Re: Defect Report

From:

cbarron3@ix.netcom.com (Carl Barron)

Newsgroups:

comp.lang.c++.moderated

Date:

Sun, 17 Feb 2008 08:11:07 CST

Message-ID:

<1icfqy0.1sydxbq1yubz7sN%cbarron3@ix.netcom.com>

Tristan Wibberley <maihem-nn1@maihem.org> wrote:

I would like to propose that the reference implementation of
"std::unique_ptr" and of "std::shared_ptr" have a private constructor
taking a pointer instead of public and that friend functions
("make_unique_ptr" and "make_shared_ptr") be provided for creating these
smart pointer types.

Rather than using "deleter" types, I propose that these friend
functions, ???and the smart pointer type they are creating would be
parameterised on a "life" type and would call the "create" member
function, forwarding their arguments via "std::forward". A
"default_life" type template would be defined, specialised for arrays
and non-array types so that the create member function may take a size_t
and evaluate "new T[size]()" for array types (after getting the element
type T), or evaluate "new T(args...)" for non array types.

   I find it strange one wants to force a C style interface on smart
ptrs:)

    there is little difference between the above and

    class foo;
    foo *create_foo();
    void delete_foo(foo *);

  I find the current interface well designed and easy to use without
added baggage in its interface a shared_ptr<T> is a shared_ptr<T>
no matter how it deletes is ptr [or perhaps does not delete it!]
The problem of a wrong new expression should not be severe with proper
usage, of problem specific classes or frww functions.

   FWIW I hope this goes nowhere fast:)

  If you are serious post a working prototype of this someplace on
the web and I may change my mind, but I doubt it .

--
      [ See http://www.gotw.ca/resources/clcm.htm for info about ]
      [ comp.lang.c++.moderated. First time posters: Do this! ]