initial overdrive import code

[sitka/sitka-tools.git] / marc_import_overdrive / overdrive-import.pl

#!/usr/bin/perl
use strict;
use warnings;

use lib '/openils/lib/perl5/';

use Error qw/:try/;
use OpenILS::Utils::Fieldmapper;
use Digest::MD5 qw/md5_hex/;
use OpenSRF::Utils::JSON;
use OpenILS::Application::AppUtils;
use Data::Dumper;
use Unicode::Normalize;
use Encode;

use FileHandle;
use Time::HiRes qw/time/;
use Getopt::Long;
use MARC::Batch;
use MARC::File::XML ( BinaryEncoding => 'utf-8' );
use MARC::Charset;
use DBI;

#MARC::Charset->ignore_errors(1);

my ($config, $idlfile, $marctype, $enc) =
	('/srv/openils/conf/opensrf_core.xml', '/srv/openils/conf/fm_IDL.xml', 'USMARC', 'utf8');

my (@files, @trash_fields, @req_fields, $quiet, $startid);

my @targetorg = ('BPR','BFN','BTE');

@req_fields = ('856');
my $overdrive_prefix = 'http\:\/\/downloads\.bclibrary\.ca\/ContentDetails\.htm\?ID\=';
my $tcn_prefix = "LtG_";

GetOptions(
	'marctype=s'	=> \$marctype, # format of MARC files being processed defaults to USMARC, often set to XML
	'encoding=s'	=> \$enc, # set assumed MARC encoding for MARC::Charset
	'config=s'	=> \$config, # location of OpenSRF core config file, defaults to /openils/conf/opensrf_core.xml
	'file=s'	=> \@files, # files to process (or you can simple list the files as unnamed arguments, i.e. @ARGV)
	'required_fields=s'	=> \@req_fields, # skip any records missing these fields
	'trash=s'	=> \@trash_fields, # fields to remove from all processed records
	'xml_idl=s'	=> \$idlfile, # location of XML IDL file, defaults to /openils/conf/fm_IDL.xml
	'startid=i'	=> \$startid, #starting ID
	'quiet'		=> \$quiet # do not output progress count
);

@trash_fields = split(/,/,join(',',@trash_fields));
@req_fields = split(/,/,join(',',@req_fields));

if ($enc) {
	MARC::Charset->ignore_errors(1);
	MARC::Charset->assume_encoding($enc);
}

if (uc($marctype) eq 'XML') {
	'open'->use(':utf8');
} else {
	bytes->use();
}

@files = @ARGV if (!@files);

Fieldmapper->import(IDL => $idlfile);

select STDERR; $| = 1;
select STDOUT; $| = 1;

my $batch = new MARC::Batch ( $marctype, @files );
$batch->strict_off();
$batch->warnings_off();

my $starttime = time;
my $rec;
my $count = 0;

my $id = 1 || $startid;

PROCESS: while ( try { $rec = $batch->next } otherwise { $rec = -1 } ) {
	next if ($rec == -1);
    print STDERR "======\n";
    $id++;
	$count++;

	# Skip records that don't contain a required field (like '245', for example)
	foreach my $req_field (@req_fields) {
		if (!$rec->field("$req_field")) {
			warn "\n!!! Record $count missing required field $req_field, skipping record.\n";
			next PROCESS;
		}
	}

# -----------------
# Overdrive - specific code
# -----------------

    my $tcn_value; 
	my $tcn_source = 'Library To Go';

    my $caption = 'DOWNLOADABLE AUDIOBOOK';

    # this is the base 856 field we're going to generate separate fields for each org unit we're scoping at
    my $baseurifield;

    # check all 856s
    URIFIELD: foreach my $uri ($rec->field('856')){

        # Overdrive uses a $3 for Excerpts, we want to keep this intact so carry on then
        next URIFIELD if ($uri->subfield('3')); 

        # we need a $u for a URL, if we don't have this it is bad
        my $url = $uri->subfield('u');
        if(!$url){
            warn "856 has no URL in rec $id.  Skipping";
        } 

        # this record has been through Evergreen if a $9 exists somewhere
        # instead, we can presumably pull the tcn from the 901 and generate our base field that way
        if($uri->subfield('9')){
            # if this doesn't match our prefix, ignore it, we only care about our current prefix
            next unless($url =~ m/($overdrive_prefix)/);

            # if it does:
            # delete $9 subfield and use this as a base uri field
            $baseurifield = $uri->clone;
            $rec->delete_fields($uri);
            next URIFIELD;
        }


        if($uri->subfield('z')){
            $caption = 'EBOOK' if ($uri->subfield('z') =~ /Book/);
            $uri->delete_subfield(code => 'z');
            $uri->add_subfields('z' => 'Click to access online (library card required)');

            next unless($url =~ m/($overdrive_prefix)/);

            # trim out Overdrive's magical GUID-looking ID thingy
            my $overdrivekey = $url;
            $overdrivekey =~ s/($overdrive_prefix)//g;

            # make it TCN-ish
	        $tcn_value = $tcn_prefix . $overdrivekey;

            # we have a base for our scoped fields
            $baseurifield = $uri->clone;

            $rec->delete_fields($uri);
        }
    }

    if(!$baseurifield){
        die "the horror!";
    }
    # add some arbitrary stuff as prescribed by our cataloguer overlords
    $rec = adjust_leader($rec);
    $rec = process_custom_fields($rec);


    # add a scoped field for each org unit in our array
    foreach(@targetorg){
        if($baseurifield){
            my $newfield = $baseurifield->clone();
            $newfield->add_subfields('9' => $_);
            $rec->insert_fields_ordered($newfield);
        } else {
            die;# $rec->as_formatted();
        }
    }

# -----------------
# END Overdrive - specific code
# -----------------


	$rec->delete_field($_) for ($rec->field(@trash_fields));

    my $field901 = MARC::Field->new(
            '901' => ('', ''),
            a => $tcn_value,
            b => $tcn_source,
            c => $id,
    );


    $rec->insert_fields_ordered($field901);

    print $rec->as_formatted();

    next PROCESS;


	$tcn_value = $rec->subfield('901' => 'a');
	$tcn_source = $rec->subfield('901' => 'b');
	$id = $rec->subfield('901' => 'c');

	(my $xml = $rec->as_xml_record()) =~ s/\n//sog;
	$xml =~ s/^<\?xml.+\?\s*>//go;
	$xml =~ s/>\s+</></go;
	$xml =~ s/\p{Cc}//go;
	$xml = OpenILS::Application::AppUtils->entityize($xml);
	$xml =~ s/[\x00-\x1f]//go;

	my $bib = new Fieldmapper::biblio::record_entry;
	$bib->id($id);
	$bib->active('t');
	$bib->deleted('f');
	$bib->marc($xml);
	$bib->creator(0);
	$bib->create_date('now');
	$bib->editor(0);
	$bib->edit_date('now');
	$bib->tcn_source($tcn_source);
	$bib->tcn_value($tcn_value);
	$bib->last_xact_id('IMPORT-'.$starttime);

	#print OpenSRF::Utils::JSON->perl2JSON($bib)."\n";

	if (!$quiet){# && !($count % 50)) {
		print STDERR "\r$count\t". $count / (time - $starttime);
	}
}

sub adjust_leader {
    my $rec = shift;
    my $leader = $rec->leader();
    $leader = substr($leader,0,5) . 'm' . substr($leader,6,length($leader));
    $rec->leader($leader);
    return $rec;
}

sub process_custom_fields{
    my $rec = shift;
    my $caption = shift;
   
    my @newfields;

    push @newfields, MARC::Field->new(
        '538' => (' ', ' '),
        a => "Requires OverDrive Media Console"
    );
    push @newfields, MARC::Field->new(
        '594' => (' ', ' '),
        a => "Library To Go"
    );

    push @newfields, MARC::Field->new(
        '655' => (' ', '4'),
        'a' => $caption 
    );    

    $rec->insert_fields_ordered(@newfields);
    return $rec;
}