From: Peter Dimov (pdimov_at_[hidden])
Date: 2002-09-06 08:25:19
From: "Gabriel Dos Reis" <gdr_at_[hidden]>
> "Peter Dimov" <pdimov_at_[hidden]> writes:
>
> [...]
>
> | Another, more practical, argument is that if you specialize std::less,
this
> | will work:
> |
> | std::set< interval > s;
> |
> | std::vector< interval > v;
> | std::sort(v.begin(), v.end(), less<interval>());
> |
> | but this (for example) will not:
> |
> | std::set< std::pair<interval, int> > s2;
>
> Yes, but the problem isn't with interval. Is it?
Yes, it is. std::pair is just an example. You can "fix" std::pair but you
can't fix every compound object in existence.
Once you decouple std::less from operator<, you now need to always define a
less<> specialization in terms of less<> whenever you would define an
operator< in terms of operator<.
This is not the original intent of the STL. The original intent is for
less<> to alias operator<, just like the other function objects alias the
other operators.
less<> was decoupled from operator< for pointers and it went downhill from
there. Fortunately pointers are totally ordered on most platforms that
matter, so this doesn't have much of an impact.