[Coldstuff] Environment list and recursive locations...

Jonathan Robertson coldstuff@cold.org
Sun, 17 Feb 2002 21:36:05 -0800 (PST)


The performance hit isn't enough to slow anything down
significantly I don't think either, I'd like to see
the second one. :D
Jonathan
--- Brandon Gillespie <brandon@roguetrader.com> wrote:
> Ok, I decided to tackle fixing some of the located
> location issues
> (i.e. a location of a couch in a location of a room,
> etc).  I thought
> about this years ago (1995 or so) and cannot recall
> ANYTHING from back
> then, so I'm starting over.  I'd appreciate some
> input on how other
> muds handle this, because frankly I can't even
> recall how MOO does it
> anymore...
> 
> I have three variants, based on a little hacking I
> just did...  All of
> the timings come from the same room on the same
> hardware, and are
> relative only to themselves...
> 
> 1-- Current ColdCore configuration...
> 
>    This configuration is very minimal... the
> environment has no concept
>    of located locations.  It is just a list of
> anything you are holding
>    or which is in the room you are in (so you could
> address the couch in
>    the room you are in, but nobody on the couch in
> the room...).
> 
>    This takes approximately 0.000086 to generate an
> environment list...
> 
> 2-- Open-ended configuration
> 
>    This allows for any object within the context of
> the environment to
>    add-to the list as it is being generated.  This
> way a couch could also
>    inject everybody sitting on the couch, etc.  It
> uses a dictionary to
>    avoid loops and replication.  However, it is also
> very slow, time is
>    approximately 0.001234 in the same room.
> 
> 3-- $located_locations only
> 
>    This goes back to the original code and just
> hacks in a check for
>    located locations, and if the target object is a
> located location it
>    then injects that object's contens.  It takes
> 0.000318 seconds.
> 
> So the issue is--stick with a non extensible
> mechanism like #3, but
> which is faster, or go with an extensible mechanism
> like #2 and just
> take the performance bump (is it really that big of
> a deal?).  Nice
> thing about #2 is it allows for things like an
> object to be able to
> inject a detail into the environment, without having
> to either make it
> be a real object or have it be a located_location. 
> Of course, one
> could argue against that, in that all addressable
> objects in the
> environment should be $things, and all located
> locations should be
> $located_locations, thus making it a non-issue...
> 
> Let me know which would be preferrable...
> 
> -Brandon
> 
> _______________________________________________
> Cold-Coldstuff mailing list
> Cold-Coldstuff@cold.org
> http://web.cold.org/mailman/listinfo/cold-coldstuff


__________________________________________________
Do You Yahoo!?
Yahoo! Sports - Coverage of the 2002 Olympic Games
http://sports.yahoo.com