28 Feb
2005
28 Feb
'05
4:38 a.m.
Beman Dawes wrote:
In 1999 I did some work on a B-tree implementation for Boost, but became discouraged. For maximum portability and very high performance, it would have had to have a lot of limitations; memcopy'able POD types only, an interface that differed in several ways from the standard library containers, etc. Even with those limitations, the prototype took twice as long for typical operations as the old C code B-tree I've used for 20 years.
Perhaps someone else will give it a try; a B-tree is an incredibly valuable tool and it would be great to have a good one in Boost.
I thought Berkeley DB from Sleepy cat was good. Unless someone comes up with something that is better than that ... joe