, 2015 1:13:54 PM EDT, "Vicente J. Botet Escriba" wrote:

...

On June 22, 2015 4:41:05 AM EDT, Oliver Kowalke wrote:

...

2015-06-22 10:29 GMT+02:00 Rob Stewart :

...

If there's only one kind of coroutine in coroutine2, why isn't it just named "coroutine"? asymmetric_coroutine<> describes the coroutine type (assymetric transfer of execution control), With only one kind, the documentation already describes how it behaves. Having asymmetric_coroutine suggests other variations are expected in the future.

...

but coroutine.hpp contains:

template< typename T > using coroutine = asymmetric_coroutine< T >;

so you could use coroutine<> if desired. If you think asymmetric_coroutine is necessary, and clearly I do not, you should reverse those. That is, asymmetric_coroutine should be

Le 22/06/15 11:10, Rob Stewart a écrit : the alias.

...

I don't agree. asymmetric_coroutine is the correct term. This let us to add again symmetric_coroutine if the need comeback.

If that must remain a possibility, then there should be no "coroutine" alias and the header should be asymmetric_coroutine.hpp. If symmetric_coroutine was removed for good reason, does that reason not preclude its return? I don't have that answer. I'm just inferring from the result. ___ Rob (Sent from my portable computation engine)