|
Boost : |
From: Gottlob Frege (gottlobfrege_at_[hidden])
Date: 2007-08-11 01:23:39
On 8/8/07, Jeremy Bruestle <jeremy.bruestle_at_[hidden]> wrote:
> If anyone has more question though, or has ideas, or just wants to
> express interest, please do so. At this point I think I'm going to
> keep working on the project even if no one is interested, because I
> see of lot of fun problems to attack, but I'm still curious if anyone
> else has use for such a structure. Thanks to all so far for the good
> feedback.
>
> -Jeremy
Your original reason for the shared map seems to be the cost of
copying. If the map was based on a vector instead of a tree (and
didn't share nodes at all) do you think that might solve your slow
copy problems?
I worked on a project where we found that copying of maps was one of
the major bottle-necks, so we converted most of them to sorted
vectors, and that fixed it.
We didn't need the sorted vectors to have all the same behaviour as
maps, but I've had a plan for vector-based maps sitting on the
back-burner for a while now. Might be something to consider.
Tony
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk