Container suggestions (was: Re: BOOST_NO_HASH)

From: deansturtevant_at_[hidden]
Date: 2001-01-07 00:11:24

Funny this topic should come up -- I was just thinking of posting a 
suggestion that rb-trees be made a part of boost, until I realized 
that the functionality I needed I could get from the standard set. 
Still, it would be more natural to work with an interface to the tree 
itself, since what I wanted was a data structure which efficiently 
supported the following operations on a container of elements with a 
comparison defined:
insert
minimum
maximum
remove_minimum
remove_maximum
empty
count

(I wanted to implement a 'priority deque', which is a double-ended 
priority queue. These are very useful in search algorithms where you 
want to bound the number of active hypotheses by tossing the least 
likely ones).

David Abrahams has provided a compelling reason for providing 
mechanisms for accessing internal tree structure.

But I don't understand why it's inappropriate to use something like 
the interfaces underlying standard set implementations. David?

As for the hash table issue -- I think it would be a good idea to 
provide some hash implementations. They can be very useful. Perhaps 
boost could provide a variety of them.

- Dean