I get to be amazed I understand as much as I do. I declare you an honorary masochist by virtue of coding Perl.

I'm still working through it, particularly trying to understand what it is supposed to look like and behave like, and noticed something which _might_ be of service:

If I understand it correctly, the _add_level method currently overwrites facts with the same first element. I imagine I'm trouncing all over established terminology, so let me illustrate instead.

You have the two facts:

$pg->add_fact( gnarfle => qw/ foo bar baz / );
$pg->add_fact( gnarfle => qw/ foo tac toe / );
[download]

           {
            '2' => [
                     undef,
                     {
                       '3' => 3
                     },
                     {
                       '4' => 4
                     }
                   ]
          }
[download]

But after the foo bar baz rule begins, it looks like a terribly disappointing

          {
            '2' => []
          }
[download]

which is then fleshed out with the 5 and 6 entries. After it's done, instead of {foo => [undef,tac,toe]}, it's {foo => [undef,bar,baz]}

The end result is that you lose part of the tree, and no longer get full results from doing rule requests.

You've probably already noticed this one, and I don't yet have enough clue to fix it correctly (I have a feeling it'll require a data structure tweak, and I don't know enough about the module to do it yet). If this was confusing anyone else, I hope I've explained what was happening.

I just learned about Data::Dumper::Deepcopy - useful when Dumper decides that all undefs are actually a reference to the same element. That was confusing, too.

Thanks for the puzzle, Ovid!

--Ferret

Darn.

I just added the following line to the end of facts.t:

    ok( $pg->gnarfle( qw/ foo bar baz /), '... this should still match
+' );
    ok( $pg->gnarfle( qw/ foo tac toe /), '... as should this.' );
[download]

The first one fails (as you said it would). I updated the number of tests and, for a lark, ran this against an older copy. Still fails. Bummer. Working on a fix now. Wish I knew how I missed that one.

Update: Quick fix, fortunately. I needed to test to see if I had an array ref there to ensure that I added to it rather than overwriting it. You can download the new code (with updated tests) here.

Cheers,
Ovid

Join the Perlmonks Setiathome Group or just click on the the link and check out our stats.

        father('arhuman','disciple')

        father('disciple','arhuman')

        color('red','green')

        color('green','red')

• There's a rudimentary unification algorithm which (seems to) work(s).
  
• I've tried to use simpler structures (that I could understand ;-)
  
• It handles also "order independent" predicates
  (when first argument is a '?')
  
• Included a dirty 'load_from_file' method
  which would allow to load facts easily regardless of the implementation
  (that's why I ruled serialization out))
  to enable use of both Perlog and Perlog2 on the same facts database...
  

#!/usr/bin/perl -w
use strict;
#use Devel::StealthDebug;

use lib '.';

use Perlog;
use Perlog2;

#my $pg = AI::Perlog->new;
my $pg2 = AI::Perlog2->new;


#$pg->load_from_file('database.txt');
$pg2->load_from_file('database.txt');

#
# Perlog2 should definitly export display ;-)
#
AI::Perlog2::display($pg2->theory(qw/$x tac $y/));
print "*********\n";
AI::Perlog2::display($pg2->theory(qw/?  $x tac $y/));
print "*********\n";
AI::Perlog2::display($pg2->theory(qw/tic tac toe/));
print "*********\n";
AI::Perlog2::display($pg2->theory(qw/tIc tac toe/));
*********try.pl**********<br>
[download]

In the spirit of TMTOWTDI people might might find "Perl and Prolog and Continuations... oh my!" of interest.

People might also find this next bit of info interesting.... but they also might find it that it spoils the fun of this thread (sorry Ovid :-). Hence the uuencoding. You have been warned.

print unpack "u", <<'EOT';
M5&AE<F4@:7,@82!-871H.CI,;V=I8SHZ4')E9&EC871E(&]N($-004X@=&AA
M="!D;V5S(&UO<W0@;V8@=VAA="!I="!L;V]K<R!L:6ME('EO=2!W86YT($%)
2.CI097)L;V<@=&\@9&\@.BTI
EOT
[download]