8 use MARC::File::XML ( BinaryEncoding => 'utf-8' );
10 use Unicode::Normalize;
12 my $conf = {}; # configuration hashref
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;
26 my $batch = undef; my $record = undef;
28 $batch = MARC::Batch->new('XML', $records);
30 $batch->warnings_off();
32 while ( eval { $record = $batch->next } ) {
34 $count++; progress_ticker();
35 unless ( defined $record )
36 { dump_exception($marc); next; }
38 my $id = $record->field($conf->{tag});
40 print XF "ERROR: Record $count in $file is missing a ",
41 $conf->{tag}, " field.\n", $record->as_formatted(), "\n=====\n";
45 $marc = populate_marc($record, $id);
46 $marc = normalize_marc($marc);
47 unless (marc_isvalid($marc))
48 { dump_exception($marc); next; }
49 dump_fingerprints($marc);
52 print "\nProcessed $count records\n" unless $conf->{quiet};
58 Constructs a hash containing the relevant MARC data for a record and
59 returns a reference to it.
64 my ($record, $id) = @_;
65 my %marc = (); $marc{isbns} = [];
68 $marc{id} = $id->as_string($conf->{subfield});
70 # record_type, bib_lvl
71 $marc{record_type} = substr($record->leader, 6, 1);
72 $marc{bib_lvl} = substr($record->leader, 7, 1);
75 my $my_008 = $record->field('008');
76 $my_008 = $my_008->as_string() if ($my_008);
77 unless (length $my_008 == 40)
78 { print XF ">> Bad 008 field length in rec. $id\n"; return \%marc }
79 $marc{date1} = substr($my_008,7,4) if ($my_008);
80 $marc{date2} = substr($my_008,11,4) if ($my_008); # UNUSED
83 if ( $marc{record_type} =~ /[gkroef]/ ) { # MAP, VIS
84 $marc{item_form} = substr($my_008,29,1) if ($my_008);
86 $marc{item_form} = substr($my_008,23,1) if ($my_008);
90 my @isbns = $record->field('020') if $record->field('020');
91 push @isbns, $record->field('024') if $record->field('024');
92 for my $f ( @isbns ) {
93 push @{ $marc{isbns} }, $1 if ( defined $f->subfield('a') and
94 $f->subfield('a')=~/(\S+)/ );
98 for my $rec_field (100, 110, 111) {
99 if ($record->field($rec_field)) {
100 $marc{author} = $record->field($rec_field)->subfield('a');
107 # issn, lccn, title, desc, pages, pub, pubyear, edition
108 $marc{lccn} = $record->field('010')->subfield('a') if $record->field('010');
109 $marc{issn} = $record->field('022')->subfield('a') if $record->field('022');
110 $marc{desc} = $record->field('300')->subfield('a') if $record->field('300');
111 $marc{pages} = $1 if (defined $marc{desc} and $marc{desc} =~ /(\d+)/);
112 $marc{title} = $record->field('245')->subfield('a')
113 if defined $record->field('245');
114 $marc{edition} = $record->field('250')->subfield('a')
115 if $record->field('250');
116 if ($record->field('260')) {
117 $marc{publisher} = $record->field('260')->subfield('b');
118 $marc{pubyear} = $record->field('260')->subfield('c');
120 (defined $marc{pubyear} and $marc{pubyear} =~ /(\d{4})/) ? $1 : '';
127 =head2 normalize_marc
129 Gently massages your data.
136 $marc->{record_type }= 'a' if ($marc->{record_type} eq ' ');
137 if ($marc->{title}) {
138 $marc->{title} = NFD($marc->{title});
139 $marc->{title} =~ s/[\x{80}-\x{ffff}]//go;
140 $marc->{title} = lc($marc->{title});
141 $marc->{title} =~ s/\W+$//go;
143 if ($marc->{author}) {
144 $marc->{author} = NFD($marc->{author});
145 $marc->{author} =~ s/[\x{80}-\x{ffff}]//go;
146 $marc->{author} = lc($marc->{author});
147 $marc->{author} =~ s/\W+$//go;
148 if ($marc->{author} =~ /^(\w+)/) {
149 $marc->{author} = $1;
152 if ($marc->{publisher}) {
153 $marc->{publisher} = NFD($marc->{publisher});
154 $marc->{publisher} =~ s/[\x{80}-\x{ffff}]//go;
155 $marc->{publisher} = lc($marc->{publisher});
156 $marc->{publisher} =~ s/\W+$//go;
157 if ($marc->{publisher} =~ /^(\w+)/) {
158 $marc->{publisher} = $1;
168 Checks MARC record to see if neccessary fingerprinting data is
175 return 1 if ($marc->{item_form} and ($marc->{date1} =~ /\d{4}/) and
176 $marc->{record_type} and $marc->{bib_lvl} and $marc->{title});
181 =head2 dump_fingerprints
185 sub dump_fingerprints {
188 if ($conf->{runtype} eq "primary") {
189 print OF join("\t",$marc->{id}, $marc->{item_form},
190 $marc->{date1}, $marc->{record_type},
191 $marc->{bib_lvl}, $marc->{title}), "\n";
193 if ((scalar @{ $marc->{isbns} } > 0) && $marc->{pages}) {
194 # case a : isbn and pages
195 foreach my $isbn ( @{ $marc->{isbns}} ) {
196 print OF join("\t", $marc->{id}, "case a",
197 $marc->{item_form}, $marc->{date1},
198 $marc->{record_type},
199 $marc->{bib_lvl}, $marc->{title},
200 $isbn, $marc->{pages}), "\n";
204 if ($marc->{edition}) { # case b : edition
205 print OF join("\t", $marc->{id}, "case b",
206 $marc->{item_form}, $marc->{date1},
207 $marc->{record_type}, $marc->{bib_lvl},
208 $marc->{title}, $marc->{edition}), "\n";
211 if ($marc->{issn}) { # case c : issn
212 print OF join("\t", $marc->{id}, "case c",
213 $marc->{item_form}, $marc->{date1},
214 $marc->{record_type}, $marc->{bib_lvl},
215 $marc->{title}, $marc->{issn}), "\n";
218 if ($marc->{lccn}) { # case d : lccn
219 print OF join("\t", $marc->{id}, "case d",
220 $marc->{item_form}, $marc->{date1},
221 $marc->{record_type}, $marc->{bib_lvl},
222 $marc->{title}, $marc->{lccn}) ,"\n";
225 # case e : author, publisher, pubyear, pages
226 if ($marc->{author} and $marc->{publisher} and $marc->{pubyear}
227 and $marc->{pages}) {
228 print OF join("\t", $marc->{id}, "case e",
229 $marc->{item_form}, $marc->{date1},
230 $marc->{record_type}, $marc->{bib_lvl},
231 $marc->{title}, $marc->{author},
232 $marc->{publisher}, $marc->{pubyear},
233 $marc->{pages}), "\n";
240 =head2 dump_exception
242 Write line of exception report
248 unless (defined $marc) {
249 print XF "Undefined record at line $count; likely bad XML\n";
252 print XF "Record ", $marc->{id}, " did not make the cut: ";
253 print XF "Missing item_form. " unless ($marc->{item_form});
254 print XF "Missing date1. " unless (defined $marc->{date1});
255 print XF "Invalid date1: ", $marc->{date1}
256 unless ($marc->{date1} =~ /\d{4}/);
257 print XF "Missing record_type. " unless ($marc->{record_type});
258 print XF "Missing bib_lvl. " unless ($marc->{bib_lvl});
259 print XF "Missing title. " unless ($marc->{title});
266 Performs boring script initialization. Handles argument parsing,
275 # set mode on existing filehandles
276 binmode(STDIN, ':utf8');
278 my $rc = GetOptions( $c,
287 show_help() unless $rc;
289 my @keys = keys %{$c};
290 show_help() unless (@ARGV and @keys);
291 for my $key ('runtype', 'tag', 'subfield', 'output', 'exception') {
292 push @missing, $key unless $c->{$key}
295 print "Required option: ", join(', ', @missing), " missing!\n";
299 show_help() if ($c->{help});
303 =head2 progress_ticker
307 sub progress_ticker {
308 return if $conf->{quiet};
310 if ($count % 100 == 0) {
312 print " $count \n" unless ($count % 1400);
313 } elsif ($count % 20 == 0) {
320 Display usage message when things go wrong
326 Usage is: fingerprinter [REQUIRED ARGS] [OPTIONS] <filelist>
328 --runtype=(primary|full) -r Do 'primary' or 'full' fingerprinting
329 --tag=N -t Which tag to use
330 --subfield=X -s Which subfield to use
331 --output=<file> -o Output filename
332 --exceptions=<file> -x Exception report filename
334 --quiet -q Don't write status messages to STDOUT