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};
19 open XF, '>', $conf->{exception};
21 for my $file (@ARGV) {
22 print XF "Processing $file\n";
23 #open my $records, '<:utf8', $file; # This hack dosn't play nice with the use MARC::File::XML ( BinaryEncoding => 'utf-8' ); hack
24 my $batch = undef; my $record = undef;
26 #$batch = MARC::Batch->new('XML', $records); # The other part of the hack
27 $batch = MARC::Batch->new($conf->{marctype}, $file);
29 $batch->warnings_off();
31 while ( $record = $batch->next ) {
32 $count++; progress_ticker();
34 unless ( defined $record )
35 { dump_exception($marc); next; }
37 my $id = $record->field($conf->{tag});
39 print XF "ERROR: Record $count in $file is missing a ",
40 $conf->{tag}, " field.\n", $record->as_formatted(), "\n=====\n";
44 $marc = populate_marc($record, $id);
45 $marc = normalize_marc($marc);
46 unless (marc_isvalid($marc))
47 { dump_exception($marc); next; }
48 dump_fingerprints($marc);
49 $scount++; progress_ticker();
53 print "\nSuccessfully processed:\t$count\n" unless $conf->{quiet};
57 Constructs a hash containing the relevant MARC data for a record and
58 returns a reference to it.
63 my ($record, $id) = @_;
64 my %marc = (); $marc{isbns} = [];
67 $marc{id} = $id->as_string($conf->{subfield});
69 # record_type, bib_lvl
70 $marc{record_type} = substr($record->leader, 6, 1);
71 $marc{bib_lvl} = substr($record->leader, 7, 1);
74 my $my_008 = $record->field('008');
75 $my_008 = $my_008->as_string() if ($my_008);
76 if (defined $my_008) {
77 unless (length $my_008 == 40) {
78 $my_008 = $my_008 . (' ' x (40 - length($my_008)));
79 print XF ">> Short 008 padded to ",length($my_008)," at rec $count\n";
81 $marc{date1} = substr($my_008,7,4) if ($my_008);
82 $marc{date2} = substr($my_008,11,4) if ($my_008); # UNUSED
84 unless ($marc{date1} and $marc{date1} =~ /\d{4}/) {
85 my $my_260 = $record->field('260');
86 my $date1 = $my_260->subfield('c') if $my_260;
87 if (defined $date1 and $date1 =~ /\d{4}/) {
88 $marc{date1} = $date1;
89 print XF ">> using 260c as date1 at rec $count\n";
94 if ( $marc{record_type} =~ /[gkroef]/ ) { # MAP, VIS
95 $marc{item_form} = substr($my_008,29,1) if ($my_008);
97 $marc{item_form} = substr($my_008,23,1) if ($my_008);
101 my @isbns = $record->field('020') if $record->field('020');
102 push @isbns, $record->field('024') if $record->field('024');
103 for my $f ( @isbns ) {
104 push @{ $marc{isbns} }, $1 if ( defined $f->subfield('a') and
105 $f->subfield('a')=~/(\S+)/ );
109 for my $rec_field (100, 110, 111) {
110 if ($record->field($rec_field)) {
111 $marc{author} = $record->field($rec_field)->subfield('a');
118 push @{ $marc{oclc} }, $record->field('001')->as_string()
119 if ($record->field('001') and $record->field('003') and
120 $record->field('003')->as_string() =~ /OCo{0,1}LC/);
121 for ($record->field('035')) {
122 my $oclc = $_->subfield('a');
123 push @{ $marc{oclc} }, $oclc
124 if (defined $oclc and $oclc =~ /\(OCoLC\)/ and $oclc =~/([0-9]+)/);
127 # "Accompanying material" (300e)
128 $marc{accomp} = $record->field('300')->subfield('e')
129 if $record->field('300');
131 # issn, lccn, title, desc, pages, pub, pubyear, edition
132 $marc{lccn} = $record->field('010')->subfield('a') if $record->field('010');
133 $marc{issn} = $record->field('022')->subfield('a') if $record->field('022');
134 $marc{desc} = $record->field('300')->subfield('a') if $record->field('300');
135 $marc{pages} = $1 if (defined $marc{desc} and $marc{desc} =~ /(\d+)/);
136 $marc{title} = $record->field('245')->subfield('a')
137 if $record->field('245');
138 $marc{edition} = $record->field('250')->subfield('a')
139 if $record->field('250');
140 if ($record->field('260')) {
141 $marc{publisher} = $record->field('260')->subfield('b');
142 $marc{pubyear} = $record->field('260')->subfield('c');
144 (defined $marc{pubyear} and $marc{pubyear} =~ /(\d{4})/) ? $1 : '';
151 =head2 normalize_marc
153 Gently massages your data.
160 $marc->{record_type }= 'a' if ($marc->{record_type} eq ' ');
161 if ($marc->{title}) {
162 $marc->{title} = NFD($marc->{title});
163 $marc->{title} =~ s/[\x{80}-\x{ffff}]//go;
164 $marc->{title} = lc($marc->{title});
165 $marc->{title} =~ s/\W+$//go;
167 if ($marc->{author}) {
168 $marc->{author} = NFD($marc->{author});
169 $marc->{author} =~ s/[\x{80}-\x{ffff}]//go;
170 $marc->{author} = lc($marc->{author});
171 $marc->{author} =~ s/\W+$//go;
172 if ($marc->{author} =~ /^(\w+)/) {
173 $marc->{author} = $1;
176 if ($marc->{publisher}) {
177 $marc->{publisher} = NFD($marc->{publisher});
178 $marc->{publisher} =~ s/[\x{80}-\x{ffff}]//go;
179 $marc->{publisher} = lc($marc->{publisher});
180 $marc->{publisher} =~ s/\W+$//go;
181 if ($marc->{publisher} =~ /^(\w+)/) {
182 $marc->{publisher} = $1;
192 Checks MARC record to see if neccessary fingerprinting data is
199 return 1 if ($marc->{item_form} and ($marc->{date1} =~ /\d{4}/) and
200 $marc->{record_type} and $marc->{bib_lvl} and $marc->{title});
205 =head2 dump_fingerprints
209 sub dump_fingerprints {
213 if ($conf->{runtype} eq "primary") {
214 print OF join("\t",$marc->{id}, $marc->{item_form},
215 $marc->{date1}, $marc->{record_type},
216 $marc->{bib_lvl}, $marc->{title}), "\n";
218 if ((scalar @{ $marc->{isbns} } > 0) and $marc->{pages}) {
219 # case a : isbn and pages
220 foreach my $isbn ( @{ $marc->{isbns}} ) {
221 print OF join("\t", $marc->{id}, "case a",
222 $marc->{item_form}, $marc->{date1},
223 $marc->{record_type},
224 $marc->{bib_lvl}, $marc->{title},
225 $isbn, $marc->{pages}), "\n";
232 Looks like case b isn't a terribly good idea most of the time -- it
233 will generate many spurious matches in series books in particular. So
234 for now, simply turning it off.
236 if ($marc->{edition}) { # case b : edition
237 print OF join("\t", $marc->{id}, "case b",
238 $marc->{item_form}, $marc->{date1},
239 $marc->{record_type}, $marc->{bib_lvl},
240 $marc->{title}, $marc->{edition}), "\n";
246 if ($marc->{issn}) { # case c : issn
247 print OF join("\t", $marc->{id}, "case c",
248 $marc->{item_form}, $marc->{date1},
249 $marc->{record_type}, $marc->{bib_lvl},
250 $marc->{title}, $marc->{issn}), "\n";
254 if ($marc->{lccn}) { # case d : lccn
255 print OF join("\t", $marc->{id}, "case d",
256 $marc->{item_form}, $marc->{date1},
257 $marc->{record_type}, $marc->{bib_lvl},
258 $marc->{title}, $marc->{lccn}) ,"\n";
262 if ($marc->{accomp}) { # case e : accomp
263 print OF join("\t", $marc->{id}, "case e",
264 $marc->{item_form}, $marc->{date1},
265 $marc->{record_type}, $marc->{bib_lvl},
266 $marc->{title}, $marc->{accomp}) ,"\n";
271 if (scalar @{$marc->{oclc} }) {
272 for (@{$marc->{oclc} }) {
273 print OF join("\t", $marc->{id}, "case o",
274 $marc->{item_form}, $marc->{date1},
275 $marc->{record_type}, $marc->{bib_lvl},
276 $marc->{title}, $_, "\n");
281 # case z : author, publisher, pubyear, pages
282 if ($marc->{author} and $marc->{publisher} and $marc->{pubyear}
283 and $marc->{pages}) {
284 print OF join("\t", $marc->{id}, "case z",
285 $marc->{item_form}, $marc->{date1},
286 $marc->{record_type}, $marc->{bib_lvl},
287 $marc->{title}, $marc->{author},
288 $marc->{publisher}, $marc->{pubyear},
289 $marc->{pages}), "\n";
293 # case poo : nothing good; dump a primary and move on
295 print OF join("\t", $marc->{id}, "case poo", $marc->{item_form},
296 $marc->{date1}, $marc->{record_type},
297 $marc->{bib_lvl}, $marc->{title}), "\n";
304 =head2 dump_exception
306 Write line of exception report
312 unless (defined $marc) {
313 print XF "Undefined record at line $count; likely bad XML\n";
316 print XF "Record ", $marc->{id}, " did not make the cut: ";
317 print XF "Missing item_form. " unless ($marc->{item_form});
318 unless (defined $marc->{date1})
319 { print XF "Missing date1. " }
321 { print XF "Invalid date1: ", $marc->{date1}, " "
322 unless ($marc->{date1} =~ /\d{4}/); }
323 print XF "Missing record_type. " unless ($marc->{record_type});
324 print XF "Missing bib_lvl. " unless ($marc->{bib_lvl});
325 print XF "Missing title. " unless ($marc->{title});
332 Performs boring script initialization. Handles argument parsing,
341 # set mode on existing filehandles
342 binmode(STDIN, ':utf8');
344 my $rc = GetOptions( $c,
356 show_help() unless $rc;
357 show_help() if ($c->{help});
359 # set defaults if told to do so
360 if ($c->{incoming}) {
361 $c->{tag} = 903 unless defined $c->{tag};
362 $c->{subfield} = 'a' unless defined $c->{subfield};
363 $c->{marctype} = 'XML' unless defined $c->{marctype};
364 $c->{output} = 'incoming.fp' unless defined $c->{output};
365 $c->{exception} = 'incoming.ex' unless defined $c->{exception};
366 $c->{runtype} = 'full' unless defined $c->{runtype};
367 } elsif ($c->{incumbent}) {
368 $c->{tag} = 901 unless defined $c->{tag};
369 $c->{subfield} = 'c' unless defined $c->{subfield};
370 $c->{marctype} = 'XML' unless defined $c->{marctype};
371 $c->{output} = 'incumbent.fp' unless defined $c->{output};
372 $c->{exception} = 'incumbent.ex' unless defined $c->{exception};
373 $c->{runtype} = 'full' unless defined $c->{runtype};
376 my @keys = keys %{$c};
377 show_help() unless (@ARGV and @keys);
378 for my $key ('runtype', 'tag', 'subfield', 'output', 'exception')
379 { push @missing, $key unless $c->{$key} }
381 print "Required option: ", join(', ', @missing), " missing!\n";
387 =head2 progress_ticker
391 sub progress_ticker {
392 return if $conf->{quiet};
393 printf("\r> %d recs seen; %d processed", $count, $scount);
394 printf(" (%d/s)", ($count / (time - $start + 1)))
395 if ($count % 500 == 0);
400 Display usage message when things go wrong
406 Usage is: $0 [REQUIRED ARGS] [OPTIONS] <filelist>
408 --runtype=(primary|full) -r Do 'primary' or 'full' fingerprinting
409 --tag=N -t Which tag to use
410 --subfield=X -s Which subfield to use
411 --output=<file> -o Output filename
412 --exceptions=<file> -x Exception report filename
414 --incoming Set -r to 'full'; -t, -s, -o, -x to incoming defaults
415 --incumbent Set -r to 'full'; -t, -s, -o, -x to incumbent defaults
417 Example: '$0 --incoming' is equivalent to
418 '$0 -r full -t 903 -s a -o incoming.fp -x incoming.ex'
420 --quiet -q Don't write status messages to STDOUT