#!/usr/bin/perl

# mods2kinosearch.pl - indexs a mods file with kinosearch

# Eric Lease Morgan <emorgan@nd.edu>
# 2007-07-14 - added namespace declaration; seems inelegant but functional
# 2007-06-24 - first cut; based on crri


# define
use constant INDEX   => '../etc/index';

# require
use strict;
use XML::LibXML;
use KinoSearch::InvIndexer;
use KinoSearch::Analysis::PolyAnalyzer;

# get the input
my $input = shift;
if ( ! $input ) {

	print "Usage: $0 filename\n";
	exit;
	
}

# initialize the index
my $analyzer = KinoSearch::Analysis::PolyAnalyzer->new( language => 'en' );
my $invindexer = KinoSearch::InvIndexer->new(
	invindex => INDEX,
	create   => 1,
	analyzer => $analyzer
);
$invindexer->spec_field( name => 'collection' );
$invindexer->spec_field( name => 'coverage' );
$invindexer->spec_field( name => 'creator' );
$invindexer->spec_field( name => 'date' );
$invindexer->spec_field( name => 'description' );
$invindexer->spec_field( name => 'format' );
$invindexer->spec_field( name => 'identifier' );
$invindexer->spec_field( name => 'language' );
$invindexer->spec_field( name => 'publisher' );
$invindexer->spec_field( name => 'relation' );
$invindexer->spec_field( name => 'rights' );
$invindexer->spec_field( name => 'subject' );
$invindexer->spec_field( name => 'title' );
$invindexer->spec_field( name => 'type' );

# initialize/configure
my $index = 0;
my $node;
binmode( STDOUT, ':utf8' );

# parse and register the mods namespace
my $parser = XML::LibXML->new;
my $collection = $parser->parse_file( $input );
$collection = XML::LibXML::XPathContext->new( $collection );
$collection->registerNs( 'mods', 'http://www.loc.gov/mods/v3' );

# loop through each mods element
foreach my $mods ( $collection->findnodes( '//mods:mods' )) {

	# register the namespace, again!
	$mods = XML::LibXML::XPathContext->new( $mods );
	$mods->registerNs( 'mods', 'http://www.loc.gov/mods/v3' );

	# increment
	$index++;
	
	# map mods to dc; see http://www.loc.gov/standards/mods/mods-dcsimple.html
	my $titles = '';
	foreach $node ( $mods->findnodes( './/mods:titleInfo/mods:title' )) { $titles .= &normalize( $node->textContent ) . '|' }
	
	my $creators = '';
	foreach $node ( $mods->findnodes( './/mods:name/mods:namePart' )) { $creators .= &normalize( $node->textContent ) . '|' }

	my $subjects = '';
	foreach $node ( $mods->findnodes( './/mods:subject/mods:topic' ))      { $subjects .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:subject/mods:name' ))       { $subjects .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:subject/mods:occupation' )) { $subjects .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:classification' ))     { $subjects .= &normalize( $node->textContent ) . '|' }

	my $descriptions = '';
	foreach $node ( $mods->findnodes( './/mods:abstract' ))        { $descriptions .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:note' ))            { $descriptions .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:tableOfContents' )) { $descriptions .= &normalize( $node->textContent ) . '|' }
	
	my $publishers = '';
	foreach $node ( $mods->findnodes( './/mods:originInfo/mods:publisher' )) { $publishers .= &normalize( $node->textContent ) . '|' }

	my $dates = '';
	foreach $node ( $mods->findnodes( './/mods:originInfo/mods:dateIssued' ))   { $dates .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:originInfo/mods:dateCreated' ))  { $dates .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:originInfo/mods:dateCaptured' )) { $dates .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:originInfo/mods:dateOther' ))    { $dates .= &normalize( $node->textContent ) . '|' }

	my $types = '';
	foreach $node ( $mods->findnodes( './/mods:typeOfResource' )) { $types .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:genre' ))          { $types .= &normalize( $node->textContent ) . '|' }

	my $formats = '';
	foreach $node ( $mods->findnodes( './/mods:physicalDescription' )) { $formats .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:internetMediaType' ))   { $formats .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:extent' ))              { $formats .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:form' ))                { $formats .= &normalize( $node->textContent ) . '|' }

	my $identifiers = '';
	foreach $node ( $mods->findnodes( './/mods:identifier' ))   { $identifiers .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:location/URL' )) { $identifiers .= &normalize( $node->textContent ) . '|' }

	my $languages = '';
	foreach $node ( $mods->findnodes( './/mods:language' )) { $languages .= &normalize( $node->textContent ) . '|' }

	my $relations = '';
	foreach $node ( $mods->findnodes( './/mods:relatedItem' )) { $relations .= &normalize( $node->textContent ) . '|' }

	my $coverages = '';
	foreach $node ( $mods->findnodes( './/mods:subject/mods:geographic' ))             { $coverages .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:subject/mods:temporal' ))               { $coverages .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:subject/mods:hierarchicalGeographic' )) { $coverages .= &normalize( $node->textContent ) . '|' }
	foreach $node ( $mods->findnodes( './/mods:subject/mods:cartographics' ))          { $coverages .= &normalize( $node->textContent ) . '|' }

	my $rights = '';
	foreach $node ( $mods->findnodes( './/accessCondition' )) { $rights .= &normalize( $node->textContent ) . '|' }

	# display
	print "        record: $index\n";
	print "         title: $titles\n";
	print "      creators: $creators\n";
	print "      subjects: $subjects\n";
	print "  descriptions: $descriptions\n";
	print "    publishers: $publishers\n";
	print "         dates: $dates\n";
	print "         types: $types\n";
	print "       formats: $formats\n";
	print "   identifiers: $identifiers\n";
	print "     languages: $languages\n";
	print "     relations: $relations\n";
	print "     coverages: $coverages\n";
	print "        rights: $rights\n";
	print "\n";
	
	# update the index
	my $doc = $invindexer->new_doc( $index );
	$doc->set_value( coverage    => $coverages );
	$doc->set_value( creator     => $creators );
	$doc->set_value( date        => $dates );
	$doc->set_value( description => $descriptions );
	$doc->set_value( format      => $formats );
	$doc->set_value( identifier  => $identifiers );
	$doc->set_value( language    => $languages );
	$doc->set_value( publisher   => $publishers );
	$doc->set_value( relation    => $relations );
	$doc->set_value( rights      => $rights );
	$doc->set_value( subject     => $subjects );
	$doc->set_value( title       => $titles );
	$doc->set_value( type        => $types );
	$doc->set_value( collection  => 'mods' );
	$invindexer->add_doc( $doc );

}

# optimize
print "Optimizing... ";
$invindexer->finish( optimize => 1 );
print "done.\n";

# done
exit;


sub normalize {

	my $s = shift;
	while ( $s =~ /^\s/ ) { $s =~ s/^\s// }
	while ( $s =~ /\s$/ ) { $s =~ s/\s$// }
	$s =~ s/\t/ /g;
	$s =~ s/\n/ /g;
	$s =~ s/\r/ /g;
	while ( $s =~ /  / ) { $s =~ s/  / / }
	return $s;
	
}