#!/usr/bin/perl

use strict;
use warnings;

my $skip = shift || 0;
my $count = 0;
$| = 1;

open MARC, '<', 'incoming.marc.xml';
open NUMARC, '>', 'incoming.clean.marc.xml';

until ($count == ($skip - 1)) {
    my $t = <MARC>;
    print NUMARC $t;
    $count++;
    printf("\rSpinning on to record %s (%2.2f%%)", $skip, ($count / $skip *100))
      unless ($count % 1000);
}
print "\nScrubbing resumes...\n" if $skip;

my $line1 = <MARC>;

while (my $line2 = <MARC>) {
    $count++;
    # catch empty datafield elements
    if ($line1 =~ m/<datafield tag="..." ind1="." ind2=".">/) {
        if ($line2 =~ m|</datafield>|) {
            $line1 = <MARC>;
            $count++;
            next;
        }
    }

    # clean up tags with spaces in them
    $line1 =~ s/tag="  /tag="00/g;
    $line1 =~ s/tag=" /tag="0/g;
    $line1 =~ s/tag="-/tag="0/g;
    $line1 =~ s/tag="(\d\d) /tag="0$1/g;

    # subfields can't be non-alphanumeric
    die "Junk in subfield at line $count: $line1"
      if $line1 =~ /<subfield code="[^[:alnum:]]"/;

    # everything looks ok
    print NUMARC $line1;
    $line1 = $line2;
}
print NUMARC $line1;