9 my $term = new Term::ReadLine 'yaz-cleanup';
10 my $OUT = $term->OUT || \*STDOUT;
12 # initialization and setup
15 populate_trash() if ($conf->{trash});
17 my @record = (); # current record storage
18 my %recmeta = (); # metadata about current record
19 my $ptr = 0; # record index pointer
21 my $input = shift || 'incoming.marc.xml';
23 open MARC, '<:utf8', $input;
24 open my $NUMARC, '>:utf8', $conf->{output};
25 print $NUMARC '<?xml version="1.0" encoding="UTF-8"?>',"\n";
26 print $NUMARC '<collection xmlns="http://www.loc.gov/MARC21/slim">',"\n";
28 open my $EXMARC, '>:utf8', $conf->{exception};
30 # this is the dispatch table which drives command selection in
32 my %commands = ( c => \&print_fullcontext,
35 O => \&insert_original,
46 '<' => \&widen_window,
47 '>' => \&narrow_window,
53 my @spinner = qw(- / | \\);
56 while ( buildrecord() ) {
57 unless ($conf->{ricount} % 100) {
58 print "\rWorking... ", $spinner[$sidx];
59 $sidx = ($sidx == $#spinner) ? 0 : $sidx + 1;
62 do_automated_cleanups();
65 until ($ptr == $#record) {
67 if ($record[$ptr] =~ /&/ && $record[$ptr] !~ /&\w+?;/)
68 { edit("Naked ampersand"); $ptr= 0; next }
70 # tags must be numeric
71 if ($record[$ptr] =~ /<datafield tag="(.+?)"/) {
74 edit("Non-numerics in tag");
79 # subfields can't be non-alphanumeric
80 if ($record[$ptr] =~ /<subfield code="(.*?)"/) {
82 if ($match =~ /\P{IsAlnum}/ or $match eq '') {
83 edit("Junk in subfield code/Null subfield code");
89 write_record($NUMARC);
91 print $NUMARC "</collection>\n";
92 print $OUT "\nDone. \n";
95 #-----------------------------------------------------------------------------------
97 #-----------------------------------------------------------------------------------
99 sub do_automated_cleanups {
101 until ($ptr == $#record) {
102 # catch empty datafield elements
103 if ($record[$ptr] =~ m/<datafield tag="..." ind1="." ind2=".">/) {
104 if ($record[$ptr + 1] =~ m|</datafield>|) {
105 my @a = @record[0 .. $ptr - 1];
106 my @b = @record[$ptr + 2 .. $#record];
108 message("Empty datafield scrubbed");
113 # and quasi-empty subfields
114 if ($record[$ptr] =~ m|<subfield code="(.*?)">(.*?)</sub|) {
115 my $code = $1; my $content = $2;
116 if ($code =~ /\W/ and ($content =~ /\s+/ or $content eq '')) {
117 my @a = @record[0 .. $ptr - 1];
118 my @b = @record[$ptr + 1 .. $#record];
120 message("Empty subfield scrubbed");
129 for $ptr (0 .. $#record) {
131 if ($record[$ptr] =~ m|<leader>(.+?)</leader>|) {
133 if (length $leader < 24) {
134 $leader .= ' ' x (20 - length($leader));
136 $record[$ptr] = "<leader>$leader</leader>\n";
137 message("Short leader padded");
140 if ($record[$ptr] =~ m|<controlfield tag="008">(.+?)</control|) {
143 if (length $content < 40) {
144 $content .= ' ' x (40 - length($content));
145 $record[$ptr] = "<controlfield tag=\"008\">$content</controlfield>\n";
146 message("Short 008 padded");
150 # clean misplaced dollarsigns
151 if ($record[$ptr] =~ m|<subfield code="\$">c?\d+\.\d{2}|) {
152 $record[$ptr] =~ s|"\$">c?(\d+\.\d{2})|"c">\$$1|;
153 message("Dollar sign corrected");
156 # clean up tags with spaces in them
157 $record[$ptr] =~ s/tag=" /tag="00/g;
158 $record[$ptr] =~ s/tag=" /tag="0/g;
159 $record[$ptr] =~ s/tag="-/tag="0/g;
160 $record[$ptr] =~ s/tag="(\d\d) /tag="0$1/g;
162 # stow tag data if we're looking at it
163 if ($record[$ptr] =~ m/<datafield tag="(.{3})" ind1="(.)" ind2="(.)">/) {
169 # automatable subfield maladies
170 $record[$ptr] =~ s/code=" ">c/code="c">/;
171 $record[$ptr] =~ s/code=" ">\$/code="c"$>/;
175 #-----------------------------------------------------------------------------------
177 #-----------------------------------------------------------------------------------
182 Handles the Term::ReadLine loop
189 return if $conf->{trash}{ $recmeta{tag} };
190 $conf->{editmsg} = $msg;
193 # stow original problem line
194 $conf->{origline} = $record[$ptr];
197 my $line = $term->readline('marc-cleanup>');
198 my @chunks = split /\s+/, $line;
200 if (length $chunks[0] == 1)
201 { next unless (defined $commands{$chunks[0]}) }
203 if (defined $commands{$chunks[0]}) {
204 my $term = $commands{$chunks[0]}->(@chunks[1..$#chunks]);
207 $record[$ptr] = "$line\n";
216 Constructs record arrays from the incoming MARC file and returns them
223 $l = <MARC> while (defined $l and $l !~ /<record>/);
224 return $l unless defined $l;
229 until ($l =~ m|</record>|)
230 { push @record, $l; $l = <MARC>; }
237 my $trash = $conf->{trash};
239 $conf->{rocount}++ if ($FH eq $NUMARC);
240 print $FH '<!-- ', $recmeta{explanation}, " -->\n"
241 if(defined $recmeta{explanation});
243 # excise unwanted tags
244 if (keys %{$trash} or $conf->{autoscrub}) {
247 for my $line (@record) {
249 $istrash = 0 if $line =~ m|</datafield|;
252 if ($line =~ m/<datafield tag="(.{3})"/) {
254 if ($trash->{$tag} or ($conf->{autoscrub} and $tag =~ /\D/)) {
259 push @trimmed, $line;
265 unless ($conf->{nocollapse}) {
266 s/\n// for (@record);
269 # add 903(?) with new record id
271 if ($conf->{'renumber-from'}) {
272 $renumber = join('', '<datafield tag="', $conf->{'renumber-tag'},
273 '" ind1=" " ind2=" ">',
274 '<subfield code="', $conf->{'renumber-subfield'}, '">',
275 $conf->{'renumber-from'}, '</subfield></datafield>');
276 $renumber .= "\n" if $conf->{nocollapse};
277 push @record, $renumber;
278 $conf->{'renumber-from'}++;
282 print $FH "</record>\n";
285 sub print_fullcontext {
286 print $OUT "\r", ' ' x 72, "\n";
287 print $OUT $conf->{editmsg},"\n";
288 print $OUT "\r Tag:",$recmeta{tag}, " Ind1:'",
289 $recmeta{ind1},"' Ind2:'", $recmeta{ind2}, "'";
295 my $upper = int($conf->{window} / 2) + 1;
296 my $lower = int($conf->{window} / 2) - 1;
297 my $start = ($ptr - $upper < 0) ? 0 : $ptr - $upper;
298 my $stop = ($ptr + $lower > $#record) ? $#record : $ptr + $lower;
300 print $OUT ' |', $record[$_] for ($start .. $ptr - 1);
301 print $OUT '==> |', $record[$ptr];
302 print $OUT ' |', $record[$_] for ($ptr + 1 .. $stop);
309 print $OUT "\r$msg at ",$conf->{ricount},"/",$conf->{rocount} + 1,"\n";
312 #-----------------------------------------------------------------------------------
314 #-----------------------------------------------------------------------------------
317 my ($line_in, @chunks) = @_;
319 my $ofrom = shift @chunks;
320 if ($ofrom =~ /^'/ or !@chunks) {
321 until ($ofrom =~ /'$/)
322 { $ofrom .= join(' ','',shift @chunks) }
323 $ofrom =~ s/^'//; $ofrom =~ s/'$//;
325 my $to = shift @chunks;
327 until ($to =~ /'$/ or !@chunks)
328 { $to .= join(' ','',shift @chunks) }
329 $to =~ s/^'//; $to =~ s/'$//;
333 for my $char (split(//,$ofrom)) {
334 $char = "\\" . $char if ($char =~ /\W/);
335 $from = join('', $from, $char);
338 $conf->{prevline} = $record[$ptr];
339 $record[$ptr] =~ s/$from/$to/;
345 $conf->{prevline} = $record[$ptr];
346 # remove <subfield stuff; extract (probably wrong) subfield code
347 $record[$ptr] =~ s/^\s*<subfield code="(.*?)">//;
348 # and move to front of line
349 $record[$ptr] = join(' ', $1 , $record[$ptr]);
350 # tear off trailing subfield tag from preceeding line
351 $record[$ptr - 1] =~ s|</subfield>\n||;
352 # join current line onto preceeding line
353 $record[$ptr - 1] = join('', $record[$ptr - 1], $record[$ptr]);
355 my @a = @record[0 .. $ptr - 1];
356 my @b = @record[$ptr + 1 .. $#record];
358 # move record pointer to previous line
365 my $temp = $record[$ptr];
366 $record[$ptr] = $conf->{prevline};
367 $conf->{prevline} = $temp;
373 $conf->{prevline} = $record[$ptr];
374 my @a = @record[0 .. $ptr - 1];
375 my @b = @record[$ptr + 1 .. $#record];
382 my @a = @record[0 .. $ptr - 1];
383 my @b = @record[$ptr .. $#record];
384 @record = (@a, $conf->{prevline}, @b);
391 print $OUT "\n", $conf->{origline}, "\n";
396 my ($line_in, @explanation) = @_;
397 $recmeta{explanation} = join(' ', 'Tag', $recmeta{tag}, @explanation);
398 write_record($EXMARC);
403 $ptr++ unless ($ptr == $#record);;
409 $ptr-- unless ($ptr == 0);
414 sub commit_edit { return 1 }
417 if ($conf->{window} == 15)
418 { print $OUT "Window can't be bigger than 15 lines\n"; return }
419 $conf->{window} += 2;
424 if ($conf->{window} == 5)
425 { print $OUT "Window can't be smaller than 5 lines\n"; return }
426 $conf->{window} -= 2;
433 Type a replacement for the indicated line, or enter a command.
435 Commands: c Show record context again ('C' for brief context)
436 k Kill indicated line (remove from record)
437 m Merge indicated line with previous line
439 s Substitute ARG1 for ARG2 in indicated line
440 t Commit changes and resume stream edit
441 x Write this record to the exception file instead of output
450 #-----------------------------------------------------------------------------------
452 #-----------------------------------------------------------------------------------
453 # defined a domain-specific language for specifying MARC tags to be dropped from
454 # records during processing. it is line oriented, and is specified as follows:
456 # each line may specify any number of tags to be included, either singly (\d{1,3})
457 # or as a range (\d{1,3}\.\.\d{1,3}
459 # if a single number is given, it must be between '000' and '999', inclusive.
461 # ranges obey the previous rule, and also the first number of the range must be less
462 # than the second number
464 # finally, any single range in a line may be followed by the keyword 'except'. every
465 # number or range after 'except' is excluded from the range specified. all these
466 # numbers must actually be within the range.
468 # specifying a tag twice is an error, to help prevent typos
471 print $OUT ">>> TRASHTAGS FILE FOUND. LOADING TAGS TO BE STRIPPED FROM OUTPUT\n";
472 open TRASH, '<', $conf->{trash}
473 or die "Can't open trash tags file!\n";
475 my $lastwasrange = 0;
476 my %lastrange = ( high => 0, low => 0);
479 my @chunks = split /\s+/;
480 while (my $chunk = shift @chunks) {
483 if ($chunk =~ /^\d{1,3}$/) {
484 trash_add($chunk, $except);
490 if ($chunk =~ /^\d{1,3}\.\.\d{1,3}$/) {
491 my ($low, $high) = trash_add_range($chunk, $except, \%lastrange);
493 %lastrange = (low => $low, high => $high)
499 if ($chunk eq 'except') {
500 die "Keyword 'except' can only follow a range (line $.)\n"
501 unless $lastwasrange;
502 die "Keyword 'except' may only occur once per line (line $.)\n"
508 die "Unknown chunk $chunk in .trashtags file (line $.)\n";
512 # remove original id sequence tag from trash hash if we know it
513 trash_add($conf->{'original-tag'}, 1)
514 if ($conf->{'original-tag'} and $conf->{trash}{ $conf->{'original-tag'} });
517 sub trash_add_range {
518 my ($chunk, $except, $range) = @_;
519 my ($low,$high) = split /\.\./, $chunk;
520 die "Ranges must be 'low..high' ($low is greater than $high on line $.)\n"
523 die "Exception ranges must be within last addition range (line $.)\n"
524 if ($low < $range->{low} or $high > $range->{high});
526 for my $tag ($low..$high) {
527 trash_add($tag, $except)
533 my ($tag, $except) = @_;
534 my $trash = $conf->{trash};
536 die "Trash values must be valid tags (000-999)\n"
537 unless ($tag >= 0 and $tag <= 999);
540 delete $trash->{$tag};
542 die "Trash tag '$tag' specified twice (line $.)\n"
548 #-----------------------------------------------------------------------
552 Performs boring script initialization. Handles argument parsing,
561 # set mode on existing filehandles
562 binmode(STDIN, ':utf8');
564 my $rc = GetOptions( $c,
569 'renumber-from|rf=i',
572 'renumber-subfield|rt=i',
576 show_help() unless $rc;
577 show_help() if ($c->{help});
580 $c->{output} = 'incoming.cleaned.marc.xml' unless defined $c->{output};
581 $c->{exception} = 'incoming.exception.marc.xml' unless defined $c->{exception};
582 $c->{'renumber-tag'} = 903 unless defined $c->{'renumber-tag'};
583 $c->{'renumber-subfield'} = 'a' unless defined $c->{'renumber-subfield'};
586 my @keys = keys %{$c};
587 show_help() unless (@ARGV and @keys);
588 #for my $key ('runtype', 'tag', 'subfield', 'output', 'exception')
589 # { push @missing, $key unless $c->{$key} }
591 # print "Required option: ", join(', ', @missing), " missing!\n";
598 Usage is: $0 [OPTIONS] <filelist>
600 --output -o Cleaned MARCXML output filename (default: incoming.cleaned.marc.xml)
601 --exception -x Exception (dumped records) MARCXML filename (incoming.exception.marc.xml)