Movatterモバイル変換

This page is a snapshot from the LWG issues list, see theLibrary Active Issues List for more information and the meaning ofResolved status.

823.identity<void> seems broken

Section: 22.2.4[forward]Status:ResolvedSubmitter: Walter BrownOpened: 2008-04-09Last modified: 2016-01-28

Priority:Not Prioritized

View all otherissues in [forward].

View all issues withResolved status.

Discussion:

N2588 seems to have added anoperator() member function to theidentity<> helper in 22.2.4[forward]. I believe this change makes it nolonger possible to instantiateidentity<void>, as it would requireforming a reference-to-void type as thisoperator()'s parameter type.

Suggested resolution: Specializeidentity<void> so as not to requirethe member function's presence.

[Sophia Antipolis:]

Jens: suggests to add a requires clause to avoid specializing onvoid.

Alisdair: also consider cv-qualifiedvoid.

Alberto provided proposed wording.

[2009-07-30 Daniel reopens:]

This issue became closed, because theReferentType requirementfixed the problem - this is no longer the case. In retrospective it seemsto be that the root of current issues aroundstd::identity (823,700⁽ⁱ⁾,939⁽ⁱ⁾)is that it was standardized as something very different (an unconditionaltype mapper) than traditional usage indicated (a function object that shouldderive fromstd::unary_function), as the SGI definition does. This issue couldbe solved, ifstd::identity is removed (one proposal of939⁽ⁱ⁾), but until thishas been decided, this issue should remain open. An alternative forremoving it, would be, to do the following:

1. Letidentity stay as areal function object, which wouldnow properlyderive fromunary_function:

   template <class T> struct identity : unary_function<T, T> {  const T& operator()(const T&) const;};

2. Invent (if needed) a generic type wrapper (corresponding to conceptIdentityOf),e.g.identity_of, and move it's prototype description back to 22.2.4[forward]:

   template <class T> struct identity_of {  typedef T type;};

   and adapt thestd::forward signature to useidentity_ofinstead ofidentity.

[2009-10 Santa Cruz:]

Mark asNAD EditorialResolved, fixed by939⁽ⁱ⁾.

Proposed resolution:

Change definition ofidentity in 22.2.4[forward], paragraph 2, to:

template <class T>  struct identity {    typedef T type;requires ReferentType<T>      const T& operator()(const T& x) const;  };

...

requires ReferentType<T>    const T& operator()(const T& x) const;

Rationale:

The point here is to able to writeT& givenT andReferentType isprecisely the concept that guarantees so, according to N2677(Foundational concepts). Because of this, it seems preferable than anexplicit check forcv void usingSameType/remove_cv as it was suggestedin Sophia. In particular, Daniel remarked that there may be types otherthancv void which aren't referent types (int[], perhaps?).