#!/usr/bin/perl -w
#---------------------------------#
#  PROGRAM: scholia_formatter.pl  #
#---------------------------------#

# Open the file supplied by the first argument
open (INFILE, $ARGV[0]) || die "Error opening input file";
@contents = <INFILE>;

$outputFile = "new_".$ARGV[0];
open (OUTFILE, "> $outputFile");

$print_begin = 0;
$dashed_list_found = 0;
$footnote = 0;
$inside_a_list = 0;
$header_two_lines = 0;
$new_footnote_paragraph = 0;

foreach $line (@contents) {

	# break at end of content				
	if ($line =~ /<!-- print end -->/) {
		last;
	}

	# TITLE
	if ($line =~ /class="art_head/) {
		$line =~ s{<p class="art_head">}{};
		$line =~ s{</p>}{};
		print OUTFILE $line;		
		next;
	}

	# BODY
	if ($line =~ /<!-- print begin -->/) {			# start printing each line only once the print actually starts
		$print_begin = 1;
		next;
	}

	if ($print_begin == 1) {
#		$line =~ s/^\s+//;		# trim whitespace at beginning of line

		# Headers				# replaces headers contained in a single line
		$line =~ s{<p class="art_sub1"(.*)>(.+)</p>}{<h2 class="art_sub1"$1>$2</h2>};	
		$line =~ s{<p class="art_sub2"(.*)>(.+)</p>}{<h3 class="art_sub2"$1>$2</h3>};						
		$line =~ s{<p class="art_sub3"(.*)>(.+)</p>}{<h4 class="art_sub3"$1>$2</h4>};						
		$line =~ s{<p class="art_sub4"(.*)>(.+)</p>}{<h4 class="art_sub4"$1>$2</h4>};						

		if ($header_two_lines ge 1 && $line =~ '</p>') {			# replaces the closing header tag on a later line
			$line =~ s#</p>#</h$header_two_lines>#;
			$header_two_lines = 0;
		}
		
		if ($line =~ 'p class="art_sub1') {							# replaces headers spanning multiple lines
			$line =~ s{<p class="art_sub1"(.*)>}{<h2 class="art_sub1"$1>};
			$header_two_lines = 2;
		}
		if ($line =~ 'p class="art_sub2') {
			$line =~ s{<p class="art_sub2"(.*)>}{<h3 class="art_sub2"$1>};
			$header_two_lines = 3;
		}
		if ($line =~ 'p class="art_sub3') {
			$line =~ s{<p class="art_sub3"(.*)>}{<h4 class="art_sub3"$1>};
			$header_two_lines = 4;
		}
		if ($line =~ 'p class="art_sub4') {
			$line =~ s{<p class="art_sub4"(.*)>}{<h4 class="art_sub4"$1>};
			$header_two_lines = 4;
		}

		# Content
		$line =~ s{<p class="leader">}{<p>};
		$line =~ s{<p class="content">}{<p>};
		$line =~ s{<p(.*) align="center"(.*)>}{<p$1 class="centered"$2>};		# center
		$line =~ s{<p(.*) align="right"(.*)>}{<p$1 class="right"$2>};			# right
		$line =~ s{font face="Symbol"}{span class="greek"}g;		# this line shouldn't be necessary, really...
		$line =~ s{<font(.*) face="Symbol"(.*)>}{<span$1 class="greek"$2>};		# Symbol font
		$line =~ s{</font>}{</span>};
		$line =~ s{<blockquote> ?<p>(.+)</p> ?</blockquote>}{<blockquote>$1</blockquote>};		### I haven't seen this occur yet...

		$line =~ s{centered" class="art_sub2"><i>Thesis (..?)</i>}{centered"><span class="art_sub3">Thesis $1</span>};	# Thesis statements
		$line =~ s{<i class="art_sub2">Thesis (..?)</i>}{<span class="art_sub3">Thesis $1</span>};
		$line =~ s{centered" class="art_sub2"><b>(.*)</b>}{centered"><span class="art_sub2">$1</span>};
#<p class="The First Category of Divine Laws

		$line =~ s{<sup>(\d+)</sup>}{<sup><a href="#ftnt_$1" name="ftnt_anchr_$1" title="Footnote #$1">$1</a></sup>}g;

		$line =~ s{^<ul>$}{<ul class="dashed">};			# dashify <ul> lists

# 		$line =~ s/\x{201C}/&#8220;/;				# open double quote
# 		$line =~ s/\x{201D}/&#8221;/;				# close double quote
# 		$line =~ s/\x{2018}/&#8216;/;				# open single quote
# 		$line =~ s/\x{2019}/&#8217;/;				# close single quote
# 		$line =~ s/\x{2026}/&#8230;/;				# ellipsis
# 		$line =~ s/\x{2013}/&#8211;/;				# en dash
# 		$line =~ s/\x{2014}/&#8212;/;				# em dash
		
		# Lists
		if ($line =~ '<p>- (.+)</p>') {
			$line =~ s{<p>- (.+)</p>}{<li>$1</li>};
			$dashed_list_found = 1;
		}
		
		#FOOTNOTES
#		if ($line =~ '<h2 class="art_sub1">Notes</h2>' && $footnote == 0) {		# some scholias don't have this line
		if ($line =~ '<ol>$' && $footnote == 0) {
			$footnote = 1;
		}
		
		if ($footnote ge 1) {
			$line =~ s{^\s*<ol>$}{};					# remove <ol>
			$line =~ s{^\s*</ol>$}{};					# remove </ol>
			if ($line =~ m{^\s*<ol start="(\d*)">$}) {	# remove <ol start="#">, but remember the number
				$footnote = $1;
				next;
			}
			if ($line =~ m{^    (    )?<li>\d*}) {
				$line =~ s{^    (    )?<li>\d*\.? ?}{<p class="references"><a class="footnote_anchr" href="#ftnt_anchr_$footnote" name="ftnt_$footnote" title="Footnote #$footnote">$footnote</a> };
				$footnote++;
				$inside_a_list = 0;				# to account for any footnotes not closed with </li>
				$new_footnote_paragraph = 0;	# to account for any multi-paragraphed footnotes with an odd number of <br> tags
			}
			if ($line =~ '<li>') {				# don't replace list item tags for lists inside footnotes
				$inside_a_list = 1;
			}
			if ($inside_a_list == 0) {
				$line =~ s{</li>}{</p>};			
			}
			if ($line =~ '</li>') {
				$inside_a_list = 0;
			}
			
			if ($line =~ '<br>' && $new_footnote_paragraph == 0) {				# multiple paragraphs in a footnote are enclosed in <p> tags
				$line =~ s{<br>}{</p>};											# assumes multiple paragraphs are denoted with two <br> tags
				$new_footnote_paragraph = 2;
			}
			if ($line =~ '<br>' && $new_footnote_paragraph == 2) {
				$line =~ s{<br>}{<p class="references$new_footnote_paragraph">};
				$new_footnote_paragraph = 0;
			}
		}
		print OUTFILE $line;
	}
}

if ($dashed_list_found == 1) {
	print OUTFILE "\ndashed list FOUND!\n";
}

close INFILE;
close OUTFILE;