boost::range vs && (rvalue ref)
Passing range by ref (&) is fine, but using (&&) results in and error. Any
ideas?
#include
Passing range by ref (&) is fine, but using (&&) results in and error. Any ideas?
[snip]
When you pass by T&&, the deduced type of T will include the reference. (So e.g. if you have a T&& argument and you pass in an int rvalue, T will be deduced as 'const int&'. If you pass an int lvalue, T will be deduced as 'int&'. This is in contrast to a regular reference, like T&, where if you pass an int lvalue, T will be deduced as just 'int'). Range metafunctions don't accept references to range types as their input; you must pass the range type itself. So, you must call remove_reference<> on the deduced type 'T' prior to passing it to a range metafunction. I wonder if we should change the range metafunctions to accept references to range types, and do the reference stripping themselves? Regards, Nate
On 11-01-2013 04:37, Nathan Ridge wrote:
Passing range by ref (&) is fine, but using (&&) results in and error. Any ideas?
[snip]
When you pass by T&&, the deduced type of T will include the reference. (So e.g. if you have a T&& argument and you pass in an int rvalue, T will be deduced as 'const int&'. If you pass an int lvalue, T will be deduced as 'int&'. This is in contrast to a regular reference, like T&, where if you pass an int lvalue, T will be deduced as just 'int').
Range metafunctions don't accept references to range types as their input; you must pass the range type itself.
So, you must call remove_reference<> on the deduced type 'T' prior to passing it to a range metafunction.
I wonder if we should change the range metafunctions to accept references to range types, and do the reference stripping themselves?
I think we should. This has come up before. -Thorsten
Thorsten Ottosen wrote:
On 11-01-2013 04:37, Nathan Ridge wrote:
Passing range by ref (&) is fine, but using (&&) results in and error. Any ideas?
[snip]
When you pass by T&&, the deduced type of T will include the reference. (So e.g. if you have a T&& argument and you pass in an int rvalue, T will be deduced as 'const int&'. If you pass an int lvalue, T will be deduced as 'int&'. This is in contrast to a regular reference, like T&, where if you pass an int lvalue, T will be deduced as just 'int').
Range metafunctions don't accept references to range types as their input; you must pass the range type itself.
So, you must call remove_reference<> on the deduced type 'T' prior to passing it to a range metafunction.
I wonder if we should change the range metafunctions to accept references to range types, and do the reference stripping themselves?
I think we should. This has come up before.
-Thorsten
My objective is to support this common type of usage: template<typename out_t> F (out_t & o) { } F (boost::make_iterator_range (blah.begin(), blah.end()));
Thorsten Ottosen wrote:
On 1--1--013 4::7,, Nathan Ridge wrote:
Passing range by ref (&) is fine, but using (&&) results in and error. Any ideas?
[snip]
[snip]
Range metafunctions don't accept references to range types as their input; you must pass the range type itself.
So, you must call remove_reference<> on the deduced type 'T' prior to passing it to a range metafunction.
I wonder if we should change the range metafunctions to accept references to range types, and do the reference stripping themselves?
I think we should. This has come up before.
-Thorsten
My objective is to support this common type of usage:
template<typename out_t> F (out_t & o) { }
F (boost::make_iterator_range (blah.begin(), blah.end()));
I filed https://svn.boost.org/trac/boost/ticket/885 .
In the meantime, you can use the remove_reference workaround I mentioned,
as follows:
template<typename out_t>
void F (out_t && o)
{
typedef typename boost::range_value
participants (3)
-
Nathan Ridge -
Neal Becker -
Thorsten Ottosen