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 == $#record) {
131 # catch empty datafield elements
132 if ($record[$ptr] =~ m/<datafield tag="..."/) {
133 if ($record[$ptr + 1] =~ m|</datafield>|) {
134 my @a = @record[0 .. $ptr - 1];
135 my @b = @record[$ptr + 2 .. $#record];
137 message("Empty datafield scrubbed");
142 # and quasi-empty subfields
143 if ($record[$ptr] =~ m|<subfield code="(.*?)">(.*?)</sub|) {
144 my $code = $1; my $content = $2;
145 if ($code =~ /\W/ and ($content =~ /\s+/ or $content eq '')) {
146 my @a = @record[0 .. $ptr - 1];
147 my @b = @record[$ptr + 1 .. $#record];
149 message("Empty subfield scrubbed");
158 for $ptr (0 .. $#record) {
160 if ($record[$ptr] =~ m|<leader>(.+?)</leader>|) {
162 if (length $leader < 24) {
163 $leader .= ' ' x (20 - length($leader));
165 $record[$ptr] = "<leader>$leader</leader>\n";
166 message("Short leader padded");
169 if ($record[$ptr] =~ m|<controlfield tag="008">(.+?)</control|) {
172 if (length $content < 40) {
173 $content .= ' ' x (40 - length($content));
174 $record[$ptr] = "<controlfield tag=\"008\">$content</controlfield>\n";
175 message("Short 008 padded");
179 # clean misplaced dollarsigns
180 if ($record[$ptr] =~ m|<subfield code="\$">c?\d+\.\d{2}|) {
181 $record[$ptr] =~ s|"\$">c?(\d+\.\d{2})|"c">\$$1|;
182 message("Dollar sign corrected");
185 # automatable subfield maladies
186 $record[$ptr] =~ s/code=" ">c/code="c">/;
187 $record[$ptr] =~ s/code=" ">\$/code="c">\$/;
192 sub stow_record_data {
193 # get tag data if we're looking at it
194 if ($record[$ptr] =~ m/<datafield tag="(?<TAG>.{3})"/) {
195 $recmeta{tag} = $+{TAG};
196 $record[$ptr] =~ m/ind1="(?<IND1>.)"/;
197 $recmeta{ind1} = $+{IND1} || '';
198 $record[$ptr] =~ m/ind2="(?<IND2>.)"/;
199 $recmeta{ind2} = $+{IND2} || '';
201 unless (defined $recmeta{tag}) {
202 message("Autokill record: no detectable tag");
203 dump_record("No detectable tag") ;
207 # and since we are looking at a tag, see if it's the original id
208 if ($conf->{'original-subfield'} and $recmeta{tag} == $conf->{'original-tag'}) {
209 my $line = $record[$ptr]; my $lptr = $ptr;
210 my $osub = $conf->{'original-subfield'};
211 $recmeta{oid} = 'NONE';
213 # skim to end of this tag
214 until ($line =~ m|</datafield>|) {
215 if ($line =~ /<subfield code="$osub">(.+?)</)
216 { $recmeta{oid} = $1 }
218 $line = $record[$lptr];
220 unless (defined $recmeta{oid}) {
221 message("Autokill record: no oldid when old2new mapping requested");
222 dump_record("No old id found");
230 #-----------------------------------------------------------------------------------
232 #-----------------------------------------------------------------------------------
236 Handles the Term::ReadLine loop
243 return if $conf->{trash}->has( $recmeta{tag} );
244 $conf->{editmsg} = $msg;
247 # stow original problem line
248 $recmeta{origline} = $record[$ptr];
251 my $line = $term->readline('marc-cleanup>');
252 my @chunks = split /\s+/, $line;
254 # lines with single-character first chunks are commands.
255 # make sure they exist.
256 if (length $chunks[0] == 1) {
257 unless (defined $commands{$chunks[0]}) {
258 print $OUT "No such command '", $chunks[0], "'\n";
263 if (defined $commands{$chunks[0]}) {
264 my $term = $commands{$chunks[0]}->(@chunks[1..$#chunks]);
267 $recmeta{prevline} = $record[$ptr];
268 $record[$ptr] = "$line\n";
272 # set pointer to top on the way out
278 Constructs record arrays from the incoming MARC file and returns them
286 my $trash = $conf->{trash};
288 $l = <MARC> while (defined $l and $l !~ /<record>/);
289 return $l unless defined $l;
294 until ($l =~ m|</record>|) {
295 # clean up tags with spaces in them
296 $l =~ s/tag=" /tag="00/g;
297 $l =~ s/tag=" /tag="0/g;
298 $l =~ s/tag="-/tag="0/g;
299 $l =~ s/tag="(\d\d) /tag="0$1/g;
301 # excise unwanted tags
303 if ($l =~ m|</datafield|)
306 { $l = <MARC>; next }
308 if ($l =~ m/<datafield tag="(.{3})"/) {
309 if ($trash->has($1) or ($conf->{autoscrub} and $1 =~ /\D/))
310 { $istrash = 1; next }
325 open $EXMARC, '>:utf8', $conf->{exception}
326 or die "Can't open exception file $!\n";
330 $conf->{rocount}++ if ($FH eq $NUMARC);
331 print $FH '<!-- ', $recmeta{explanation}, " -->\n"
332 if(defined $recmeta{explanation});
334 # add 903(?) with new record id
336 if ($conf->{'renumber-from'}) {
337 $recmeta{nid} = $conf->{'renumber-from'};
338 $renumber = join('', ' <datafield tag="', $conf->{'renumber-tag'},
339 '" ind1=" " ind2=" "> <subfield code="',
340 $conf->{'renumber-subfield'},
341 '">', $recmeta{nid}, "</subfield></datafield>\n");
342 my @tmp = @record[0 .. $#record - 1];
343 my $last = $record[$#record];
344 @record = (@tmp, $renumber, $last);
345 $conf->{'renumber-from'}++;
348 # scrub newlines (unless told not to or writing exception record)
349 unless ($conf->{nocollapse} or $FH eq $EXMARC)
350 { s/\n// for (@record) }
352 # write to old->new map file if needed
353 if ($conf->{'renumber-from'} and $conf->{'original-subfield'}) {
354 print $OLD2NEW $recmeta{oid}, "\t", $recmeta{nid}, "\n"
357 # actually write the record
358 print $FH @record,"\n";
360 # if we were dumping to exception file, nuke the record and set ptr
361 # to terminate processing loop
366 sub print_fullcontext {
367 print $OUT "\r", ' ' x 72, "\n";
368 print $OUT $conf->{editmsg},"\n";
369 print $OUT "\r Tag:",$recmeta{tag}, " Ind1:'",
370 $recmeta{ind1},"' Ind2:'", $recmeta{ind2}, "'";
371 print $OUT " @ ", $conf->{ricount}, "/", $conf->{rocount} + 1;
377 my $upper = int($conf->{window} / 2) + 1;
378 my $lower = int($conf->{window} / 2) - 1;
379 my $start = ($ptr - $upper < 0) ? 0 : $ptr - $upper;
380 my $stop = ($ptr + $lower > $#record) ? $#record : $ptr + $lower;
382 print $OUT ' |', $record[$_] for ($start .. $ptr - 1);
383 print $OUT '==> |', $record[$ptr];
384 print $OUT ' |', $record[$_] for ($ptr + 1 .. $stop);
391 print $OUT "\r$msg at ",$conf->{ricount},"/",$conf->{rocount} + 1,"\n";
394 #-----------------------------------------------------------------------------------
396 #-----------------------------------------------------------------------------------
401 my $ofrom = shift @chunks;
402 if ($ofrom =~ /^'/) {
403 until ($ofrom =~ /'$/ or !@chunks)
404 { $ofrom .= join(' ','',shift @chunks) }
405 $ofrom =~ s/^'//; $ofrom =~ s/'$//;
407 my $to = shift @chunks;
409 until ($to =~ /'$/ or !@chunks)
410 { $to .= join(' ','',shift @chunks) }
411 $to =~ s/^'//; $to =~ s/'$//;
415 for my $char (split(//,$ofrom)) {
416 $char = "\\" . $char if ($char =~ /\W/);
417 $from = join('', $from, $char);
420 $recmeta{prevline} = $record[$ptr];
421 $record[$ptr] =~ s/$from/$to/;
427 $recmeta{prevline} = $record[$ptr];
428 # remove <subfield stuff; extract (probably wrong) subfield code
429 $record[$ptr] =~ s/^\s*<subfield code="(.*?)">//;
430 # and move to front of line
431 $record[$ptr] = join(' ', $1 , $record[$ptr]);
432 # tear off trailing subfield tag from preceeding line
433 $record[$ptr - 1] =~ s|</subfield>\n||;
434 # join current line onto preceeding line
435 $record[$ptr - 1] = join('', $record[$ptr - 1], $record[$ptr]);
437 my @a = @record[0 .. $ptr - 1];
438 my @b = @record[$ptr + 1 .. $#record];
440 # move record pointer to previous line
447 unless ($recmeta{prevline})
448 { print $OUT "No previously edited line to flip\n"; return }
449 my $temp = $record[$ptr];
450 $record[$ptr] = $recmeta{prevline};
451 $recmeta{prevline} = $temp;
457 $recmeta{killline} = $record[$ptr];
458 my @a = @record[0 .. $ptr - 1];
459 my @b = @record[$ptr + 1 .. $#record];
466 unless ($recmeta{killline})
467 { print $OUT "No killed line to yank\n"; return }
468 my @a = @record[0 .. $ptr - 1];
469 my @b = @record[$ptr .. $#record];
470 @record = (@a, $conf->{killline}, @b);
475 sub insert_original {
476 $record[$ptr] = $recmeta{origline};
482 print $OUT "\nOrig. edit line :", $recmeta{origline};
483 print $OUT "Current flip line:", $recmeta{prevline} if $recmeta{prevline};
484 print $OUT "Last killed line :", $recmeta{killline} if $recmeta{killline};
490 my (@explanation) = @_;
491 print $OUT @explanation;
492 $recmeta{explanation} = join(' ', 'Tag', $recmeta{tag}, @explanation);
493 write_record($EXMARC);
498 $ptr++ unless ($ptr == $#record);;
504 $ptr-- unless ($ptr == 0);
509 sub commit_edit { return 1 }
512 if ($conf->{window} == 15)
513 { print $OUT "Window can't be bigger than 15 lines\n"; return }
514 $conf->{window} += 2;
519 if ($conf->{window} == 5)
520 { print $OUT "Window can't be smaller than 5 lines\n"; return }
521 $conf->{window} -= 2;
527 Type a replacement for the indicated line, or enter a command.
529 DISPLAY COMMANDS | LINE AUTO-EDIT COMMANDS
530 < Expand context window | k Kill current line
531 > Contract context window | y Yank last killed line
532 p Move pointer to prev line | m Merge current line into preceding line
533 n Move pointer to next line | o Insert original line
534 c Print line context | f Flip current line and last edited line
535 d Print current saved lines |
536 -----------------------------+-------------------------------------------
537 s Subtitute; replace ARG1 in current line with ARG2. If either ARG
538 contains spaces, it must be single-quoted
539 t Commit changes and resume automated operations
540 x Dump record to exception file
549 #-----------------------------------------------------------------------
553 Performs boring script initialization. Handles argument parsing,
562 # set mode on existing filehandles
563 binmode(STDIN, ':utf8');
565 my $rc = GetOptions( $c,
571 'renumber-from|rf=i',
573 'renumber-subfield|rs=s',
575 'original-subfield|os=s',
583 show_help() unless $rc;
584 show_help() if ($c->{help});
585 show_trashhelp() if ($c->{trashhelp});
588 my $pfx = $c->{prefix} // "bibs";
589 $c->{output} = join('.',$c->{prefix},'clean','marc','xml');
590 $c->{exception} = join('.',$c->{prefix},'exception','marc','xml');
591 $c->{'renumber-tag'} = 903 unless defined $c->{'renumber-tag'};
592 $c->{'renumber-subfield'} = 'a' unless defined $c->{'renumber-subfield'};
595 if ($c->{trashfile}) {
596 $c->{trash} = Equinox::Migration::SimpleTagList->new(file => $conf->{trashfile})
598 $c->{trash} = Equinox::Migration::SimpleTagList->new;
600 # remove original id sequence tag from trash hash if we know it
601 $c->{trash}->remove_tag($c->{'original-tag'})
602 if ( $c->{'original-tag'} and $c->{trash}->has($c->{'original-tag'}) );
604 # autotrash 901, 903 if strip-nines
605 if ($c->{'strip9'}) {
606 $c->{trash}->add_tag(901);
607 $c->{trash}->add_tag(903);
610 my @keys = keys %{$c};
611 show_help() unless (@ARGV and @keys);
616 Usage is: marc-cleanup [OPTIONS] <filelist>
618 --output -o Cleaned MARCXML output filename
619 --exception -x Exception (dumped records) MARCXML filename
621 --prefix=<PREFIX>> -p Shared prefix for output/exception files. Will produce
622 PREFIX.clean.marc.xml and PREFIX.exception.marc.xml
624 --renumber-from -rf Begin renumbering id sequence with this number
625 --renumber-tag -rt Tag to use in renumbering (default: 903)
626 --renumber-subfield -rs Subfield code to use in renumbering (default: a)
627 --original-tag -ot Original id tag; will be kept in output even if
628 it appears in the trash file
629 --original-subfield -os Original id subfield code. If this is specified
630 and renumbering is in effect, an old-to-new mapping
631 file (old2new.map) will be generated.
633 --autoscrub -a Automatically remove non-numeric tags in data
634 --nocollapse -n Don't compress records to one line on output
635 --strip9 Automatically remove any existing 901/903 tags in data
636 --no-strip9 Don't complain about 901/903 tags in data
637 --trashfile -t File containing trash tag data (see --trashhelp)
640 --script Store human-initiated ops in scriptfile (.mcscript)
650 http://intra.lan.hq.esilibrary.com/dokuwiki/doku.php?id=migration:tag_files
652 for tag file syntax information.