On 5/5/11 2:31 AM, Brandon Moore wrote:
2011-5-4 8:53:50 PM, wren ng thornton On 5/1/11 7:58 PM, Edward Z. Yang wrote:
OK, to summarize the current discussion: - It would be nice to have a general-purpose priority queue in containers. I'm not interested in this goal per se, but I do view it as the cleanest way to get what I want.
With regards to this point, is there any reason why the following does not suffice (albeit not in base)?
http://code.google.com/p/priority-queues/
They're not priority *search* queues, but I'm not sure how much to read into your leaving out that word...
That word is pretty significant. In a priority queue every element also has a key, which can be used to efficiently remove it or adjust the priority.
That package wouldn't support cancelling a timeout, for example.
Yes, I know it's important. Which is why I brought up the question about whether the omission in ezyang's summary was intentional or an unfortunate oversight. If the goal is only to provide general-purpose priority queues, then I argue that we should be promoting that package. Algorithmically, they're asymptotically optimal; practically, they come with proofs of correctness; and, in practice, I've been quite happy with them. The only reason to promote something else is as a special case, if someone has a PQ which has better constant factors (for folks whose data is small enough not to care about asymptotics). However, if the goal in question is to provide general-purpose priority search queues, then clearly that project is irrelevant. -- Live well, ~wren