Re: [boost] hash table library proposal (was Extensions in Boost.Compatibility)

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-12-13 14:50:19

At 04:51 PM 12/13/2002, Toon Knapen wrote:

 >On Thursday 12 December 2002 20:06, Beman Dawes wrote:
 >> At 08:17 PM 12/11/2002, Jeremy Maitin-Shepard wrote:
 >>  >I forgot to mention in my previous post the following proposed ideas:
 >>  >
 >>  >A hash map and hash multimap in which a single value type is 
specified
 >>  >along with a traits class that is used to extract the key from the
 >>  >value type.  The std::pair-specific hash_map and hash_multimap would
 >>  >then be based upon this generic implementation.
 >>  >
 >>  >If there is interest in this library, I will create an 
implementation.
 >>
 >> Are you aware that a hash_map proposal from Matt Austern is almost sure 

 >to
 >> be included in the upcoming C++ Standard Library Technical Report?
 >>
 >> See http://std.dkuug.dk/jtc1/sc22/wg21/docs/papers/2002/n1399.html
 >>
 >> The only reason that proposal wasn't accepted in Santa Cruz was a 
naming
 >> controversy and some minor technical quibbles about tuning types being
 >> specified with more precision than can be delivered.
 >
 >Still it makes sense to have it already in boost (until it's implemented 
by
 >all compilers)

Yes, but I think we should try to stick closely to the LWG proposal. In 
fact, it might be an advantage for standardization to have an exact 
implementation. (It might be worth asking Matt Austern how he proposes to 
change the interface to deal with the issues still on the table.)

Something completely different might be OK, too, but it doesn't serve 
anyone's interest to do something "almost standard". That was the point of 
the question.

--Beman