Re: [boost] [Proposal] A little reasonable modification of boost::bind_t

On May 11, 2013 12:59 AM, "Arthur Athrun" wrote:

Hi, everyone,

I'd like to propose a little modification of boost:: bind to expose more information in boost:: _bi:: bind_t.

Here is my motivation problem. Suppose this is a function which takes

boost:

:_bi::bind_t as a parameter. But at the same time, I'd like to enumerate all binded parameters and do something for specific types, like the following pseudo code shows.

template< class F> void foo(F & f) { foreach(T& arg, f.arguments()) { if(boost::is_same::value) //here ID_t is a user specific type //do something for arg. } }

I found that all binded arguments are stored in a listN object. But unfortunately, the list object (L _l in boost::_bi::bind_t) is set to private access and there is no method to boost::_bi::bind_t to access the listN object and the type of listN.

I don't know why you developers didn't provide a method to access the listN object. But it seems reasonable to provide a public method in , like this

typedef L argu_type;

const L & arguments() const {return l_;}

Also, the class "template<class T> class value" in doesn't provide any access to T. It will be reasonable to add access of type T, like this

public: typedef T val_type;

With such modifications, we can access all binded arguments in the boost::_bi::bind_t. Here is a showcase.

templatevoid Test(VT t, T * p = 0) { //This parameter is not binded, so ignore it! }templatevoid Test(VT t, typename T::val_type * p = 0) { //T::val_type is the parameter's type， //and can use boost::is_same::value to check it! std::cout<<"great! we get a value here!"<

template void retrive_args(boost::_bi::bind_t > & f) { Test<A1>(f.arguments()[_1]); Test<A2>(f.arguments()[_2]); }

int f(int a, double b) { std::cout<

int main(int argc, char * argv[]) { double t = 1.0; retrive_args(boost::bind(f, _1, 0.2)); //Here will retrive the second parameters only. return 0; }

I don't know if I get the right way to implement it. Please let me know if you have another implementation. Any comments or ideas will be appreciated. Thanks.

Sounds like something that might already be possible with Boost.Phoenix...?