specializing std::less

[Exits Funnel]

I've got this code:

//Begin foo.cpp
#include <utility>
#include <bits/stl_function.h>

typedef std::pair<int, int> int_pair_t;

template<>
struct std::less<int_p air_t> { };
//End foo.cpp

g++ has this to say on the issue:

foo.cpp:7: specializing `struct std::less<int_p air_t>' in different
namespace
/usr/include/c++/3.2.3/bits/stl_function.h: 195: from definition of `
template<class _Tp> struct std::less'

This seems reasonable (though for what it's worth, it compiles under
VC7.1). I can fix the problem by wrapping the specialization in a
'namespace std' block but I'm wondering if this is prudent or is there
some other solution. Thanks.

-exits

[Victor Bazarov]

Re: specializing std::less

"Exits Funnel" <exitsNOfunnelS PAM@yahoo.com> wrote...[color=blue]
> [...] I can fix the problem by wrapping the specialization in a
> 'namespace std' block but I'm wondering if this is prudent or is there
> some other solution. Thanks.[/color]

It's expressly allowed by the Standard.

Victor

[Sharad Kala]

Re: specializing std::less


"Exits Funnel" <exitsNOfunnelS PAM@yahoo.com> wrote in message[color=blue]
> I've got this code:
>[/color]
[snip]
[color=blue]
> This seems reasonable (though for what it's worth, it compiles under
> VC7.1). I can fix the problem by wrapping the specialization in a
> 'namespace std' block but I'm wondering if this is prudent or is there
> some other solution. Thanks.[/color]

That's fine (Read 17.4.3.1/1)

Sharad

[Dietmar Kuehl]

Re: specializing std::less


Victor Bazarov wrote:[color=blue]
> "Exits Funnel" <exitsNOfunnelS PAM@yahoo.com> wrote...[color=green]
> > [...] I can fix the problem by wrapping the specialization in a
> > 'namespace std' block but I'm wondering if this is prudent or is[/color][/color]
there[color=blue][color=green]
> > some other solution. Thanks.[/color]
>
> It's expressly allowed by the Standard.[/color]

It is allowed to [partially] specialize class templates from the
standard library in namespace 'std' *IF* the specialization involves
a user defined type. 'std::pair<int, int>' as used in the original
article does not qualify for a user defined specialization!

That said, it is probably save to specialize things which don't exist
according to the standard - only there is no guarantee that it indeed
works...
--
<mailto:dietmar _kuehl@yahoo.co m> <http://www.dietmar-kuehl.de/>
Phaidros eaSE - Easy Software Engineering: <http://www.phaidros.co m/>

[Sharad Kala]

Re: specializing std::less


"Dietmar Kuehl" <dietmar_kuehl@ yahoo.com> wrote in message
[color=blue]
>
> Victor Bazarov wrote:[color=green]
> > "Exits Funnel" <exitsNOfunnelS PAM@yahoo.com> wrote...[color=darkred]
> > > [...] I can fix the problem by wrapping the specialization in a
> > > 'namespace std' block but I'm wondering if this is prudent or is[/color][/color]
> there[color=green][color=darkred]
> > > some other solution. Thanks.[/color]
> >
> > It's expressly allowed by the Standard.[/color]
>
> It is allowed to [partially] specialize class templates from the
> standard library in namespace 'std' *IF* the specialization involves
> a user defined type. 'std::pair<int, int>' as used in the original
> article does not qualify for a user defined specialization![/color]

True, but this is a well known work around that works on all the modern
compilers I have tested it on. But as you say it is not legal as quoted in
the Standard. Is there a proposal to change it (for std::pair) ?

Sharad

[Exits Funnel]

Re: specializing std::less

Sharad Kala wrote:[color=blue]
> "Dietmar Kuehl" <dietmar_kuehl@ yahoo.com> wrote in message
>
>[color=green]
>>Victor Bazarov wrote:
>>[color=darkred]
>>>"Exits Funnel" <exitsNOfunnelS PAM@yahoo.com> wrote...
>>>
>>>>[...] I can fix the problem by wrapping the specialization in a
>>>>'namespac e std' block but I'm wondering if this is prudent or is[/color]
>>
>>there
>>[color=darkred]
>>>>some other solution. Thanks.
>>>
>>>It's expressly allowed by the Standard.[/color]
>>
>>It is allowed to [partially] specialize class templates from the
>>standard library in namespace 'std' *IF* the specialization involves
>>a user defined type. 'std::pair<int, int>' as used in the original
>>article does not qualify for a user defined specialization![/color]
>
>
> True, but this is a well known work around that works on all the modern
> compilers I have tested it on. But as you say it is not legal as quoted in
> the Standard. Is there a proposal to change it (for std::pair) ?
>
> Sharad
>
>[/color]

Thanks Victor, Sharad, Dietmar. I appreciate the information.

-exits