Re: [boost] [futures] Request for Comments

From: Matt Calabrese (rivorus_at_[hidden])
Date: 2005-08-16 22:49:49

I love the concept of futures and look forward to the completion of
this library. However, do have one critisism.

On 8/16/05, Thorsten Schuett <schuett_at_[hidden]> wrote:
> //double_[] creates a simple_future<double>

Forgive me if am missing the obvious, but is there a reason as to why
you have double_ and int_, etc. with overloaded operator[] to create a
simple_future as opposed to simply having a "make_future" function
which automatically detects which instantiation of simple_future
should be made based off of the return type of the function
pointer/function object being passed.

In other words, why have

double_[ &double_func ];
int_[ &int_func ];

when you can simply have

make_future( &double_func ); // detects the double return type and
returns a simple_future< double >
make_future( &int_func ); // detects the int return type and returns a
simple_future< int >

Having a make_future function gets rid of the need to explicitly state
the return type of the function in the future's creation.

Even still, if you actually want to be explicit, then why would you
make the double_ and int_ types when the desired functionality already
exists by just specifying the template arguments of simple_future?
double_[ &double_func ] seems to offer no benefit over simple_future<
double >( &double_func ), other than possibly less code to type. It
seems to me like completely uneccessary code, especially since you
have similar types for int and I'd imagine other built-ins. I don't
really see an added benefit of these objects concerning the manner in
which you are currently using them.

-Matt Calabrese