opencog-dev team mailing list archive

[Gustavo Machado Campagnani Gama <gama@xxxxxxxxxxxxx>]

Hi Joel,

I had a quick chat with Senna this afternoon about this. We agreed that
we'll eventually migrate the TLB permanently to use some data structure
that can validate the handles [side note: we both agreed that a map
(rbtree) might not be best option, but this is something to be discussed
some other time.]

But for the time being, we think it's important to keep the "cast-based"
version because it serves as a base-point. For instance, once we have a
proper benchmark, it'll probably be useful to have something to compare
to when we're trying to quantify the overhead of a given data structure.

And if your code depends on the map, that's fine. All we want is to keep
the cast-based implementation for the time being.

--
Gustavo


On Thu, 2008-06-26 at 15:06 +1200, Joel Pitt wrote:
> Hi all,
> 
> Is everyone okay if we remove the TLB support for not using std::map?
> 
> There isn't anyway to check if a Handle is valid or not unless
> std::map is being used, which I (just realized I) need for the
> ForgettingAgent.
> 
> Well, I could hackily work around this, but surely the TLB should
> eventually have a way to say "that handle doesn't exist", along with
> "that handle is on disk/in memory" (in future)?
> 
> J