#!/usr/local/bin/perl5.8.0
# WebGenDB Utility
# This script replace all the < with entities for the tags that are text within a batchLoad tag
# e.g. <Title>First draft of <title>JavaScript</title></Title>
# becomes <Title>First draft of &lt;title&gt;JavaScript&lt;/title&gt;</Title>
use Getopt::Long;

if ($#ARGV <1) {
    print "usage: replaceTags.pl -f <inputFile> -o <outputFile>\n";
    exit;
}

GetOptions ('f=s' => \$input,
	    'o=s' => \$output,
	    'h'   => \$help);

if (defined $help) {
    print "usage: replaceTags.pl -f <inputFile> -o <outputFile>\n";
    exit;
}

die "usage: replaceTags.pl -f <inputFile> -o <outputFile>\n" unless (defined $input && defined $output);

# Read all to the end of the file
$/ = EOF;
open(INPUT, "$input");
$inFile = <INPUT>;

# Assumption: all the EAD tags are always all in small caps
#             all BatchLoad tags either start or contain a capitalized letter

# For all the <tag> which only contain small caps, replace the brakets with entities
# need to account for <tag> </tag> <tag/>
# other to keep into account: <title type="simple">,  <note show="new" id="fn1">
# <archref show="embed" actuate="onload" href="http://www.oac.cdlib.org/findaid/ark:/13030/kt200014h4">
# <note label="General Note">
$inFile =~ s/<(\/?[a-z=_"\s\d\@:.\/]*\/?)>/&lt;$1&gt;/g;

# Fix <note label="General Note">
$inFile =~ s/<([a-z]+ [a-zA-Z:="\s]*)>/&lt;$1&gt;/g;

# get rid of all the \n which have been introduced from the EAD
$inFile =~ s/\n/ /g;

# collapse multiple spaces to one
$inFile =~ s/\s+/ /g;

# put a \n after every </.*> and after &lt;/abc&gt; tag for readibility 
# and because of Exception of String too long
$inFile =~ s/(<\/.*?>)/$1\n/g;
# This commented out because the EAD doesn't need to have extra \n in the content
#$inFile =~ s/(&lt;\/.*?&gt;)/$1\n/g;

open (OUTPUT, ">$output");
print OUTPUT $inFile;