Le 06/03/2016 15:29, Louis Dionne a écrit :

...

On Mar 6, 2016, at 04:59, Vicente J. Botet Escriba wrote:

Le 06/03/2016 09:59, Andrey Semashev a écrit :

...

On 2016-03-06 05:21, Vicente J. Botet Escriba wrote:

...

Others could be considered also as function.hpp, lambda.hpp and lift.hpp, as the macros are there to workaround some missing language features, but those are much more specialized (Boost.Core?) I don't think Boost.Core is the right place for them as this library is for small generally useful components used by many libraries. The components you pointed out seem too specialized to me, and Boost.Fit looks like the right place for them.

Andrey, I was thinking in Boost.Core as these are language-like emulation features, like e.g. addressof, enable_if, explicit_operator_bool, ignorunused, no_exception_support, noncopyanble, scoped_enum, typeinfo.

IIUC, BOOST_FIT_STATIC_FUNCTION try to fix a standard Core issue (pending issue 2104 in CWG) identified by Eric Nibler, with the a solution based on proposal's Eric (http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4381.html). The library has also a macro BOOST_FIT_DECLARE_STATIC_VAR to do it for any data variable. I would like something that solves this problem in Boost.

BOOST_FIT_STATIC_LAMBDA try to covers the C++17 feature constexpr lambdas.

I suspect that Boost.Hana should have something like that. Louis, could you tell us how do you manage with these issues? I don’t manage these issues. Hana uses only hand-written function objects that can be marked constexpr, and lambdas are completely excluded from the codebase. There’s no way to properly workaround the limitations of constexpr lambdas, since the lambda’s operator() won’t be constexpr even with BOOST_FIT_STATIC_LAMBDA.

Sorry it was me that did the association of BOOST_FIT_STATIC_LAMBDA to C++17 constexpr lambdas.

"The |BOOST_FIT_STATIC_LAMBDA| macro allows initializing non-capturing lambdas at compile-time in a |constexpr| expression."

Nevertheless we get the ability to declare them at the namespace level, which is already a good step, or do you consider that this is not useful? Paul, if Louis is right, I believe that the documentation should state clearly that even if the resulting function can be assigned to a constexpr variable the user can not call it at compile-time, and even if this is obvious for the experts. Vicente