9 use Equinox::Migration::SimpleTagList;
11 my $term = new Term::ReadLine 'yaz-cleanup';
12 my $OUT = $term->OUT || \*STDOUT;
13 binmode STDOUT, ":utf8";
14 binmode $OUT, ":utf8";
18 # initialization and setup
22 # set up files, since everything appears to be in order
23 my $marcfile = shift || 'incoming.marc.xml';
24 open MARC, '<:utf8', $marcfile
25 or die "Can't open input file $!\n";
26 open my $NUMARC, '>:utf8', $conf->{output}
27 or die "Can't open output file $!\n";
28 open my $OLD2NEW, '>', 'old2new.map'
29 if ($conf->{'renumber-from'} and $conf->{'original-subfield'});
31 print $NUMARC "<collection>\n";
33 $conf->{totalrecs} = `grep -c '<record' $marcfile`;
34 chomp $conf->{totalrecs};
37 my @record = (); # current record storage
38 my %recmeta = (); # metadata about current record
39 my $ptr = 0; # record index pointer
41 # this is the dispatch table which drives command selection in
43 my %commands = ( c => \&print_fullcontext,
46 '<' => \&widen_window,
47 '>' => \&narrow_window,
49 o => \&insert_original,
63 my @spinner = qw(- \\ | /);
66 while ( buildrecord() ) {
67 unless ($conf->{ricount} % 50) {
68 $conf->{percent} = int(($conf->{ricount} / $conf->{totalrecs}) * 100);
69 print "\rWorking (",$conf->{percent},"%) ", $spinner[$sidx];
70 $sidx = ($sidx == $#spinner) ? 0 : $sidx + 1;
73 my $rc = do_automated_cleanups();
77 until ($ptr == $#record) {
78 # get datafield/tag data if we have it
79 my $rc = stow_record_data();
83 if ($record[$ptr] =~ /&/ && $record[$ptr] !~ /&\w+?;/)
84 { edit("Naked ampersand"); $ptr= 0; next }
86 if ($record[$ptr] =~ /<datafield tag="(.+?)"/) {
88 # tags must be numeric
90 edit("Non-numerics in tag") unless $conf->{autoscrub};
93 # test for existing 901/903 unless we're autocleaning them
94 unless ($conf->{'strip9'} or $conf->{'no-strip9'}) {
95 if ($match == 901 or $match == 903) {
96 edit("Incoming 901/903 found in data");
102 # subfields can't be non-alphanumeric
103 if ($record[$ptr] =~ /<subfield code="(.*?)"/) {
104 if ($1 =~ /\P{IsAlnum}/ or $1 eq '') {
105 edit("Junk in subfield code/Null subfield code");
109 # subfields can't be non-alphanumeric
110 if ($record[$ptr] =~ /<subfield code="(\w{2,})"/) {
111 edit("Subfield code larger than 1 char");
117 write_record($NUMARC);
119 print $NUMARC "</collection>\n";
120 print $OUT "\nDone. \n";
123 #-----------------------------------------------------------------------------------
125 #-----------------------------------------------------------------------------------
127 sub do_automated_cleanups {
129 until ($ptr == $#record) {
130 # catch empty datafield elements
131 if ($record[$ptr] =~ m/<datafield tag="..."/) {
132 if ($record[$ptr + 1] =~ m|</datafield>|) {
133 my @a = @record[0 .. $ptr - 1];
134 my @b = @record[$ptr + 2 .. $#record];
136 message("Empty datafield scrubbed");
141 # and quasi-empty subfields
142 if ($record[$ptr] =~ m|<subfield code="(.*?)">(.*?)</sub|) {
143 my $code = $1; my $content = $2;
144 if ($code =~ /\W/ and ($content =~ /\s+/ or $content eq '')) {
145 my @a = @record[0 .. $ptr - 1];
146 my @b = @record[$ptr + 1 .. $#record];
148 message("Empty subfield scrubbed");
157 for $ptr (0 .. $#record) {
159 if ($record[$ptr] =~ m|<leader>(.+?)</leader>|) {
161 if (length $leader < 24) {
162 $leader .= ' ' x (20 - length($leader));
164 $record[$ptr] = "<leader>$leader</leader>\n";
165 message("Short leader padded");
168 if ($record[$ptr] =~ m|<controlfield tag="008">(.+?)</control|) {
171 if (length $content < 40) {
172 $content .= ' ' x (40 - length($content));
173 $record[$ptr] = "<controlfield tag=\"008\">$content</controlfield>\n";
174 message("Short 008 padded");
178 # clean misplaced dollarsigns
179 if ($record[$ptr] =~ m|<subfield code="\$">c?\d+\.\d{2}|) {
180 $record[$ptr] =~ s|"\$">c?(\d+\.\d{2})|"c">\$$1|;
181 message("Dollar sign corrected");
184 # clean up tags with spaces in them
185 $record[$ptr] =~ s/tag=" /tag="00/g;
186 $record[$ptr] =~ s/tag=" /tag="0/g;
187 $record[$ptr] =~ s/tag="-/tag="0/g;
188 $record[$ptr] =~ s/tag="(\d\d) /tag="0$1/g;
190 # automatable subfield maladies
191 $record[$ptr] =~ s/code=" ">c/code="c">/;
192 $record[$ptr] =~ s/code=" ">\$/code="c">\$/;
197 sub stow_record_data {
198 # get tag data if we're looking at it
199 if ($record[$ptr] =~ m/<datafield tag="(?<TAG>.{3})"/) {
200 $recmeta{tag} = $+{TAG};
201 $record[$ptr] =~ m/ind1="(?<IND1>.)"/;
202 $recmeta{ind1} = $+{IND1} || '';
203 $record[$ptr] =~ m/ind2="(?<IND2>.)"/;
204 $recmeta{ind2} = $+{IND2} || '';
206 unless (defined $recmeta{tag}) {
207 message("Autokill record: no detectable tag");
208 dump_record("No detectable tag") ;
212 # and since we are looking at a tag, see if it's the original id
213 if ($conf->{'original-subfield'} and $recmeta{tag} == $conf->{'original-tag'}) {
214 my $line = $record[$ptr]; my $lptr = $ptr;
215 my $osub = $conf->{'original-subfield'};
216 $recmeta{oid} = 'NONE';
218 # skim to end of this tag
219 until ($line =~ m|</datafield>|) {
220 if ($line =~ /<subfield code="$osub">(.+?)</)
221 { $recmeta{oid} = $1 }
223 $line = $record[$lptr];
225 unless (defined $recmeta{oid}) {
226 message("Autokill record: no oldid when old2new mapping requested");
227 dump_record("No old id found");
235 #-----------------------------------------------------------------------------------
237 #-----------------------------------------------------------------------------------
241 Handles the Term::ReadLine loop
248 return if $conf->{trash}->has( $recmeta{tag} );
249 $conf->{editmsg} = $msg;
252 # stow original problem line
253 $recmeta{origline} = $record[$ptr];
256 my $line = $term->readline('marc-cleanup>');
257 my @chunks = split /\s+/, $line;
259 # lines with single-character first chunks are commands.
260 # make sure they exist.
261 if (length $chunks[0] == 1) {
262 unless (defined $commands{$chunks[0]}) {
263 print $OUT "No such command '", $chunks[0], "'\n";
268 if (defined $commands{$chunks[0]}) {
269 my $term = $commands{$chunks[0]}->(@chunks[1..$#chunks]);
272 $recmeta{prevline} = $record[$ptr];
273 $record[$ptr] = "$line\n";
277 # set pointer to top on the way out
283 Constructs record arrays from the incoming MARC file and returns them
290 $l = <MARC> while (defined $l and $l !~ /<record>/);
291 return $l unless defined $l;
296 until ($l =~ m|</record>|)
297 { push @record, $l; $l = <MARC>; }
304 my $trash = $conf->{trash};
308 open $EXMARC, '>:utf8', $conf->{exception}
309 or die "Can't open exception file $!\n";
313 $conf->{rocount}++ if ($FH eq $NUMARC);
314 print $FH '<!-- ', $recmeta{explanation}, " -->\n"
315 if(defined $recmeta{explanation});
317 # excise unwanted tags
318 if (defined $trash or $conf->{autoscrub}) {
321 for my $line (@record) {
323 $istrash = 0 if $line =~ m|</datafield|;
326 if ($line =~ m/<datafield tag="(.{3})"/) {
328 if ($trash->has($tag) or ($conf->{autoscrub} and $tag =~ /\D/)) {
333 push @trimmed, $line;
338 # add 903(?) with new record id
340 if ($conf->{'renumber-from'}) {
341 $recmeta{nid} = $conf->{'renumber-from'};
342 $renumber = join('', ' <datafield tag="', $conf->{'renumber-tag'},
343 '" ind1=" " ind2=" "> <subfield code="',
344 $conf->{'renumber-subfield'},
345 '">', $recmeta{nid}, "</subfield></datafield>\n");
346 my @tmp = @record[0 .. $#record - 1];
347 my $last = $record[$#record];
348 @record = (@tmp, $renumber, $last);
349 $conf->{'renumber-from'}++;
352 # scrub newlines (unless told not to or writing exception record)
353 unless ($conf->{nocollapse} or $FH eq $EXMARC)
354 { s/\n// for (@record) }
356 # write to old->new map file if needed
357 if ($conf->{'renumber-from'} and $conf->{'original-subfield'}) {
358 print $OLD2NEW $recmeta{oid}, "\t", $recmeta{nid}, "\n"
361 # actually write the record
362 print $FH @record,"\n";
364 # if we were dumping to exception file, nuke the record and set ptr
365 # to terminate processing loop
370 sub print_fullcontext {
371 print $OUT "\r", ' ' x 72, "\n";
372 print $OUT $conf->{editmsg},"\n";
373 print $OUT "\r Tag:",$recmeta{tag}, " Ind1:'",
374 $recmeta{ind1},"' Ind2:'", $recmeta{ind2}, "'";
375 print $OUT " @ ", $conf->{ricount}, "/", $conf->{rocount} + 1;
381 my $upper = int($conf->{window} / 2) + 1;
382 my $lower = int($conf->{window} / 2) - 1;
383 my $start = ($ptr - $upper < 0) ? 0 : $ptr - $upper;
384 my $stop = ($ptr + $lower > $#record) ? $#record : $ptr + $lower;
386 print $OUT ' |', $record[$_] for ($start .. $ptr - 1);
387 print $OUT '==> |', $record[$ptr];
388 print $OUT ' |', $record[$_] for ($ptr + 1 .. $stop);
395 print $OUT "\r$msg at ",$conf->{ricount},"/",$conf->{rocount} + 1,"\n";
398 #-----------------------------------------------------------------------------------
400 #-----------------------------------------------------------------------------------
405 my $ofrom = shift @chunks;
406 if ($ofrom =~ /^'/) {
407 until ($ofrom =~ /'$/ or !@chunks)
408 { $ofrom .= join(' ','',shift @chunks) }
409 $ofrom =~ s/^'//; $ofrom =~ s/'$//;
411 my $to = shift @chunks;
413 until ($to =~ /'$/ or !@chunks)
414 { $to .= join(' ','',shift @chunks) }
415 $to =~ s/^'//; $to =~ s/'$//;
419 for my $char (split(//,$ofrom)) {
420 $char = "\\" . $char if ($char =~ /\W/);
421 $from = join('', $from, $char);
424 $recmeta{prevline} = $record[$ptr];
425 $record[$ptr] =~ s/$from/$to/;
431 $recmeta{prevline} = $record[$ptr];
432 # remove <subfield stuff; extract (probably wrong) subfield code
433 $record[$ptr] =~ s/^\s*<subfield code="(.*?)">//;
434 # and move to front of line
435 $record[$ptr] = join(' ', $1 , $record[$ptr]);
436 # tear off trailing subfield tag from preceeding line
437 $record[$ptr - 1] =~ s|</subfield>\n||;
438 # join current line onto preceeding line
439 $record[$ptr - 1] = join('', $record[$ptr - 1], $record[$ptr]);
441 my @a = @record[0 .. $ptr - 1];
442 my @b = @record[$ptr + 1 .. $#record];
444 # move record pointer to previous line
451 unless ($recmeta{prevline})
452 { print $OUT "No previously edited line to flip\n"; return }
453 my $temp = $record[$ptr];
454 $record[$ptr] = $recmeta{prevline};
455 $recmeta{prevline} = $temp;
461 $recmeta{killline} = $record[$ptr];
462 my @a = @record[0 .. $ptr - 1];
463 my @b = @record[$ptr + 1 .. $#record];
470 unless ($recmeta{killline})
471 { print $OUT "No killed line to yank\n"; return }
472 my @a = @record[0 .. $ptr - 1];
473 my @b = @record[$ptr .. $#record];
474 @record = (@a, $conf->{killline}, @b);
479 sub insert_original {
480 $record[$ptr] = $recmeta{origline};
486 print $OUT "\nOrig. edit line :", $recmeta{origline};
487 print $OUT "Current flip line:", $recmeta{prevline} if $recmeta{prevline};
488 print $OUT "Last killed line :", $recmeta{killline} if $recmeta{killline};
494 my (@explanation) = @_;
495 print $OUT @explanation;
496 $recmeta{explanation} = join(' ', 'Tag', $recmeta{tag}, @explanation);
497 write_record($EXMARC);
502 $ptr++ unless ($ptr == $#record);;
508 $ptr-- unless ($ptr == 0);
513 sub commit_edit { return 1 }
516 if ($conf->{window} == 15)
517 { print $OUT "Window can't be bigger than 15 lines\n"; return }
518 $conf->{window} += 2;
523 if ($conf->{window} == 5)
524 { print $OUT "Window can't be smaller than 5 lines\n"; return }
525 $conf->{window} -= 2;
531 Type a replacement for the indicated line, or enter a command.
533 DISPLAY COMMANDS | LINE AUTO-EDIT COMMANDS
534 < Expand context window | k Kill current line
535 > Contract context window | y Yank last killed line
536 p Move pointer to prev line | m Merge current line into preceding line
537 n Move pointer to next line | o Insert original line
538 c Print line context | f Flip current line and last edited line
539 d Print current saved lines |
540 -----------------------------+-------------------------------------------
541 s Subtitute; replace ARG1 in current line with ARG2. If either ARG
542 contains spaces, it must be single-quoted
543 t Commit changes and resume automated operations
544 x Dump record to exception file
553 #-----------------------------------------------------------------------
557 Performs boring script initialization. Handles argument parsing,
566 # set mode on existing filehandles
567 binmode(STDIN, ':utf8');
569 my $rc = GetOptions( $c,
575 'renumber-from|rf=i',
577 'renumber-subfield|rs=s',
579 'original-subfield|os=s',
587 show_help() unless $rc;
588 show_help() if ($c->{help});
589 show_trashhelp() if ($c->{trashhelp});
592 my $pfx = $c->{prefix} // "bibs";
593 $c->{output} = join('.',$c->{prefix},'clean','marc','xml');
594 $c->{exception} = join('.',$c->{prefix},'exception','marc','xml');
595 $c->{'renumber-tag'} = 903 unless defined $c->{'renumber-tag'};
596 $c->{'renumber-subfield'} = 'a' unless defined $c->{'renumber-subfield'};
599 if ($c->{trashfile}) {
600 $c->{trash} = Equinox::Migration::SimpleTagList->new($conf->{trashfile})
602 $c->{trash} = Equinox::Migration::SimpleTagList->new;
604 # remove original id sequence tag from trash hash if we know it
605 $c->{trash}->remove_tag($c->{'original-tag'})
606 if ( $c->{'original-tag'} and $c->{trash}->has($c->{'original-tag'}) );
608 # autotrash 901, 903 if strip-nines
609 if ($c->{'strip9'}) {
610 $c->{trash}->add_tag(901);
611 $c->{trash}->add_tag(903);
614 my @keys = keys %{$c};
615 show_help() unless (@ARGV and @keys);
620 Usage is: marc-cleanup [OPTIONS] <filelist>
622 --output -o Cleaned MARCXML output filename
623 --exception -x Exception (dumped records) MARCXML filename
625 --prefix=<PREFIX>> -p Shared prefix for output/exception files. Will produce
626 PREFIX.clean.marc.xml and PREFIX.exception.marc.xml
628 --renumber-from -rf Begin renumbering id sequence with this number
629 --renumber-tag -rt Tag to use in renumbering (default: 903)
630 --renumber-subfield -rs Subfield code to use in renumbering (default: a)
631 --original-tag -ot Original id tag; will be kept in output even if
632 it appears in the trash file
633 --original-subfield -os Original id subfield code. If this is specified
634 and renumbering is in effect, an old-to-new mapping
635 file (old2new.map) will be generated.
637 --autoscrub -a Automatically remove non-numeric tags in data
638 --nocollapse -n Don't compress records to one line on output
639 --strip9 Automatically remove any existing 901/903 tags in data
640 --no-strip9 Don't complain about 901/903 tags in data
641 --trashfile -t File containing trash tag data (see --trashhelp)
644 --script Store human-initiated ops in scriptfile (.mcscript)
654 http://intra.lan.hq.esilibrary.com/dokuwiki/doku.php?id=migration:tag_files
656 for tag file syntax information.