#!/opt/bin/perl
##
## Spam Recognition System -- fully-functional prototype in perl
## (to be recoded with additional functionality in C or Objective-C)
##
## Based on algorithms described by Paul Graham in:
## http://www.paulgraham.com/spam.html
##
## $Id: check.pl,v 1.4 2002/08/22 02:53:07 dgc Exp $
## dgc@uchicago.edu
##

($pdir = $0) =~ s!/[^/]+$!!;
eval 'push(@INC, $pdir); require "lib.pl"';

$DEBUG = 0;

sub usage {
	print STDERR "usage: $0 [-v] file [...]\n";
}

sub main {
	my ($C, $R, $P, $N);
	my ($file, $pretty, @sel, $rating);
	my (@k, $n);
	my $verbose = undef;
	my @indicators = qw(-- == ++);
	my $Rprime;

	lockdb();
	($P, $N) = mapcontext();
	$R = maprate();

	if ($ARGV[0] eq "-v") {
		$verbose = 1;
		shift @ARGV;
	}

	@ARGV = qw(-) if ($#ARGV < 0);
        for $file (@ARGV) {
		$C = newcontext();
		unless (scanfiles($C, 0, $file)) {
			print "$0: can't open \"$file\": $!\n";
			next;
		}
		$Rprime = selectwords($C, $R);
		@sel = keys %{$Rprime};

		$rating = rateselection($Rprime, $P, $N, @sel);
		$pretty = $file;
		$pretty = "stdin" if ($file eq "-");
		print "$pretty: $rating\n";
		if ($verbose) {
			#print "\tBased upon:\n";
			map {
				printf("\t%s %3d x %0.3f %s\n",
					$indicators[($Rprime->{$_} <=> 0.5) + 1],
					$C->{stats}->{$_},
					$Rprime->{$_},
					$_);
			} sort {
				$Rprime->{$b} <=> $Rprime->{$a} ||
				$C->{stats}->{$b} <=> $C->{stats}->{$a} ||
				$a cmp $b;
			} @sel;
			print "\n";
			@k = keys %{$R};
			$n = $#k+1;
			print <<END_TEXT
	Positive records: $P->{nmsgs}
	Negative records: $N->{nmsgs}
	Tokens indexed: $n

	* "++" indicates tokens which suggest unsolicited bulk e-mail content.
	* "--" indicates tokens which suggest ordinary mail.
	* A rating of 0.500 indicates that a token counts equally for and
	  against a message. It suggests that you need more data in your
	  knowledge base.
END_TEXT
		}
	}

	unmapcontext($P, $N);
	unmaprate($R);
	unlockdb();
}

&main;