Re: [sigc] std::function<> and std::bind() in libsigc++

[Chris Vine <chris cvine freeserve co uk>]

On Sat, 26 Oct 2013 11:22:15 +0200
Kjell Ahlstedt <kjell ahlstedt bredband net> wrote:
> As a result of the discussion in
> https://mail.gnome.org/archives/libsigc-list/2013-July/msg00007.html
> I have at last pushed a libsigc++ commit, where I explain how 
> std::function<> can be used in libsigc++, and why it's not that easy
> to use std::bind().
> https://git.gnome.org/browse/libsigc++2/commit/?id=49e753a7e2d2ce85bc090da47b80c78a24c065d6
>
> I said in July that std::function should be mentioned wherever lambda 
> expressions are mentioned. That's not exactly what I've done. I've 
> realized that std::function can be an alternative to a sigc::slot
> rather than an alternative to a lambda expression.

I wouldn't say that std::function can be an "alternative" to
sigc::slot: it is that sigc::slot can encapsulate any functor and
therefore can wrap a std::function object.

However, you would not often want to do that.  You would normally pass
the underlying functor that std::function might wrap (say, a lambda
expression) to sigc::slot directly.  As I mentioned in the
correspondence to which you refer, to encapsulate the underlying
functor using std::function, and then encapsulate that using
sigc::slot, causes double virtual dispatch (as well as additional
memory allocation). Both sigc::slot and std::function perform type
erasure.

It follows that I think the text "It is often possible to replace
sigc::slot<> by the C++11 class std::function<>, for instance ..." is a
little misleading.  It is not replacing it, it is wrapping it by
implicit conversion. But as I say, better to encourage people to pass
the lambda expression (or whatever) directly.

> As is now explained in the documentation of
> sigc::signal<>::connect(), the problem with std::bind() is that it
> creates a functor whose operator()() is a variadic template. 
> SIGC_FUNCTORS_DEDUCE_RESULT_TYPE_WITH_DECLTYPE can't deduce the
> result type of such a functor. I've no idea how (or if) that can be
> fixed.
>
> _______________________________________________
> libsigc-list mailing list
> libsigc-list gnome org
> https://mail.gnome.org/mailman/listinfo/libsigc-list