8 use Unicode::Normalize;
9 use MARC::File::XML ( BinaryEncoding => 'utf-8' );
11 my $conf = {}; # configuration hashref
12 my $count = 0; my $scount = 0;
18 open OF, '>', $conf->{output};
20 open XF, '>', $conf->{exception};
23 for my $file (@ARGV) {
24 print XF "Processing $file\n";
25 #open my $records, '<:utf8', $file; # This hack dosn't play nice with the use MARC::File::XML ( BinaryEncoding => 'utf-8' ); hack
26 my $batch = undef; my $record = undef;
28 #$batch = MARC::Batch->new('XML', $records); # The other part of the hack
29 $batch = MARC::Batch->new($conf->{marctype}, $file);
31 $batch->warnings_off();
33 while ( $record = $batch->next ) {
34 $count++; progress_ticker();
36 unless ( defined $record )
37 { dump_exception($marc); next; }
39 my $id = $record->field($conf->{tag});
41 print XF "ERROR: Record $count in $file is missing a ",
42 $conf->{tag}, " field.\n", $record->as_formatted(), "\n=====\n";
46 $marc = populate_marc($record, $id);
47 $marc = normalize_marc($marc);
48 unless (marc_isvalid($marc))
49 { dump_exception($marc); next; }
50 dump_fingerprints($marc);
51 $scount++; progress_ticker();
55 print "\nSuccessfully processed:\t$count\n" unless $conf->{quiet};
59 Constructs a hash containing the relevant MARC data for a record and
60 returns a reference to it.
65 my ($record, $id) = @_;
66 my %marc = (); $marc{isbns} = [];
69 $marc{id} = $id->as_string($conf->{subfield});
71 # record_type, bib_lvl
72 $marc{record_type} = substr($record->leader, 6, 1);
73 $marc{bib_lvl} = substr($record->leader, 7, 1);
76 my $my_008 = $record->field('008');
77 $my_008 = $my_008->as_string() if ($my_008);
78 if (defined $my_008) {
79 unless (length $my_008 == 40) {
80 $my_008 = $my_008 . (' ' x (40 - length($my_008)));
81 print XF ">> Short 008 padded to ",length($my_008)," at rec $count\n";
83 $marc{date1} = substr($my_008,7,4) if ($my_008);
84 $marc{date2} = substr($my_008,11,4) if ($my_008); # UNUSED
86 unless ($marc{date1} and $marc{date1} =~ /\d{4}/) {
87 my $my_260 = $record->field('260');
88 my $date1 = $my_260->subfield('c') if $my_260;
89 if (defined $date1 and $date1 =~ /\d{4}/) {
90 $marc{date1} = $date1;
91 print XF ">> using 260c as date1 at rec $count\n";
96 if ( $marc{record_type} =~ /[gkroef]/ ) { # MAP, VIS
97 $marc{item_form} = substr($my_008,29,1) if ($my_008);
99 $marc{item_form} = substr($my_008,23,1) if ($my_008);
103 my @isbns = $record->field('020') if $record->field('020');
104 push @isbns, $record->field('024') if $record->field('024');
105 for my $f ( @isbns ) {
106 push @{ $marc{isbns} }, $1 if ( defined $f->subfield('a') and
107 $f->subfield('a')=~/(\S+)/ );
111 for my $rec_field (100, 110, 111) {
112 if ($record->field($rec_field)) {
113 $marc{author} = $record->field($rec_field)->subfield('a');
120 push @{ $marc{oclc} }, $record->field('001')->as_string()
121 if ($record->field('001') and $record->field('003') and
122 $record->field('003')->as_string() =~ /OCo{0,1}LC/);
123 for ($record->field('035')) {
124 my $oclc = $_->subfield('a');
125 push @{ $marc{oclc} }, $oclc
126 if (defined $oclc and $oclc =~ /\(OCoLC\)/ and $oclc =~/([0-9]+)/);
129 # "Accompanying material" (300e)
130 $marc{accomp} = $record->field('300')->subfield('e')
131 if $record->field('300');
133 # issn, lccn, title, desc, pages, pub, pubyear, edition
134 $marc{lccn} = $record->field('010')->subfield('a') if $record->field('010');
135 $marc{issn} = $record->field('022')->subfield('a') if $record->field('022');
136 $marc{desc} = $record->field('300')->subfield('a') if $record->field('300');
137 $marc{pages} = $1 if (defined $marc{desc} and $marc{desc} =~ /(\d+)/);
138 $marc{title} = $record->field('245')->subfield('a')
139 if $record->field('245');
140 $marc{edition} = $record->field('250')->subfield('a')
141 if $record->field('250');
142 if ($record->field('260')) {
143 $marc{publisher} = $record->field('260')->subfield('b');
144 $marc{pubyear} = $record->field('260')->subfield('c');
146 (defined $marc{pubyear} and $marc{pubyear} =~ /(\d{4})/) ? $1 : '';
153 =head2 normalize_marc
155 Gently massages your data.
162 $marc->{record_type }= 'a' if ($marc->{record_type} eq ' ');
163 if ($marc->{title}) {
164 $marc->{title} = NFD($marc->{title});
165 $marc->{title} =~ s/[\x{80}-\x{ffff}]//go;
166 $marc->{title} = lc($marc->{title});
167 $marc->{title} =~ s/\W+$//go;
169 if ($marc->{author}) {
170 $marc->{author} = NFD($marc->{author});
171 $marc->{author} =~ s/[\x{80}-\x{ffff}]//go;
172 $marc->{author} = lc($marc->{author});
173 $marc->{author} =~ s/\W+$//go;
174 if ($marc->{author} =~ /^(\w+)/) {
175 $marc->{author} = $1;
178 if ($marc->{publisher}) {
179 $marc->{publisher} = NFD($marc->{publisher});
180 $marc->{publisher} =~ s/[\x{80}-\x{ffff}]//go;
181 $marc->{publisher} = lc($marc->{publisher});
182 $marc->{publisher} =~ s/\W+$//go;
183 if ($marc->{publisher} =~ /^(\w+)/) {
184 $marc->{publisher} = $1;
194 Checks MARC record to see if neccessary fingerprinting data is
201 return 1 if ($marc->{item_form} and ($marc->{date1} =~ /\d{4}/) and
202 $marc->{record_type} and $marc->{bib_lvl} and $marc->{title});
207 =head2 dump_fingerprints
211 sub dump_fingerprints {
215 if ($conf->{runtype} eq "primary") {
216 print OF join("\t",$marc->{id}, $marc->{item_form},
217 $marc->{date1}, $marc->{record_type},
218 $marc->{bib_lvl}, $marc->{title}), "\n";
220 if ((scalar @{ $marc->{isbns} } > 0) and $marc->{pages}) {
221 # case a : isbn and pages
222 foreach my $isbn ( @{ $marc->{isbns}} ) {
223 print OF join("\t", $marc->{id}, "case a",
224 $marc->{item_form}, $marc->{date1},
225 $marc->{record_type},
226 $marc->{bib_lvl}, $marc->{title},
227 $isbn, $marc->{pages}), "\n";
234 Looks like case b isn't a terribly good idea most of the time -- it
235 will generate many spurious matches in series books in particular. So
236 for now, simply turning it off.
238 if ($marc->{edition}) { # case b : edition
239 print OF join("\t", $marc->{id}, "case b",
240 $marc->{item_form}, $marc->{date1},
241 $marc->{record_type}, $marc->{bib_lvl},
242 $marc->{title}, $marc->{edition}), "\n";
248 if ($marc->{issn}) { # case c : issn
249 print OF join("\t", $marc->{id}, "case c",
250 $marc->{item_form}, $marc->{date1},
251 $marc->{record_type}, $marc->{bib_lvl},
252 $marc->{title}, $marc->{issn}), "\n";
256 if ($marc->{lccn}) { # case d : lccn
257 print OF join("\t", $marc->{id}, "case d",
258 $marc->{item_form}, $marc->{date1},
259 $marc->{record_type}, $marc->{bib_lvl},
260 $marc->{title}, $marc->{lccn}) ,"\n";
264 if ($marc->{accomp}) { # case e : accomp
265 print OF join("\t", $marc->{id}, "case e",
266 $marc->{item_form}, $marc->{date1},
267 $marc->{record_type}, $marc->{bib_lvl},
268 $marc->{title}, $marc->{accomp}) ,"\n";
273 if (scalar @{$marc->{oclc} }) {
274 for (@{$marc->{oclc} }) {
275 print OF join("\t", $marc->{id}, "case o",
276 $marc->{item_form}, $marc->{date1},
277 $marc->{record_type}, $marc->{bib_lvl},
278 $marc->{title}, $_, "\n");
283 # case z : author, publisher, pubyear, pages
284 if ($marc->{author} and $marc->{publisher} and $marc->{pubyear}
285 and $marc->{pages}) {
286 print OF join("\t", $marc->{id}, "case z",
287 $marc->{item_form}, $marc->{date1},
288 $marc->{record_type}, $marc->{bib_lvl},
289 $marc->{title}, $marc->{author},
290 $marc->{publisher}, $marc->{pubyear},
291 $marc->{pages}), "\n";
295 # case poo : nothing good; dump a primary and move on
297 print OF join("\t", $marc->{id}, "case poo", $marc->{item_form},
298 $marc->{date1}, $marc->{record_type},
299 $marc->{bib_lvl}, $marc->{title}), "\n";
306 =head2 dump_exception
308 Write line of exception report
314 unless (defined $marc) {
315 print XF "Undefined record at line $count; likely bad XML\n";
318 print XF "Record ", $marc->{id}, " did not make the cut: ";
319 print XF "Missing item_form. " unless ($marc->{item_form});
320 unless (defined $marc->{date1})
321 { print XF "Missing date1. " }
323 { print XF "Invalid date1: ", $marc->{date1}, " "
324 unless ($marc->{date1} =~ /\d{4}/); }
325 print XF "Missing record_type. " unless ($marc->{record_type});
326 print XF "Missing bib_lvl. " unless ($marc->{bib_lvl});
327 print XF "Missing title. " unless ($marc->{title});
334 Performs boring script initialization. Handles argument parsing,
343 # set mode on existing filehandles
344 binmode(STDIN, ':utf8');
346 my $rc = GetOptions( $c,
358 show_help() unless $rc;
359 show_help() if ($c->{help});
361 # set defaults if told to do so
362 if ($c->{incoming}) {
363 $c->{tag} = 903 unless defined $c->{tag};
364 $c->{subfield} = 'a' unless defined $c->{subfield};
365 $c->{marctype} = 'XML' unless defined $c->{marctype};
366 $c->{output} = 'incoming.fp' unless defined $c->{output};
367 $c->{exception} = 'incoming.ex' unless defined $c->{exception};
368 $c->{runtype} = 'full' unless defined $c->{runtype};
369 } elsif ($c->{incumbent}) {
370 $c->{tag} = 901 unless defined $c->{tag};
371 $c->{subfield} = 'c' unless defined $c->{subfield};
372 $c->{marctype} = 'XML' unless defined $c->{marctype};
373 $c->{output} = 'incumbent.fp' unless defined $c->{output};
374 $c->{exception} = 'incumbent.ex' unless defined $c->{exception};
375 $c->{runtype} = 'full' unless defined $c->{runtype};
378 my @keys = keys %{$c};
379 show_help() unless (@ARGV and @keys);
380 for my $key ('runtype', 'tag', 'subfield', 'output', 'exception')
381 { push @missing, $key unless $c->{$key} }
383 print "Required option: ", join(', ', @missing), " missing!\n";
389 =head2 progress_ticker
393 sub progress_ticker {
394 return if $conf->{quiet};
395 printf("\r> %d recs seen; %d processed", $count, $scount);
396 printf(" (%d/s)", ($count / (time - $start + 1)))
397 if ($count % 500 == 0);
402 Display usage message when things go wrong
408 Usage is: $0 [REQUIRED ARGS] [OPTIONS] <filelist>
410 --runtype=(primary|full) -r Do 'primary' or 'full' fingerprinting
411 --tag=N -t Which tag to use
412 --subfield=X -s Which subfield to use
413 --output=<file> -o Output filename
414 --exceptions=<file> -x Exception report filename
416 --incoming Set -r to 'full'; -t, -s, -o, -x to incoming defaults
417 --incumbent Set -r to 'full'; -t, -s, -o, -x to incumbent defaults
419 Example: '$0 --incoming' is equivalent to
420 '$0 -r full -t 903 -s a -o incoming.fp -x incoming.ex'
422 --quiet -q Don't write status messages to STDOUT