dmd-Preservation-IssueInfoDC

#!/usr/bin/env perl

package dmdPreservationIssueinfoDC;

use strict;
use warnings;
use FindBin;
use lib "$FindBin::RealBin/../lib";
use Getopt::Long;
use Fcntl qw(:DEFAULT :flock);
use Try::Tiny;
use Log::Log4perl;
use File::Find;
use Data::Dumper;
use Digest::MD5;
use JSON;
use Poppler;
use XML::LibXML;
use CIHM::Swift::Client;
use URI::Escape;
use DateTime::Format::ISO8601;
use HTTP::Date qw(:DEFAULT time2isoz);
use File::Spec;
use File::stat;
use XML::LibXML;
use utf8;
use Switch;
use CIHM::Meta::DCCSV;

Log::Log4perl->init_once("/etc/canadiana/tdr/log4perl.conf");
my $logger = Log::Log4perl::get_logger("CIHM::TDR");

my $inThisFile;

sub log_warnings {
    my $warning = shift;
    chomp $warning;

    # Strip wide characters before  trying to log
    ( my $stripped = $warning ) =~ s/[^\x00-\x7f]//g;
    $stripped =~ s/\R/ /g;

    if ( defined $inThisFile ) {
        print STDERR "$stripped in $inThisFile\n";
        $logger->warn("$stripped in $inThisFile");
    }
    else {
        $logger->warn($stripped);
        print STDERR "$stripped\n";
    }
}
local $SIG{__WARN__} = sub { &log_warnings };

my $dmddir = '/crkn-nas-wip/_Metadata_Synchronised';
my $csvdir = '/crkn-nas-wip/_Metadata_Synchronised/Fixed/IssueinfoDC';

$logger->info("DMD Preservation Issueinfo DC: start");

GetOptions(
    'dmddir:s' => \$dmddir,
    'csvdir:s' => \$csvdir,
);

# Used to show a different processname during processing
my $dmddcprog = $0;

#
my $dccsv;

## Test an ID
#$dccsv = CIHM::Meta::DCCSV->new( { csvdir => $csvdir } );
#$dccsv->addElement( "oocihm.7a9790", "thisis" , "atest" );
#exit 0;

my $preservationdir = File::Spec->catfile( $dmddir, 'preservation' );
opendir( my $dh, $preservationdir ) || die "Can't open $preservationdir: $!";
while ( readdir $dh ) {
    if ( substr( $_, 0, 1 ) ne "." ) {
        my $prefixdir = $_;
        my $preservationcsvdir =
          File::Spec->catfile( $csvdir, 'preservation', $prefixdir );
        my $preservationsubdir =
          File::Spec->catfile( $dmddir, 'preservation', $prefixdir );

        $dccsv = CIHM::Meta::DCCSV->new( { csvdir => $preservationcsvdir } );

        $0 = $dmddcprog . " find in $preservationsubdir";
        print "Find for $prefixdir\n";

        find( \&matching_dmd_file, $preservationsubdir );

        print "Write for $prefixdir\n";

        $dccsv->writeCSV();
        undef $dccsv;

    }
}
closedir $dh;

exit 0;

sub matching_dmd_file {
    if ( -f $_ ) {
        $0 = $dmddcprog . " looking for Issueinfo in $File::Find::dir";
        if (/^(.*)-(\w*)\.xml$/) {
            my $id      = $1;
            my $dmdType = $2;

            if ( $dmdType eq 'issueinfo' ) {
                $inThisFile = $File::Find::name;

                # Add AIP ID as an identifier
                $dccsv->addElement( $id, "identifier", $id );

                open my $fh, '<:encoding(UTF-8)', $File::Find::name
                  or die "can't open IssueInfo file=$File::Find::name: $!\n";

                my $xmlin = do { local $/; <$fh> };
                close $fh;
                undef $inThisFile;

                my $doc = XML::LibXML->new->parse_string($xmlin);

                my $xpc = XML::LibXML::XPathContext->new;
                $xpc->registerNs( 'issueinfo',
                    'http://canadiana.ca/schema/2012/xsd/issueinfo' );

                my @allnodes = $xpc->findnodes( "//*", $doc );

                foreach my $node (@allnodes) {
                    my $content = normaliseSpace( $node->textContent );
                    my $name    = lc( $node->nodeName );

                    if ( $name eq 'coverage' ) {
                        my ( $start, $end );
                        my @attributes = $node->attributes();
                        foreach my $attribute (@attributes) {
                            switch ( $attribute->nodeName ) {
                                case "start" {
                                    $start = $attribute->value;
                                }
                                case "end" {
                                    $end = $attribute->value;
                                }
                                else {
                                    warn
"$File::Find::name Unknown attribute name: "
                                      . $attribute->nodeName
                                      . " value: "
                                      . $attribute->value . " \n";
                                }
                            }
                        }
                        $content = "$start/$end";
                    }

                    if ( length($content) ) {
                        switch ($name) {
                            case "issueinfo" {    #  Skip top level
                            }
                            case "series" {
                                $dccsv->addElement( $id, "relation", $content );
                            }
                            case "title" {
                                $dccsv->addElement( $id, "title", $content );
                            }
                            case "sequence" {     #  Unused
                            }
                            case "language" {
                                $dccsv->addElement( $id, "language", $content );
                            }
                            case "coverage" {
                                $dccsv->addElement( $id, "coverage", $content );
                            }
                            case "published" {
                                $dccsv->addElement( $id, "date", $content );
                            }
                            case "pubstatement" {
                                $dccsv->addElement( $id, "publisher",
                                    $content );
                            }
                            case "source" {
                                $dccsv->addElement( $id, "source", $content );
                            }
                            case "identifier" {
                                $dccsv->addElement( $id, "identifier",
                                    $content );
                            }
                            case "note" {
                                $dccsv->addElement( $id, "description",
                                    $content );
                            }
                            else {
                                warn "Unknown issueinfo node name: "
                                  . $node->nodeName . "\n";
                            }
                        }
                    }
                }
            }
        }
        else {
            warn $_ . "  -- doesn't match pattern (in $File::Find::dir )\n";
        }
    }
}

sub normaliseSpace {
    my $content = shift;

    $content =~ s/^\s+|\s+$//g;    # Trim space at end and beginning.
    $content =~ s/\s+/ /g;         # Remove extra spaces

    return $content;
}