A file to split a DwC-A file into triples

delimit2rdf.pl

#!/usr/bin/perl -w

use strict;
use warnings;

use v5.10.0;

=head1 NAME

delimit2rdf.pl -- Delimited file to RDF

=head1 SYNOPSIS

    delimit2rdf.pl < input.txt > output.rdf

=head1 DESCRIPTION

This script is basically the core of a DwC-A to
RDF converter. It uses information core/extension
nodes in meta.xml (see http://rs.tdwg.org/dwc/terms/guides/text/index.htm#metafile
for more details) to convert a tab delimited file
into an RDF file.

=cut

use RDF::Helper;

# Put all the conversion information here.
my $encoding = undef;           # Not implemented yet.
my $fieldsTerminatedBy = "\t";
my $linesTerminatedBy = "\n";
my $fieldsEnclosedBy = undef;   # Not implemented yet.
my $ignoreHeaderLines = 0;
my $rowType = "http://rs.tdwg.org/dwc/terms/Taxon";

my @fields = (
    undef,      # Use this field as a primary id.
    'http://rs.tdwg.org/dwc/terms/parentNameUsageID',
    'http://rs.tdwg.org/dwc/terms/acceptedNameUsageID',
    'http://rs.tdwg.org/dwc/terms/scientificName',
    'http://rs.gbif.org/terms/1.0/canonicalName',
    'http://rs.tdwg.org/dwc/terms/taxonRank',
    'http://rs.tdwg.org/dwc/terms/taxonomicStatus',
    'http://rs.tdwg.org/dwc/terms/nomenclaturalStatus',
    'http://rs.tdwg.org/dwc/terms/genus',
    'http://rs.tdwg.org/dwc/terms/specificEpithet',
    'http://rs.tdwg.org/dwc/terms/infraspecificEpithet',
    'http://rs.tdwg.org/dwc/terms/namePublishedIn',
    'http://rs.tdwg.org/dwc/terms/nameAccordingTo',
    'http://rs.tdwg.org/dwc/terms/kingdom',
    'http://rs.tdwg.org/dwc/terms/phylum',
    'http://rs.tdwg.org/dwc/terms/class',
    'http://rs.tdwg.org/dwc/terms/order',
    'http://rs.tdwg.org/dwc/terms/family'
);

# Check the input.
my %primary_key_check;

die "encoding not yet supported! Use utf8 please." if defined $encoding;
die "fieldsEnclosedBy is not yet supposed!" if defined $fieldsEnclosedBy;
die "The first field must be a primary key index!" unless(not defined $fields[0]);

# Conversion time!
=comment
my $rdf = RDF::Helper->new(
    BaseInterface => 'RDF::Trine',
    namespaces => {
        'rdf' =>        'http://www.w3.org/1999/02/22-rdf-syntax-ns#',
        'dwc' =>        'http://rs.tdwg.org/dwc/terms/',
        'gbifterms' =>  'http://rs.gbif.org/terms/1.0/',
        'gbifnub' =>    'http://gaurav.github.com/gbif-nub-dwca/'
    },
    ExpandQNames => 1
);
=cut

say <<'HEADER';
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix dwc: <http://rs.tdwg.org/dwc/terms/> .
@prefix gbifterms: <http://rs.gbif.org/terms/1.0/> .
@prefix gbifnub_id: <http://gaurav.github.com/gbif-nub-dwca/id_> .
@prefix gbifnub_name: <http://gaurav.github.com/gbif-nub-dwca/name_> .

HEADER

say STDERR "Started conversion.";
my $count_resources = 0;
my $count_triples = 0;
{
    local $/ = $linesTerminatedBy;  # Line terminator
    <> if $ignoreHeaderLines;       # Skip a line.
    while(<>) {
        chomp;
        my $line = $_;

        my @records = split($fieldsTerminatedBy, $line);
        my $index = $records[0];

        die "Duplicate id: $index" if exists $primary_key_check{$index};
        $primary_key_check{$index} = 1;

        my $resource = 'gbifnub_id:' . $index;
        # $rdf->assert_resource($resource, 'rdf:type', $rowType);

        say "<$resource> rdf:type <$rowType> ;\n";

        my $count = 0;
        foreach my $x (@records) {
            # Skip the first record.
            if($count == 0) {
                die "Invalid first field" unless defined $x;
                $count++;
                next;
            }

            my $term_name = $fields[$count];
            my $term_value = convert_term_value($term_name, $x);
            
            $count++;
            next unless defined $term_value;

            say "\t<$term_name> $term_value;";
            $count_triples++;
        }

        say ".";

        say STDERR "Processed '$records[3]' (#$index).";
        $count_resources++;
    }
}

# print $rdf->serialize(format => 'rdfxml');
 
say STDERR "Conversion completed, $count_resources resources processed, $count_triples created.";

sub convert_term_value {
    my ($term_name, $term_value) = @_;

    given($term_name) {
        when('http://rs.tdwg.org/dwc/terms/parentNameUsageID') {
            return undef if $term_value eq '';
            return "<gbifnub_id:$term_value>";
        }
    	when('http://rs.tdwg.org/dwc/terms/acceptedNameUsageID') {
            return undef if $term_value eq '';
            return "\"$term_value\"";
        }
    	when('http://rs.tdwg.org/dwc/terms/scientificName') {
            return undef if $term_value eq '';
            return "\"$term_value\"";
        }
#    	when('http://rs.gbif.org/terms/1.0/canonicalName',
#    	when('http://rs.tdwg.org/dwc/terms/taxonRank',
#    	when('http://rs.tdwg.org/dwc/terms/taxonomicStatus',
#    	when('http://rs.tdwg.org/dwc/terms/nomenclaturalStatus',
#    	when('http://rs.tdwg.org/dwc/terms/genus',
#    	when('http://rs.tdwg.org/dwc/terms/specificEpithet',
#    	when('http://rs.tdwg.org/dwc/terms/infraspecificEpithet',
#    	when('http://rs.tdwg.org/dwc/terms/namePublishedIn',
#    	when('http://rs.tdwg.org/dwc/terms/nameAccordingTo',
#    	when('http://rs.tdwg.org/dwc/terms/kingdom',
#    	when('http://rs.tdwg.org/dwc/terms/phylum',
#    	when('http://rs.tdwg.org/dwc/terms/class',
#    	when('http://rs.tdwg.org/dwc/terms/order',
#    	when('http://rs.tdwg.org/dwc/terms/family'
#
    }

    return "\"$term_value\"";
}