Computes the ancestry path of GeoPlanet places

compute_ancestry.pl

use strict;
use warnings;

use constant {
    # Original Yahoo! TSV.
    GPP => 'geoplanet_places_7.6.0.tsv',

    # Output, same Yahoo! TSV with an extra ancestry column.
    ANC => 'geoplanet_places_with_ancestry_7.6.0.tsv'
};

# We compute woeid -> ancestor list in this hash.
# 
# We need to deal with about 5.7 million records, so this script
# is optimized for speed an memory consumption. In particular this
# hash is global and will be modified in place.
my %ancestors = ();

# Initializes %ancestors with woeid -> [parent_id].
sub initialize_ancestors {
    open my $gpp_fh, GPP or die $!;
    while (my $line = <$gpp_fh>) {
        next unless $line =~ /^\d/; # data records start with a woeid
        chomp $line;

        my @fields = split /\t/, $line;
        my ($woeid, $parent_id) = @fields[0, -1];
        $ancestors{$woeid} = [$parent_id];
    }    
}

# Once %ancestors is initialized, this subroutine returns a list
# with all the ancestors of $start up to the root Earth node.
#
# Note that the code leverages branches already computed, since
# they are also stored in %ancestors.
sub branch_up_to_earth {
    my $start = shift;
    if ($start == 0) {
        ();
    } elsif ($start == 1) {
        ($start);
    } elsif (@{$ancestors{$start}} > 1) {
        ($start, @{$ancestors{$start}});
    } else {
        ($start, branch_up_to_earth($ancestors{$start}[0]));
    }
}

# Ouput the original TSV plus the ancestry for each record.
sub append_ancestry_to_tsv {
    open my $gpp_fh, GPP or die $!;
    open my $anc_fh, '>', ANC or die $!;

    my $header = <$gpp_fh>;
    chomp $header;
    print $anc_fh "$header\tAncestry\n";

    while (my $line = <$gpp_fh>) {
        if ($line =~ /^\d+/) {
            my $woeid = $&;

            chomp $line;
            print $anc_fh "$line\t";

            $ancestors{$woeid} = [branch_up_to_earth($ancestors{$woeid}[0])];
            print $anc_fh '""', join('/', reverse(@{$ancestors{$woeid}})), '""' if @{$ancestors{$woeid}};
            print $anc_fh "\n";        
        }
    }
}

initialize_ancestors;
append_ancestry_to_tsv;