11 my $term = new Term::ReadLine 'yaz-cleanup';
12 my $OUT = $term->OUT || \*STDOUT;
20 my %trash = (); # hash for tags to be dumped
22 # initialization and setup
25 # read in trash tags file if it exists
26 populate_trash() if (-e '.trashtags');
28 my @record = (); # current record storage
29 my %recmeta = (); # metadata about current record
30 my @context= (); # last 5 lines of file
32 my $input = shift || 'incoming.marc.xml';
34 open MARC, '<', $input;
35 open my $NUMARC, '>', $conf->{output};
36 print $NUMARC '<?xml version="1.0" encoding="UTF-8"?>',"\n";
37 print $NUMARC '<collection xmlns="http://www.loc.gov/MARC21/slim">',"\n";
39 open my $EXMARC, '>', $conf->{exception};
40 print $EXMARC '<?xml version="1.0" encoding="UTF-8"?>',"\n";
41 print $EXMARC '<collection xmlns="http://www.loc.gov/MARC21/slim">',"\n";
42 open MARC2, '<', $input;
45 # this is the dispatch table which drives command selection in
47 my %commands = ( c => \&print_context,
48 C => \&print_linecontext,
61 my @spinner = qw(- / | \\);
64 while (my $line = getline()) {
65 unless ($count % 2000) {
66 print "\rWorking... ", $spinner[$sidx];
67 $sidx = ($sidx == $#spinner) ? 0 : $sidx + 1;
71 next if ($line =~ m|</record>|);
73 # catch empty datafield elements
74 if ($line =~ m|</datafield>|) {
75 if ($record[-2] =~ m/<datafield tag="..." ind1="." ind2=".">/) {
76 pop @record; pop @record;
77 message("Empty datafield scrubbed");
83 if ($line =~ m|<leader>(.+?)</leader>|) {
85 if (length $leader < 24) {
86 $leader .= ' ' x (20 - length($leader));
88 $line = "<leader>$leader</leader>\n";
89 message("Short leader padded");
93 # clean misplaced dollarsigns
94 if ($line =~ m|<subfield code="\$">c?\d+\.\d{2}|) {
95 $line =~ s|"\$">c?(\d+\.\d{2})|"c">\$$1|;
96 message("Dollar sign corrected");
99 # clean up tags with spaces in them
100 $line =~ s/tag=" /tag="00/g;
101 $line =~ s/tag=" /tag="0/g;
102 $line =~ s/tag="-/tag="0/g;
103 $line =~ s/tag="(\d\d) /tag="0$1/g;
105 # stow tag data if we're looking at it
106 if ($line =~ m/<datafield tag="(.{3})" ind1="(.)" ind2="(.)">/) {
112 # and stow line back in record
116 if ($line =~ /&/ && $line !~ /&\w+?;/)
117 { edit("Naked ampersand", $line); next }
119 # tags must be numeric
120 if ($line =~ /<datafield tag="(.+?)"/) {
122 if ($match =~ /\D/) {
123 edit("Non-numerics in tag", $line);
128 # subfields can't be non-alphanumeric
129 if ($line =~ /<subfield code="(.+?)"/) {
131 if ($match =~ /\P{IsAlnum}/) {
132 edit("Junk in subfield code", $line);
138 print $NUMARC "</collection>\n";
139 print $EXMARC "</collection>\n";
140 print $OUT "\nDone. \n";
144 Handles the Term::ReadLine loop
149 my ($msg, $line_in) = @_;
150 return if $trash{$recmeta{tag}};
155 my $line = $term->readline('marc-cleanup>');
156 my @chunks = split /\s+/, $line;
158 if (length $chunks[0] == 1)
159 { next unless (defined $commands{$chunks[0]}) }
161 if (defined $commands{$chunks[0]}) {
162 my $term = $commands{$chunks[0]}->($line_in, @chunks[1..$#chunks]);
165 if ($context[3] eq " [LINE KILLED]\n") {
166 push @record, "$line\n"
168 $record[-1] = "$line\n";
170 $context[3] = "$line\n";
178 Reads from the incoming MARC file; returns lines into the driver
179 loop. Batches records for output, and maintains the context listing.
187 if ($l =~ /<record>/) {
191 } elsif ($l =~ m|</record>|) {
192 write_record($NUMARC) if $reccount;
202 $oreccount++ if ($FH eq $NUMARC);
203 print $FH '<!-- ', $recmeta{explanation}, " -->\n"
204 if(defined $recmeta{explanation});
206 # excise unwanted tags
207 if (keys %trash or $conf->{autoscrub}) {
210 for my $line (@record) {
212 $istrash = 0 if $line =~ m|</datafield|;
215 if ($line =~ m/<datafield tag="(.{3})"/) {
217 if ($trash{$tag} or ($conf->{autoscrub} and $tag =~ /\D/)) {
222 push @trimmed, $line;
228 unless ($conf->{nocollapse}) {
229 s/\n// for (@record);
232 # add 903(?) with new record id
234 if ($conf->{'renumber-from'}) {
235 $renumber = join('', '<datafield tag="', $conf->{'renumber-tag'}, '">',
236 '<subfield code="', $conf->{'renumber-subfield'}, '">',
237 $conf->{'renumber-from'}, '</subfield></datafield>');
238 $renumber .= "\n" if $conf->{nocollapse};
239 push @record, $renumber;
240 $conf->{'renumber-from'}++;
244 print $FH "</record>\n";
247 sub update_linecontext {
249 push @context, $line2;
250 shift @context if (@context > 5);
255 print $OUT "\r$msg at record $reccount/",$oreccount + 1,"\n";
259 #-----------------------------------------------------------------------------------
261 #-----------------------------------------------------------------------------------
264 my ($line_in, @chunks) = @_;
265 my $ofrom = shift @chunks;
266 if ($ofrom =~ /^'/ or !@chunks) {
267 until ($ofrom =~ /'$/)
268 { $ofrom .= join(' ','',shift @chunks) }
269 $ofrom =~ s/^'//; $ofrom =~ s/'$//;
271 my $to = shift @chunks;
273 until ($to =~ /'$/ or !@chunks)
274 { $to .= join(' ','',shift @chunks) }
275 $to =~ s/^'//; $to =~ s/'$//;
279 for my $char (split(//,$ofrom)) {
280 $char = "\\" . $char if ($char =~ /\W/);
281 $from = join('', $from, $char);
283 $record[-1] =~ s/$from/$to/;
284 $context[3] = $record[-1];
290 my $last = pop @record;
292 $record[-1] =~ s/\n//;
293 $record[-1] = join('', $record[-1], $last);
295 push @temp, @context[0..1];
296 $temp[3] = $record[-1];
297 $temp[4] = $context[4];
305 $context[3] = " [LINE KILLED]\n";
311 my ($line_in, @explanation) = @_;
312 $recmeta{explanation} = join(' ', 'Tag', $recmeta{tag}, @explanation);
313 my $line = <MARC>; $count++;
314 update_linecontext();
315 until ($line =~ m|</record>|) {
317 $line = <MARC>; $count++;
318 update_linecontext();
321 write_record($EXMARC);
325 sub commit_edit { return 1 }
328 print "\n Tag:",$recmeta{tag}, " Ind1:'",
329 $recmeta{ind1},"' Ind2:'", $recmeta{ind2}, "'";
334 sub print_linecontext {
335 print $OUT "\n", join(' |','',@context[0..2]);
336 print $OUT '==> |', $context[3];
337 print $OUT ' |', $context[4],"\n";
343 print $OUT "\n$line_in\n";
350 Type a replacement for the indicated line, or enter a command.
352 Commands: c Show record context ('C' for brief context)
353 k Kill indicated line (remove from record)
354 m Merge indicated line with previous line
356 s Substitute ARG1 for ARG2 in indicated line
357 t Commit changes and resume stream edit
358 x Write this record to the exception file instead of output
367 #-----------------------------------------------------------------------------------
369 #-----------------------------------------------------------------------------------
370 # defined a domain-specific language for specifying MARC tags to be dropped from
371 # records during processing. it is line oriented, and is specified as follows:
373 # each line may specify any number of tags to be included, either singly (\d{1,3})
374 # or as a range (\d{1,3}\.\.\d{1,3}
376 # if a single number is given, it must be between '000' and '999', inclusive.
378 # ranges obey the previous rule, and also the first number of the range must be less
379 # than the second number
381 # finally, any single range in a line may be followed by the keyword 'except'. every
382 # number or range after 'except' is excluded from the range specified. all these
383 # numbers must actually be within the range.
385 # specifying a tag twice is an error, to help prevent typos
388 print $OUT ">>> TRASHTAGS FILE FOUND. LOADING TAGS TO BE STRIPPED FROM OUTPUT...\n";
389 open TRASH, '<', '.trashtags';
391 my $lastwasrange = 0;
392 my %lastrange = ( high => 0, low => 0);
395 my @chunks = split /\s+/;
396 while (my $chunk = shift @chunks) {
399 if ($chunk =~ /^\d{1,3}$/) {
400 trash_add($chunk, $except);
406 if ($chunk =~ /^\d{1,3}\.\.\d{1,3}$/) {
407 my ($low, $high) = trash_add_range($chunk, $except, \%lastrange);
409 %lastrange = (low => $low, high => $high)
415 if ($chunk eq 'except') {
416 die "Keyword 'except' can only follow a range (line $.)\n"
417 unless $lastwasrange;
418 die "Keyword 'except' may only occur once per line (line $.)\n"
424 die "Unknown chunk $chunk in .trashtags file (line $.)\n";
428 # remove original id sequence tag from trash hash if we know it
429 trash_add($conf->{'original-tag'}, 1)
430 if ($conf->{'original-tag'} and $trash{$conf->{'original-tag'}});
433 sub trash_add_range {
434 my ($chunk, $except, $range) = @_;
435 my ($low,$high) = split /\.\./, $chunk;
436 die "Ranges must be 'low..high' ($low is greater than $high on line $.)\n"
439 die "Exception ranges must be within last addition range (line $.)\n"
440 if ($low < $range->{low} or $high > $range->{high});
442 for my $tag ($low..$high) {
443 trash_add($tag, $except)
449 my ($tag, $except) = @_;
450 die "Trash values must be valid tags (000-999)\n"
451 unless ($tag >= 0 and $tag <= 999);
455 die "Trash tag '$tag' specified twice (line $.)\n"
461 #-----------------------------------------------------------------------
465 Performs boring script initialization. Handles argument parsing,
474 # set mode on existing filehandles
475 binmode(STDIN, ':utf8');
477 my $rc = GetOptions( $c,
482 'renumber-from|rf=i',
485 'renumber-subfield|rt=i',
488 show_help() unless $rc;
489 show_help() if ($c->{help});
492 $c->{output} = 'incoming.cleaned.marc.xml' unless defined $c->{output};
493 $c->{exception} = 'incoming.exception.marc.xml' unless defined $c->{exception};
494 $c->{'renumber-tag'} = 903 unless defined $c->{'renumber-tag'};
495 $c->{'renumber-subfield'} = 'a' unless defined $c->{'renumber-subfield'};
497 my @keys = keys %{$c};
498 show_help() unless (@ARGV and @keys);
499 #for my $key ('runtype', 'tag', 'subfield', 'output', 'exception')
500 # { push @missing, $key unless $c->{$key} }
502 # print "Required option: ", join(', ', @missing), " missing!\n";
509 Usage is: $0 [OPTIONS] <filelist>
511 --output -o Cleaned MARCXML output filename (default: incoming.cleaned.marc.xml)
512 --exception -x Exception (dumped records) MARCXML filename (incoming.exception.marc.xml)