Re: perl and lojban, sitting in a tree..

[David Brookshire Conner <nellardo@concentric.net>]

Robin Turner writes:
 > From: Robin Turner <robin@Bilkent.EDU.TR>

 > Why Perl, in particular?  I don't know much about languages, but
 > I would have thought Prolog would be more suitable for what you
 > have in mind.

Basic Prolog is based on predicate calculus, which would make one
think that it would be a good match to "la lojban."

I'm of a mind that Prolog would be decent for programming in
lojban. I.e., you define a Prolog predicate for every gismu and then
can define each gismu in terms of each other (possibly).  You type in
Prolog but use the well-defined lojban gismu rather than ambiguous
English names of predicates. But this approach doesn't let you handle
e.g., lujvo or cmene the way you'd like to. You'd have to define every 
lujvo explicitly. So what you really need is a good meta-object system 
or else a new programming language (same thing, really).

For parsing, I'm of the opinion that a stream-based approach in a lazy 
functional language is much easier to use (Haskell comes to
mind). What you want to parse becomes a list of characters, and you
can manipulate it like any list. In theory you don't worry about
reading in a character or that nonsense.  Here's some code that
outlines a parser for lojban at a high level. Note that this is pretty 
much legit Haskell:

data Bridi = (Selbri, [Sumti]). -- A sentence is a preda and its args

type Discourse = [ Bridi ] -- paragraph is a list of sentences

-- take a string and open it as a file to parse
lojbanParseFromFile :: String -> LojbanDiscourse
lojbanParseFromFile file = lojbanParse openFile file


-- parse a string into a series of lojban statements
lojbanParse :: String -> Discourse
lojbanParse stuff = 
   breakIntoBridi stuff >>=  -- Get back a list of strings that are
                             -- bridi
   map (\ oneBridi ->
          ( extractSelbri oneBridi,
            extractSumti  oneBridi)) -- for each bridi, pull out the
                                     -- selbri and sumti
                        >>=
    return -- and that's it

extractSelbri and extractSumti can be written for a simple case - they 
know that they are only getting a single bridi - no
fluff. extractSelbri will look for a brivla of some sort and work out
from there to stick on anything that applies (e.g., for tanru and
such).

       
Long-winded perhaps, but much more likely to be readable the second
time than perl (the write-only language :-)....

Speaking of long-winded, that's enough from me.....

co'o mi'e brukcayr.

---------
% sleep with me 
bad character

---------
Fancy. Myth. Magic.
http://www.concentric.net/~nellardo/