# script for converting from ugly html files to Movable Type's import/export format
# it will need extensive modification to accomodate your individual entry template

# Written by Yami McMoots - http://greengabbro.net/
# ... and released to the public domain. I hope someone else finds it useful.

#Path to test file:

$testoutput = '/Users/maria/Sites/gabbro/googlecached/2006.txt';
$tmp = '/Users/maria/Sites/gabbro/googlecached/tmp.txt';
$ctmp = '/Users/maria/Sites/gabbro/googlecached/ctmp.txt';
$btmp = '/Users/maria/Sites/gabbro/googlecached/btmp.txt';

open(OUT, ">>$testoutput") || die "Error opening output: $!";

# Directory containing all the saved html files, and nothing else
opendir(DIR,"/Users/maria/Sites/gabbro/googlecached/2006") || die "Error reading directory $!";
@files = readdir(DIR);

foreach(@files) {

	$file = $_;

	open(DATA, "$file") || die "Error opening $file: $!";

	# Entry/comment cycle:
	# 0: Nothing
	# 1: Just found the beginning of an entry/comment body.
	# 2: Just finished reading the body. The next line is metadata.
	# 3: Just finished writing metadata. Time to write the body.
	$entry = 0;
	$comment = 0;

	$entrymeta = 0;

	# Are there comments on this entry?
	$ctog = 0;

	while(<DATA>) {	# read each line, one at a time, in turn.
	
		$L = $_;
	
		if( $L =~ m/<\/title>/ ) {	# Entry title
			$L =~ s/\t//g;	# Strip stupid tabs
			$L =~ s/<\/title>//;	# Get rid of the end title tag
		
			print OUT "TITLE: ";
			print OUT $L;
		}
	
		if( $L =~ m/filed under <a.*?>(.*?)<\/a>/ ) {	# Entry categories
			while( $L =~ m/<a.*?>(.*?)<\/a>/g ) {
				print OUT "CATEGORY: $1\n";
			}
				
			$entry = 3;
			$comment = 0;
		}
		
		if( $entry == 2 || $comment == 2) {	# Just finished an entry, time for metadata
			$match = 0;
			
			if( $entry == 2 && $L =~ /(\w+?) &middot; (\d\d?):(\d\d) &middot; (\d*) (\S{3}) (\d{4})/ ) {
				$author = $1;
				$hours = $2;
				$mins = $3;
				$day = $4;
				$month = $5;
				$year = $6;
							
				$entrymeta = 1;
				$match = 1;
				$entry = 0;
			}
			# Comment has an author URL
			elsif ( $L =~ m/<a href='/downloads/downloads/(.*?)'.*?>(.*?)<\/a> &middot; (\w*?) (\d{1,2})\w\w, (\d{4}) &middot; (\d\d):(\d\d)/ && $comment == 2 ) {
				$url = $1;
				$author = $2;
				$month = $3;
				$day = $4;
				$year = $5;
				$hours = $6;
				$mins = $7;

				$match = 1;
			}
			# Comment with no author URL
			elsif ( $L =~ /(.*?) &middot; (.*?) (\d{1,2}).., (\d{4}) &middot; (\d\d):(\d\d)/ && $comment == 2 ) {
				$author = $1;
				$month = $2;
				$day = $3;
				$year = $4;
				$hours = $5;
				$mins = $6;
				$url = '';

				$match = 1;
			}
			else {
				if( $entrymeta > 0 || $comment == 2 ) {
					print "Metadata regex failure on line $_";
				}
			}
		
			# Month lookup
			if( $match == 1 ) {
				if ( $month eq "Jan" || $month eq "January" ) {
					$month = "01";
				} elsif ( $month eq "Feb" || $month eq "February") {
					$month = "02";
				} elsif ( $month eq "Mar" || $month eq "March" ) {
					$month = "03";
				} elsif ( $month eq "Apr" || $month eq "April" ) {
					$month = "04";
				} elsif ( $month eq "May" || $month eq "May" ) {
					$month = "05";
				} elsif ( $month eq "Jun" || $month eq "June" ) {
					$month = "06";
				} elsif ( $month eq "Jul" || $month eq "July" ) {
					$month = "07";
				} elsif ( $month eq "Aug" || $month eq "August" ) {
					$month = "08";
				} elsif ( $month eq "Sep" || $month eq "September" ) {
					$month = "09";
				} elsif ( $month eq "Oct" || $month eq "October" ) {
					$month = "10";
				} elsif ( $month eq "Nov" || $month eq "November" ) {
					$month = "11";
				} elsif ( $month eq "Dec" || $month eq "December" ) {
					$month = "12";
				} else {
					print "Month regex failure: $month\n";
				}
								
				if ($day < 10) {
					$day = "0".$day;
				}
			
				if( $entrymeta == 1 ) {
					print OUT "AUTHOR: $author\n";
					$author = '';
					print OUT "DATE: $month/$day/$year $hours:$mins:00\n";
					$month = '';
					$day = '';
					$year = '';
					$hours = '';
					$mins = '';
				}
			
				if( $comment == 2) {
					open(CTMP, ">>$ctmp");
					
					print CTMP "COMMENT:\n";
					print CTMP "AUTHOR: $author\n";
					$author = '';
					print CTMP "DATE: $month/$day/$year $hours:$mins:00\n";
					$month = '';
					$day = '';
					$year = '';
					$hours = '';
					$mins = '';
					if ($url) {
						print CTMP "URL: $url\n";
					}
				
					close(CTMP);
				}
			}
			
			if( $comment == 2 ) { 
				$comment = 3; 
			}
			
			$match = 0;
			$entrymeta = 0;
		
		}	
			

		# Print the entry/comment body
		if( $entry == 3 ) {
			print OUT "-----\n";
			print OUT "BODY:\n";
			open(BTMP, "$btmp") || die "Error opening btmp $!";
				
			while(<BTMP>) {
				print OUT $_;
			}
			
			print OUT "-----\n";
			close(BTMP);
			open(BTMP, ">$btmp");
			print BTMP "";
			close(BTMP);
			
			$entry = 0;
			if( $ctog == 1 ) {
				open(CTMP, "$ctmp") || die "Error opening ctmp $!";
				while(<CTMP>) {
					print OUT $_;
				}
				
				close(CTMP);
				open(CTMP, ">$ctmp");
				print CTMP "";
				close(CTMP);
			}
		}
		# Comment metadata parsed: write comment body to ctmp
		elsif( $comment == 3 ) {
			open(TMP, "$tmp") || die "Error opening tmp $!";
			
			open(CTMP, ">>$ctmp") || die "Error opening ctmp $!";
			while(<TMP>) {
				print CTMP $_;
			}
			
			print CTMP "-----\n";
			close(TMP);
			close(CTMP);
			
			$comment = 0;
			
			open(TMP, ">$tmp") || die "Error opening tmp $!";
			print TMP "";
			close(TMP);
		}	

		if( $entry == 1 && $L =~ m/<\/div>/) { # Entry body end
			$entry = 2;
			close(BTMP);
		}
		if( $comment == 1 && $L =~ m/<div class="meta">/) { # Comment body end
			$comment = 2;
			close(TMP);
		}
		
		# If we're in the entry body, print the line to the scratch file:
		if( $entry == 1 && $L =~ /\S/ ) {
			$L =~ s/\t//g;
			print BTMP $L;
		}
		# If we're in a comment body, print to comment scratch:
		if( $comment == 1 && $L =~ /\S/ ) {
			$L =~ s/<\/?p>//g;
			$L =~ s/\t//g;			
	
			print TMP $L;
		}
	
		if( $L =~ m/entrytext/) {	# Entry body start
			print OUT "STATUS: Publish\n";
			open(BTMP, ">>$btmp") || die "Error opening btmp $!";
			$entry = 1;
		}
		if( $L =~ m/<li.*comment-.*>/) {	# Comment start
			open(TMP, ">>$tmp") || die "Error opening scratch $!";
			$comment = 1;
			if( $ctog == 0 ) { $ctog = 1; }
		}
		
	}

	print OUT "--------\n";

	close(DATA);
}
	
close(OUT);