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-tag'});
31 print $NUMARC "<collection xmlns=\"http://www.loc.gov/MARC21/slim\">\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 $rc = stow_record_data() if ($conf->{'renumber-from'} and $conf->{'original-tag'});
83 if ($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 $conf->{autoscrub};
95 # subfields can't be non-alphanumeric
96 if ($record[$ptr] =~ /<subfield code="(.*?)"/) {
97 if ($1 =~ /\P{IsAlnum}/ or $1 eq '') {
98 edit("Junk in subfield code/Null subfield code");
102 # subfields can't be non-alphanumeric
103 if ($record[$ptr] =~ /<subfield code="(\w{2,})"/) {
104 edit("Subfield code larger than 1 char");
110 write_record($NUMARC);
112 print $NUMARC "</collection>\n";
113 print $OUT "\nDone. ",$conf->{ricount}," in; ",$conf->{rocount}," dumped \n";
116 #-----------------------------------------------------------------------------------
118 #-----------------------------------------------------------------------------------
120 sub do_automated_cleanups {
122 until ($ptr == $#record) {
124 # catch empty datafield elements
125 if ($record[$ptr] =~ m/<datafield tag="..."/) {
126 if ($record[$ptr + 1] =~ m|</datafield>|) {
127 my @a = @record[0 .. $ptr - 1];
128 my @b = @record[$ptr + 2 .. $#record];
130 @a = undef; @b = undef;
131 message("Empty datafield scrubbed");
136 # and quasi-empty subfields
137 if ($record[$ptr] =~ m|<subfield code="(.*?)">(.*?)</sub|) {
138 my $code = $1; my $content = $2;
139 if ($code =~ /\W/ and ($content =~ /\s+/ or $content eq '')) {
140 my @a = @record[0 .. $ptr - 1];
141 my @b = @record[$ptr + 1 .. $#record];
143 @a = undef; @b = undef;
144 message("Empty subfield scrubbed");
153 for $ptr (0 .. $#record) {
155 if ($record[$ptr] =~ m|<leader>(.+?)</leader>|) {
157 if (length $leader < 24) {
158 $leader .= ' ' x (20 - length($leader));
160 $record[$ptr] = "<leader>$leader</leader>\n";
161 message("Short leader padded");
164 if ($record[$ptr] =~ m|<controlfield tag="008">(.+?)</control|) {
167 if (length $content < 40) {
168 $content .= ' ' x (40 - length($content));
169 $record[$ptr] = "<controlfield tag=\"008\">$content</controlfield>\n";
170 message("Short 008 padded");
174 # clean misplaced dollarsigns
175 if ($record[$ptr] =~ m|<subfield code="\$">c?\d+\.\d{2}|) {
176 $record[$ptr] =~ s|"\$">c?(\d+\.\d{2})|"c">\$$1|;
177 message("Dollar sign corrected");
180 # automatable subfield maladies
181 $record[$ptr] =~ s/code=" ">c/code="c">/;
182 $record[$ptr] =~ s/code=" ">\$/code="c">\$/;
187 sub stow_record_data {
188 # get tag data if we're looking at it
190 if ($record[$ptr] =~ m/<(control|data)field tag="(?<TAG>.{3})"/) {
191 $recmeta{tag} = $+{TAG};
192 $tag = $recmeta{tag};
193 $record[$ptr] =~ m/ind1="(?<IND1>.)"/;
194 $recmeta{ind1} = $+{IND1} || '';
195 $record[$ptr] =~ m/ind2="(?<IND2>.)"/;
196 $recmeta{ind2} = $+{IND2} || '';
199 message("Autokill record: no detectable tag");
200 dump_record("No detectable tag") ;
204 # and since we are looking at a tag, see if it's the original id
205 if ($tag == $conf->{'original-tag'}) {
209 if ($record[$ptr] =~ m|<controlfield tag="$tag">(.+?)</controlfield>|)
210 { $oid = $1; print $OLD2NEW "$oid\t", $recmeta{nid}, "\n" }
211 } elsif ($tag >= 10 and $conf->{'original-subfield'}) {
213 my $line = $record[$ptr]; my $lptr = $ptr;
214 my $osub = $conf->{'original-subfield'};
215 # skim to end of this tag
216 until ($line =~ m|</datafield>|) {
217 if ($line =~ /<subfield code="$osub">(.+?)</)
218 { $oid = $1; print $OLD2NEW "$oid\t", $recmeta{nid}, "\n" }
220 $line = $record[$lptr];
226 # didn't find the old id!
228 message("Autokill record: no oldid when old2new mapping requested");
229 dump_record("No old id found");
233 # got it; write to old->new map file
234 if ($conf->{'renumber-from'} and $conf->{'original-subfield'}) {
242 #-----------------------------------------------------------------------------------
244 #-----------------------------------------------------------------------------------
248 Handles the Term::ReadLine loop
255 return if $conf->{trash}->has( $recmeta{tag} );
256 if ( $conf->{fullauto} )
257 { dump_record($msg); return }
259 $conf->{editmsg} = $msg;
262 # stow original problem line
263 $recmeta{origline} = $record[$ptr];
266 my $line = $term->readline('marc-cleanup>');
267 my @chunks = split /\s+/, $line;
269 # lines with single-character first chunks are commands.
270 # make sure they exist.
271 if (length $chunks[0] == 1) {
272 unless (defined $commands{$chunks[0]}) {
273 print $OUT "No such command '", $chunks[0], "'\n";
278 if (defined $commands{$chunks[0]}) {
279 my $term = $commands{$chunks[0]}->(@chunks[1..$#chunks]);
282 $recmeta{prevline} = $record[$ptr];
283 $record[$ptr] = "$line\n";
287 # set pointer to top on the way out
293 Constructs record arrays from the incoming MARC file and returns them
301 my $trash = $conf->{trash};
303 $l = <MARC> while (defined $l and $l !~ /<record>/);
304 return $l unless defined $l;
307 for (keys %recmeta) { $recmeta{$_} = undef }
308 for (0 .. @record) { delete $record[$_] }
311 until ($l =~ m|</record>|) {
312 # clean up tags with spaces in them
313 $l =~ s/tag=" /tag="00/g;
314 $l =~ s/tag=" /tag="0/g;
315 $l =~ s/tag="-/tag="0/g;
316 $l =~ s/tag="(\d\d) /tag="0$1/g;
318 # excise unwanted tags
320 $istrash = 0 if ($l =~ m|</datafield|);
324 if ($l =~ m/<datafield tag="(.{3})"/) {
325 if ($trash->has($1) or ($conf->{autoscrub} and $1 =~ /\D/))
326 { $istrash = 1; next }
334 # add 903(?) with new record id
335 if ($conf->{'renumber-from'}) {
336 $recmeta{nid} = $conf->{'renumber-from'};
337 push @record, join('', ' <datafield tag="', $conf->{'renumber-tag'},
338 '" ind1=" " ind2=" "> <subfield code="',
339 $conf->{'renumber-subfield'},
342 "</subfield></datafield>\n");
343 $conf->{'renumber-from'}++;
356 open $EXMARC, '>:utf8', $conf->{exception}
357 or die "Can't open exception file $!\n";
361 print $FH '<!-- ', $recmeta{explanation}, " -->\n"
362 if(defined $recmeta{explanation});
364 # scrub newlines (unless told not to or writing exception record)
365 unless ($conf->{nocollapse} or $FH eq $EXMARC)
366 { s/\n// for (@record) }
368 # actually write the record
369 print $FH @record,"\n";
371 # increment output record count (if not exception)
372 $conf->{rocount}++ if ($FH eq $EXMARC);
374 # if we were dumping to exception file, nuke the record and set ptr
375 # to terminate processing loop
380 sub print_fullcontext {
381 print $OUT "\r", ' ' x 72, "\n";
382 print $OUT $conf->{editmsg},"\n";
383 print $OUT "\r Tag:",$recmeta{tag}, " Ind1:'",
384 $recmeta{ind1},"' Ind2:'", $recmeta{ind2}, "'";
385 print $OUT " @ ", $conf->{ricount}, "/", $conf->{totalrecs};
391 my $upper = int($conf->{window} / 2) + 1;
392 my $lower = int($conf->{window} / 2) - 1;
393 my $start = ($ptr - $upper < 0) ? 0 : $ptr - $upper;
394 my $stop = ($ptr + $lower > $#record) ? $#record : $ptr + $lower;
396 print $OUT ' |', $record[$_] for ($start .. $ptr - 1);
397 print $OUT '==> |', $record[$ptr];
398 print $OUT ' |', $record[$_] for ($ptr + 1 .. $stop);
405 print $OUT "\r$msg at ",$conf->{ricount},"/",$conf->{totalrecs}, "\n";
408 #-----------------------------------------------------------------------------------
410 #-----------------------------------------------------------------------------------
415 my $ofrom = shift @chunks;
416 if ($ofrom =~ /^'/) {
417 until ($ofrom =~ /'$/ or !@chunks)
418 { $ofrom .= join(' ','',shift @chunks) }
419 $ofrom =~ s/^'//; $ofrom =~ s/'$//;
421 my $to = shift @chunks;
423 until ($to =~ /'$/ or !@chunks)
424 { $to .= join(' ','',shift @chunks) }
425 $to =~ s/^'//; $to =~ s/'$//;
429 for my $char (split(//,$ofrom)) {
430 $char = "\\" . $char if ($char =~ /\W/);
431 $from = join('', $from, $char);
434 $recmeta{prevline} = $record[$ptr];
435 $record[$ptr] =~ s/$from/$to/;
436 $ofrom = undef; $to = undef; $from = undef;
442 $recmeta{prevline} = $record[$ptr];
443 # remove <subfield stuff; extract (probably wrong) subfield code
444 $record[$ptr] =~ s/^\s*<subfield code="(.*?)">//;
445 # and move to front of line
446 $record[$ptr] = join(' ', $1 , $record[$ptr]);
447 # tear off trailing subfield tag from preceeding line
448 $record[$ptr - 1] =~ s|</subfield>\n||;
449 # join current line onto preceeding line
450 $record[$ptr - 1] = join('', $record[$ptr - 1], $record[$ptr]);
452 my @a = @record[0 .. $ptr - 1];
453 my @b = @record[$ptr + 1 .. $#record];
455 # move record pointer to previous line
462 unless ($recmeta{prevline})
463 { print $OUT "No previously edited line to flip\n"; return }
464 my $temp = $record[$ptr];
465 $record[$ptr] = $recmeta{prevline};
466 $recmeta{prevline} = $temp;
473 $recmeta{killline} = $record[$ptr];
474 my @a = @record[0 .. $ptr - 1];
475 my @b = @record[$ptr + 1 .. $#record];
477 @a = undef; @b = undef;
483 unless ($recmeta{killline})
484 { print $OUT "No killed line to yank\n"; return }
485 my @a = @record[0 .. $ptr - 1];
486 my @b = @record[$ptr .. $#record];
487 @record = (@a, $conf->{killline}, @b);
488 @a = undef; @b = undef;
493 sub insert_original {
494 $record[$ptr] = $recmeta{origline};
500 print $OUT "\nOrig. edit line :", $recmeta{origline};
501 print $OUT "Current flip line:", $recmeta{prevline} if $recmeta{prevline};
502 print $OUT "Last killed line :", $recmeta{killline} if $recmeta{killline};
508 my (@explanation) = @_;
509 $recmeta{explanation} = join(' ', 'DUMPING RECORD: Tag', $recmeta{tag}, @explanation);
510 message( $recmeta{explanation} );
511 write_record($EXMARC);
516 $ptr++ unless ($ptr == $#record);;
522 $ptr-- unless ($ptr == 0);
527 sub commit_edit { return 1 }
530 if ($conf->{window} == 15)
531 { print $OUT "Window can't be bigger than 15 lines\n"; return }
532 $conf->{window} += 2;
537 if ($conf->{window} == 5)
538 { print $OUT "Window can't be smaller than 5 lines\n"; return }
539 $conf->{window} -= 2;
545 Type a replacement for the indicated line, or enter a command.
547 DISPLAY COMMANDS | LINE AUTO-EDIT COMMANDS
548 < Expand context window | k Kill current line
549 > Contract context window | y Yank last killed line
550 p Move pointer to prev line | m Merge current line into preceding line
551 n Move pointer to next line | o Insert original line
552 c Print line context | f Flip current line and last edited line
553 d Print current saved lines |
554 -----------------------------+-------------------------------------------
555 s Subtitute; replace ARG1 in current line with ARG2. If either ARG
556 contains spaces, it must be single-quoted
557 t Commit changes and resume automated operations
558 x Dump record to exception file
567 #-----------------------------------------------------------------------
571 Performs boring script initialization. Handles argument parsing,
580 # set mode on existing filehandles
581 binmode(STDIN, ':utf8');
583 my $rc = GetOptions( $c,
590 'renumber-from|rf=i',
592 'renumber-subfield|rs=s',
594 'original-subfield|os=s',
601 show_help() unless $rc and @ARGV;
602 show_help() if ($c->{help});
603 show_trashhelp() if ($c->{trashhelp});
606 my $pfx = $c->{prefix} // "bibs";
609 $c->{output} = join('.',$c->{prefix},'clean','marc','xml');
610 $c->{exception} = join('.',$c->{prefix},'exception','marc','xml');
611 $c->{'renumber-tag'} = 903 unless defined $c->{'renumber-tag'};
612 $c->{'renumber-subfield'} = 'a' unless defined $c->{'renumber-subfield'};
615 if ($c->{trashfile}) {
616 $c->{trash} = Equinox::Migration::SimpleTagList->new(file => $conf->{trashfile})
618 $c->{trash} = Equinox::Migration::SimpleTagList->new;
620 # autotrash 901, 903 unless no strip-nines
621 unless ($c->{'no-strip9'}) {
622 $c->{trash}->add_tag(901);
623 $c->{trash}->add_tag(903);
625 # remove original id sequence tag from trash hash if we know it
626 $c->{trash}->remove_tag($c->{'original-tag'})
627 if ( $c->{'original-tag'} and $c->{trash}->has($c->{'original-tag'}) );
629 my @keys = keys %{$c};
630 show_help() unless (@ARGV and @keys);
635 Usage is: marc-cleanup [OPTIONS] <filelist>
637 --output -o Cleaned MARCXML output filename
638 --exception -x Exception (dumped records) MARCXML filename
640 --prefix=<PREFIX>> -p Shared prefix for output/exception files. Will produce
641 PREFIX.clean.marc.xml and PREFIX.exception.marc.xml
643 --renumber-from -rf Begin renumbering id sequence with this number
644 --renumber-tag -rt Tag to use in renumbering (default: 903)
645 --renumber-subfield -rs Subfield code to use in renumbering (default: a)
646 --original-tag -ot Original id tag; will be kept in output even if
647 it appears in the trash file
648 --original-subfield -os Original id subfield code. If this is specified
649 and renumbering is in effect, an old-to-new mapping
650 file (old2new.map) will be generated.
652 --autoscrub -a Automatically remove non-numeric tags in data
653 --nocollapse -n Don't compress records to one line on output
654 --no-strip9 Don't autoremove 901/903 tags in data
655 --trashfile -t File containing trash tag data (see --trashhelp)
657 --fullauto No manual edits. All problematic records dumped to
660 --script Store human-initiated ops in scriptfile (.mcscript)
670 http://intra.lan.hq.esilibrary.com/dokuwiki/doku.php?id=migration:tag_files
672 for tag file syntax information.