#!/usr/bin/perl -w
# sample_bayesian.pl
# Sample Bayesian classifier to filter spam and nonspam messages 
# (c) 2007 Srilatha Attaluri, Sean Gilpin

use strict;

our %spam_dict;
our %nspam_dict;
our $prior_prob_spam;


sub countFrequency
{
	my $choice = shift(@_);
	my (@emailWords) = (@_);
	foreach(@emailWords)
	{
		if($_ ne "")
		{
			if($choice == 1)
			{
				if (exists($spam_dict{$_})) {$spam_dict{$_}++;}
				else { $spam_dict{$_} = 1 ; }
			}
			else
			{
				if (exists($nspam_dict{$_})) {$nspam_dict{$_}++;}
				else { $nspam_dict{$_} = 1 ; }
			}
		}
	}
	return;
}

sub readFile
{
	my $file = $_[0];
	open(IN, $file) or return (); # if the file doesn't open return ""
	chomp(my @lines = <IN>);
	close(IN);
	my @words = ();
	foreach(@lines)
	{
		(@words) = (@words,split(/ /,$_));
	}
	return (@words);
}

sub train
{
	my %hash = ();
	my $index = 0;
	my @total = (0,0);
	#TODO:ignore the top 10

	#if content is zero we add 1 to each of the contents
	for($index = 0; $index < 2; $index++)
	{
		my $containsZero = 0;
		if($index == 0) {%hash = %spam_dict;} else {%hash = %nspam_dict;}
		SKIP:
		{
			while(my($k,$v) = each(%hash))
			{
				if($v == 0) {$containsZero = 1; last SKIP;}
			}
		}
		if($index == 0) {%hash = %spam_dict;} else {%hash = %nspam_dict;}
		if($containsZero) #if there is atleast one zero add 1 to all elements in hash
		{
			while(my($k,$v) = each(%hash))
			{
				$hash{$k} = $hash{$k} + 1;
				$total[$index]+= $hash{$k};
			}
		}
		if($index == 0) {%spam_dict = %hash;}
		elsif($index == 1) {%nspam_dict = %hash;}
	}

	#calculate the probabilities
	for($index = 0; $index < 2; $index++)
	{
		if($index == 0) {%hash = %spam_dict;} else {%hash = %nspam_dict;}
		while(my($k,$v) = each(%hash))
		{
			$hash{$k} = sprintf("%.10f",$v/$total[$index]);
		}
		if($index == 0) {%spam_dict = %hash;} else {%nspam_dict = %hash;}
	}

	return @total;
}

sub classify
{
	my @message_words = @_;

	my $crit_value =  log($prior_prob_spam) - log(1 - $prior_prob_spam);

	foreach my $word (@message_words)
	{
		if(exists($spam_dict{$word}) and exists($nspam_dict{$word}))
		{
			$crit_value = $crit_value + (log ($spam_dict{$word}) - log($nspam_dict{$word}));
		}
	}
	if($crit_value > 0)
	{
		return "spam";
	}
	else
	{
		return "nonspam";
	}
}

sub classifyFile
{
	my $fileName = shift(@_);
	my @wordsUnknown = readFile($fileName);

	#change all the words into lowercase
	@wordsUnknown  = map(lc($_),@wordsUnknown);

	return "" if scalar(@wordsUnknown) == 0;
	#filter special symbols and digits
	foreach(@wordsUnknown)
	{
		$_ =~ s/(http:\/\/)?(https:\/\/)?(www(\.|-))?(([a-z]|\d)+\d*\.)+([a-z]{1,3})*(\/([a-z]*\d*(\.|_|-|\?|=|&|@)*)*)*//g;#filter web sites
		$_ =~ s/#|\d|\.|,|!|:|-|\'|'|;|\?|\"|\*|[(]|[)]|{|}|\[|\]|_|\^|\$|`|’|\\|®|<|>|\t*|“|\+|\/|~|\!|%|@|=|`|&|™//g;
	}

	return classify(@wordsUnknown);
}



sub singleFileMode()
{
        print "\nEnter file name to classified\n: ";
	my $fileName = <STDIN>;
	chomp $fileName;

	my $classification = classifyFile($fileName);
	if($classification eq "")
	{
		print "Bad file name.\n\n";
	}
	else
	{
		print "\n", $fileName, " ", $classification, "\n\n";
	}
}

sub batchMode
{
	# Read all files in current directory with file name unknownXX.txt where XX is two or more digits
	my @unknownFiles = <./unknown??.txt>;

	print "\n";
	foreach my $fileName (@unknownFiles)
	{
		print $fileName, " ", classifyFile($fileName), "\n";
	}
	print "\n";
}

sub initialize
{
	my @spamFileNames = <./spam/*.txt>;
	my $spamFile = "";

	foreach my $filename(@spamFileNames)
	{
		$spamFile = $filename;
		my @wordsSpam = readFile($spamFile);

		#change all the words into lowercase
		@wordsSpam = map(lc($_),@wordsSpam);
		#filter special symbols and digits
		foreach(@wordsSpam)
		{
			$_ =~ s/(http:\/\/)?(https:\/\/)?(www(\.|-))?(([a-z]|\d)+\d*\.)+([a-z]{1,3})*(\/([a-z]*\d*(\.|_|-|\?|=|&|@)*)*)*//g;#filter web sites
			$_ =~ s/#|\d|\.|,|!|:|-|\'|'|;|\?|\"|\*|[(]|[)]|{|}|\[|\]|_|\^|\$|`|’|\\|®|<|>|\t*|“|\+|\/|~|\!|%|@|=|`|&|™//g;
		}

		#count the frequencies
		countFrequency(1,@wordsSpam); #counts the freq of spam
	}

	my @nspamFileNames =  <./nonspam/*.txt>;
	my $nonSpamFile = "";
	foreach my $filename(@nspamFileNames)
	{
		$nonSpamFile = $filename;
		my @wordsNSpam = readFile($nonSpamFile);
		#change all the words into lowercase
		@wordsNSpam = map(lc($_),@wordsNSpam);
		#filter special symbols and digits
		foreach(@wordsNSpam)
		{
			$_ =~ s/(http:\/\/)?(https:\/\/)?(www(\.|-))?(([a-z]|\d)+\d*\.)+([a-z]{1,3})*(\/([a-z]*\d*(\.|_|-|\?|=|&|@)*)*)*//g;#filter web sites
			$_ =~ s/#|\d|\.|,|!|:|-|\'|'|;|\?|\"|\*|[(]|[)]|{|}|\[|\]|_|\^|\$|`|’|\\|®|<|>|\t*|“|\+|\/|~|\!|%|@|=|`|&|™//g;
		}

		#count the frequencies
		countFrequency(2,@wordsNSpam); #counts the freq of non-spam
         }

	#delete nulls from the hashes
	while(my($k,$v) = each(%spam_dict))
	{
		$k =~ s/(\\s)*//g;
		chomp($spam_dict{$k});
		delete $spam_dict{$k} if ($k eq "");
	}
	while(my($k,$v) = each(%nspam_dict))
	{
		$k =~ s/(\\s)*//g;
		chomp($k);
		delete $nspam_dict{$k} if ($k eq "");
	}

	calculateSpamPriorProb(); # Do this before merging arrays

	#merge the arrays
	while(my($k,$v) = each(%spam_dict))
	{
		$nspam_dict{$k} = 0 if (not exists($nspam_dict{$k}));
	}
	while(my($k,$v) = each(%nspam_dict))
	{
		$spam_dict{$k} = 0 if (not exists($spam_dict{$k}));
	}

	#train the model
	my($total_spam,$total_nspam) = train(); #trains the model with corresponding probabilities
}

sub calculateSpamPriorProb
{
	#Make sure to call this before merging arrays
	#calculate prior probabilties
	my $spam_wordcount = 0;
	foreach( values(%spam_dict))
	{
		$spam_wordcount += $_;
	}
	my $nspam_wordcount = 0;
	foreach( values(%nspam_dict))
	{
		$nspam_wordcount += $_;
	}

	#die "Spam dictionary or nonspam is empty\n" if(($spam_wordcount == 0) | ($nspam_wordcount == 0));

	#Set global variable
	$prior_prob_spam = $spam_wordcount/($spam_wordcount + $nspam_wordcount);
}

MAIN:
{
	initialize();

	MENU:
	{
		print "****************************************\n";
		print "*                                      *\n";
		print "* Bayesian Classifier Menu             *\n";
		print "*                                      *\n";
		print "* Press 1 for single file mode         *\n";
		print "* Press 2 for batch classify mode      *\n";
		print "* Press q to quit                      *\n";
		print "*                                      *\n";
		print "****************************************\n";
		print "\n: ";

		my $choice = <STDIN>;
		chomp($choice);

		if($choice eq "1")
		{
			singleFileMode();
		}
		elsif($choice eq "2")
		{
			batchMode();
		}
		elsif(lc($choice) eq "q")
		{
			last MENU;
		}
		else
		{
			print "Please enter a valid option from menu.  Press <Enter> to continue\n";
			<STDIN>;
		}
		redo MENU;
	}

	exit;
}