Re: [Boost-users] boost::any constructors

Subject: Re: [Boost-users] boost::any constructors
From: Antony Polukhin (antoshkka_at_[hidden])
Date: 2016-02-22 03:37:51

2016-02-16 0:31 GMT+03:00 David <list_at_[hidden]>:

> Hi,
>
> boost::any has a perfect forward constructor declared as:
>
>     template<typename ValueType>
>     any(ValueType&& value
>         , typename boost::disable_if<boost::is_same<any&, ValueType>
> >::type* = 0 // disable if value has type `any&`
>         , typename boost::disable_if<boost::is_const<ValueType> >::type* =
> 0) // disable if value has type `const ValueType&&`
>       : content(new holder< typename decay<ValueType>::type
> >(static_cast<ValueType&&>(value)))
>     {}
>
> The is_const SFINAE exclusion forces const types to the regular copy
> constructor:
>     template<typename ValueType>
>     any(const ValueType & value)
>       : content(new holder<
>             BOOST_DEDUCED_TYPENAME remove_cv<BOOST_DEDUCED_TYPENAME
> decay<const ValueType>::type>::type
>         >(value))
>     {}
>
> What is different about how the regular copy constructor treats a constant
> value than what the perfect forward construct would do if the is_const
> exclusion were removed?
>

Regular copy constructor has been there before the perfect forwarding
constructor.
When perfect forwarding constructor was added, the idea was to enable it
*only* when the value could be moved (to avoid ambiguity). So it was
disabled for all the const ValueType&& and const ValueType.

The difference with regular copy constructor is in remove_cv.

-- 
Best regards,
Antony Polukhin