#!/usr/bin/perl

# Terrible, horrible, no-good very bad Perl script for parsing flat-file HTML files into a format that
# can be imported by Movable Type.
#
# SYNTAX: ./jotsheet_conversion.pl [inputfile] > [outputfile]
#
# This script really chokes if there are DIV's within the "contentbox" 
# DIV's.  I tried to address this but failed. :)
#
# I wrote it, blame me.  --tom sherman [_bleach at yahoo dot com]
#
###
#
# This script deals with this kind of HTML from the old flat files.  There 
# are some variations on this structure, so the regex is a bit complex.
#
# <h3 class="jotsheet"><a name="2002-07-01"></a><a href="/content/jotsheet/archive/jotsheet_2002-07.shtml#2002-07-01" title="permlink for this jotsheet entry">01 jul 02</a> : title</h3>
#
# <div class="contentbox">
#
# [content]
#
# </div>


$dateAndTitleH3 = '<h3 class="jotsheet"><a name="(\d{4})-(\d{2})-(\d{2})"><\/a><a[^>]+>([^:]+):?\s*([^<]*)<\/h3>';
$dateAndTitleH4 = '<h4 class="jotsheet"><a name="([^"]*)"><\/a>\[?<a[^>]*>([^<]*)<\/a>\s?:?\s?([^\]]*)\]?\s*<\/h4>';
$mark = 0;

open(IN, "<$ARGV[0]");
while(<IN>)
{
        $line = $_;
        
# Primary entry [h3], replace with MT syntax
        if ($line =~ /$dateAndTitleH3/)
        {
                $year = $1;
                $month = $2;
                $day = $3;
                $entryDateInTitle = $4;
                $entryTitle = $5;
                if (!($entryTitle))
                {
                	$entryTitle = $entryDateInTitle;
                	$entryTitle =~ s/^([^>]+)<\/a>$/$1/;
                }
                $line =~ s/.*/-----\nKEYWORDS:\n\n-----\n\n\n--------\nAUTHOR: tom\nTITLE: $entryTitle\nSTATUS: Publish\nALLOW COMMENTS: 0\nCONVERT BREAKS: 0\nALLOW PINGS: 0\nPRIMARY CATEGORY: Old Jotsheet Entries\n\nDATE: $month\/$day\/$year 12:01:00 AM\n-----\nBODY:/;
        }

# Secondary entry [h4], properly format in HTML
        if ($line =~ /$dateAndTitleH4/)
        {
                $entryPermlink = $1;
                $entryNumWord = $2;
                $entryTitle = $3;
                if (!($entryTitle))
                {
                	$entryTitle = $entryNumWord;
                	$entryTitle =~ s/^([^>]+)<\/a>$/$1/;
                }
                $line =~ s/.*/<a name="$entryPermlink\"><\/a><h4 class="jotsheet"><a href="#$entryPermlink">$entryTitle\<\/a><\/h4>/;
        }        
        elsif ($line =~ /<!--#include virtual="\/ssi\/common\/bottom\.ssi"\s*-->/)
        {
                $line =~ s/.*/-----\nKEYWORDS:\n\n-----\n/;
        }
        elsif ($line =~ /<\/body><\/html>/)
        {
        	$line =~ s/.*//;
        }
        elsif ($line =~ /^<div class="contentbox">/)
        {
        	$line =~ s/.*//;
        	$mark = 1;
        }

# We need to allow non-"contentbox" DIVs
        elsif ($line =~ /^<div/)		
        {
        	$mark = $mark + 1;
        }
        elsif ($line =~ /^<\/div>/)
        {
        	if ($mark = 1)
        	{
        		$line =~ s/.*//;
        	}
        	else
        	{
        	       	$mark = $mark - 1;
        	}
        }
        print "$line";
}
close(IN);

exit;