[34] in Coldmud discussion meeting

root meeting help first previous next next in chain last in chain last

parsing

daemon@ATHENA.MIT.EDU (Tue Nov 16 18:03:04 1993 )

From: stewarta@netcom.com (Alex Stewart)
To: coldstuff@MIT.EDU
Date: Tue, 16 Nov 93 14:49:28 PST

Well, my mail program appears to be dropping some of my mail into oblivion
instead of the folder I specify, so I'll have to respond to some things from
memory, it seems..

The example given of bogging down the system whenever someone gave a command in
the startup room seems to me to be far more an example of spectacularly bad
architecture than of verb-matching problems.  In my opinion, there should never
be so many objects in a single room that this becomes at all noticeable and if
there are, the room should be cleaned out anyway.  This, however, is not the
real issue anyway.

I actually started modifying the parsing to elimiate what seemed to me to be an
incredibly inefficient way of dealing with remotely-referenced verbs (which I
still think to be the case).  In the process I ended up doing away with the
verb cache as well, in the interest of better encapsulation and a more flexible
system.  I'm not really adverse to the notion of a verb cache as such, but
looking over the system more closely, I really wonder whether the benefits
would outweigh the quite convoluted implementation necessary.  It seems to me
that in most cases (including the scenario of the startup room), it would have
very little, if any, positive affect the way it is now.  I'm going to look into
it further, however.  I may change my mind.

I shouldn't have mentioned I was working on it..  Inevitably whenever I mention
I'm working on something, something more important (usually RL) comes up and
drags me away from it.  Hopefully I'll be back at The ColdDark within a couple
of days to finish things up.

It also appears I invited people around without telling them where to go.
Oops.  paradise.declab.usu.edu 6666.  Due to local issues, it's generally only
up during the nighttime (US time), and sometimes in the evenings.  This
will change when we get the impending new computer to run it on.

Anyway, for those interested, the parsing structure I'm currently working on is
along the following lines:

  $player.parse_command does the normal match_command checking, then gets a
list of all the objects in the player and room, and calls obj.match_verb(str)
on each.

  obj.match_verb checks the passed string to determine whether it fits an
appropriate verb by asking each of its ancestors that define verbs to match it,
and returning any results in the form
[[obj,template,match_result,method],...]  In the process, it also checks to see
whether it's in the local environment of the sender, and disregards any local
verbs being called remotely.

  There is also a simple form of cacheing I'm going to implement on the verb
handling objects, which will remember the last passed command string and what
the result was from it and all of its ancestors, thus a good part of the
overhead on checking all of the objects in the room is reduced, since no matter
how many descendants of $foo there are, $foo's ancestors would only get
asked to match the command once.

  Remote matching of verbs is handled simply by searching the command string
for any explicit dbrefs and including them in the search list.

As I said, the my coded stuff on The ColdDark isn't as far along as I expected
it to be, so you probably won't get too much more out of looking at it right
now than what I just listed here.  I'll post when I get something more
complete down.

On the subject of regexps:  I also tend to lean towards not including regexp
matching simply because it's been my experience that people who understand them
tend to lean far too heavily on them, and they're often not the best tool for
the job (except for quite complicated tasks, which can often be broken up into
a set of simpler tasks, leading to increased readability and more easily
modifyable systems).  MUDs, after all, are not text editors or command-line
search programs, and do have their own programming languages which work quite
nicely at many of the more sophisticated tasks regexps were designed for.

-Richelieu
--