[boost] Re: Type of Review for Unordered Associative Containers & Hash Functions

From: Daniel James (daniel_at_[hidden])
Date: 2005-02-24 04:52:21

Douglas Gregor wrote:

> Boost.Unordered should be a fast-track review. We're only looking at 
> code, tests, and docs: the interface is fixed or we're talking to the 
> wrong crowd.

This is a little pedantic, but the interface isn't entirely fixed. There 
are a couple of cases where it is affected by active issues. First, 
there's the behaviour of swap when allocators aren't equal, as described in:

http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2004/n1599.html

Secondly, there's the return type of erase(const_iterator) and 
erase(const_iterator, const_iterator). In the draft they return void but 
Thorsten asked if I could change them to return an iterator, as 
described in:

http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2005/n1753.html#130

This has a small performance hit with small nodes, as erase has to 
search for the next empty bucket to return the iterator. But seems like 
a good idea to me.

These are small points, but should be included in the review. (I'll add 
something to the documentation).

Daniel