Re: associative_vector: request for formal review

From: Dave Abrahams (abrahams_at_[hidden])
Date: 2000-05-30 14:52:52

--- In boost_at_[hidden], jsiek_at_l... wrote:
> Hi Dave,
> 
> A quick first glance at the code brought up 2 issues for me:
> 
> 1. It seems to me that if someone provides their own allocator there
> is going to be a compiler error, since the Alloc parameter is not 
used
> in the typedef for Impl. Or am I missing something?

Yikes! Good catch. Serves me right for not testing allocator 
parametrization!

> 2. I've got the exact same code sitting in GGCL. How about we
>   put tied in utility.hpp?
> 
> template <class X, class Y>
> struct tied
> {
>     tied(X& x, Y& y) : m_x(x), m_y(y) {}
>     
>     template <class T1, class T2>
>     void operator=(const std::pair<T1,T2>& p)
>         { m_x = p.first; m_y = p.second; }
>     mutable X& m_x;
>     mutable Y& m_y;
> };
> 
> template <class X, class Y>
> tied<X, Y> tie(X& x, Y& y) { return tied<X,Y>(x,y); }

Well, I think that code is just detritus; I was using it but ended up 
refactoring the code and all the uses went away, so I should just 
take it out of the test. OTOH I agree that this idiom is a great idea 
for utility.hpp.
-Dave