Open-ILS/src/perlmods/OpenILS/Utils/ModsParser.pm

   1 package OpenILS::Utils::ModsParser;
   2 use strict; use warnings;
   3
   4 use OpenSRF::EX qw/:try/;
   5 use XML::LibXML;
   6 use XML::LibXSLT;
   7 use Time::HiRes qw(time);
   8 use OpenILS::Utils::Fieldmapper;
   9 use OpenSRF::Utils::SettingsClient;
  10 use Data::Dumper;
  11
  12 my $parser              = XML::LibXML->new();
  13 my $xslt                        = XML::LibXSLT->new();
  14 my $mods_sheet;
  15
  16 # ----------------------------------------------------------------------------------------
  17 # XPATH for extracting info from a MODS doc
  18 my $isbn_xpath                  = "//mods:mods/mods:identifier[\@type='isbn']";
  19 my $resource_xpath      = "//mods:mods/mods:typeOfResource";
  20 my $pub_xpath                   = "//mods:mods/mods:originInfo//mods:dateIssued[\@encoding='marc']|" .
  21                                                                 "//mods:mods/mods:originInfo//mods:dateIssued[1]";
  22 my $tcn_xpath                   = "//mods:mods/mods:recordInfo/mods:recordIdentifier";
  23 my $publisher_xpath     = "//mods:mods/mods:originInfo//mods:publisher[1]";
  24 my $edition_xpath               = "//mods:mods/mods:originInfo//mods:edition[1]";
  25 my $abstract_xpath      = "//mods:mods/mods:abstract";
  26 my $toc_xpath                   = "";
  27 my $related_xpath               = "";
  28 my $online_loc_xpath = "(//mods:location/mods:url|//mods:location/mods:url/\@displayLabel)";
  29
  30 my $xpathset = {
  31
  32         title => {
  33                 abbreviated =>
  34                         "//mods:mods/mods:titleInfo[mods:title and (\@type='abreviated')]",
  35                 translated =>
  36                         "//mods:mods/mods:titleInfo[mods:title and (\@type='translated')]",
  37                 uniform =>
  38                         "//mods:mods/mods:titleInfo[mods:title and (\@type='uniform')]",
  39                 proper =>
  40                         "//mods:mods/mods:titleInfo[mods:title and not (\@type)]",
  41         },
  42
  43         author => {
  44                 corporate =>
  45                         "//mods:mods/mods:name[\@type='corporate']/*[local-name()='namePart']".
  46                                 "[../mods:role/mods:text[text()='creator']][1]",
  47                 personal =>
  48                         "//mods:mods/mods:name[\@type='personal']/*[local-name()='namePart']".
  49                                 "[../mods:role/mods:text[text()='creator']][1]",
  50                 conference =>
  51                         "//mods:mods/mods:name[\@type='conference']/*[local-name()='namePart']".
  52                                 "[../mods:role/mods:text[text()='creator']][1]",
  53                 other =>
  54                         "//mods:mods/mods:name[\@type='personal']/*[local-name()='namePart']",
  55         },
  56
  57         subject => {
  58
  59                 topic =>
  60                         "//mods:mods/mods:subject/*[local-name()!='geographicCode']/parent::mods:subject",
  61
  62 #               geographic =>
  63 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='geographic']",
  64 #               name =>
  65 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='name']",
  66 #               temporal =>
  67 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='temporal']",
  68 #               topic =>
  69 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='topic']",
  70         },
  71         #keyword => { keyword => "//mods:mods/*[not(local-name()='originInfo')]", },
  72
  73         series => {
  74                 series => "//mods:mods/mods:relatedItem[\@type='series']/mods:titleInfo"
  75         }
  76 };
  77 # ----------------------------------------------------------------------------------------
  78
  79
  80
  81 sub new { return bless( {}, shift() ); }
  82
  83 sub get_field_value {
  84
  85         my( $self, $mods, $xpath ) = @_;
  86
  87         my @string;
  88         my $root = $mods->documentElement;
  89         $root->setNamespace( "http://www.loc.gov/mods/v3", "mods", 1 );
  90
  91         # grab the set of matching nodes
  92         my @nodes = $root->findnodes( $xpath );
  93         for my $value (@nodes) {
  94
  95                 # grab all children of the node
  96                 my @children = $value->childNodes();
  97                 my @child_text;
  98                 for my $child (@children) {
  99                         next unless( $child->nodeType != 3 );
 100
 101                         if($child->childNodes) {
 102                                 my @a;
 103                                 for my $c (@{$child->childNodes}){
 104                                         push @a, $c->textContent;
 105                                 }
 106                                 push(@child_text, join(' ', @a));
 107
 108                         } else {
 109                                 push(@child_text, $child->textContent);
 110                         }
 111
 112                 }
 113                 if(@child_text) {
 114                         push(@string, \@child_text);
 115                 }
 116
 117                 if( !@child_text  ) {
 118                         push(@string, $value->textContent );
 119                 }
 120         }
 121         return @string;
 122 }
 123
 124 =head
 125 sub _modsdoc_to_values {
 126         my( $self, $mods ) = @_;
 127         my $data = {};
 128         for my $class (keys %$xpathset) {
 129                 $data->{$class} = {};
 130                 for my $type(keys %{$xpathset->{$class}}) {
 131                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 132                         if( $class eq "subject" ) {
 133                                 push( @{$data->{$class}->{$type}},  @value );
 134                         } else {
 135                                 $data->{$class}->{$type} = $value[0];
 136                         }
 137                 }
 138         }
 139         return $data;
 140 }
 141 =cut
 142
 143 sub modsdoc_to_values {
 144         my( $self, $mods ) = @_;
 145         my $data = {};
 146
 147         {
 148                 my $class = "subject";
 149                 $data->{$class} = {};
 150                 for my $type(keys %{$xpathset->{$class}}) {
 151                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 152                         for my $arr (@value) {
 153                                 push( @{$data->{$class}->{$type}},  $arr);
 154                         }
 155                 }
 156         }
 157
 158         {
 159                 my $class = "title";
 160                 $data->{$class} = {};
 161                 for my $type(keys %{$xpathset->{$class}}) {
 162                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 163                         for my $arr (@value) {
 164                                 if( ref($arr) ) {
 165                                         $data->{$class}->{$type} = join(" ", @$arr);
 166                                 } else {
 167                                         $data->{$class}->{$type} = $arr;
 168                                 }
 169                         }
 170                 }
 171         }
 172
 173         {
 174                 my $class = "author";
 175                 $data->{$class} = {};
 176                 for my $type(keys %{$xpathset->{$class}}) {
 177                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 178                         $data->{$class}->{$type} = $value[0];
 179                 }
 180         }
 181
 182         {
 183                 my $class = "series";
 184                 $data->{$class} = {};
 185                 for my $type(keys %{$xpathset->{$class}}) {
 186                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 187                         for my $arr (@value) {
 188                                 if( ref($arr) ) {
 189                                         push(@{$data->{$class}->{$type}}, join(" ", @$arr));
 190                                 } else {
 191                                         push( @{$data->{$class}->{$type}}, $arr );
 192                                 }
 193                         }
 194                 }
 195
 196         }
 197
 198         return $data;
 199 }
 200
 201
 202
 203
 204 # ---------------------------------------------------------------------------
 205 # Grabs the data 'we want' from the MODS doc and returns it in hash form
 206 # ---------------------------------------------------------------------------
 207 sub mods_values_to_mods_slim {
 208         my( $self, $modsperl ) = @_;
 209
 210         my $title = "";
 211         my $author = "";
 212         my $subject = [];
 213         my $series      = [];
 214
 215         my $tmp = $modsperl->{title};
 216
 217
 218         if(!$tmp) { $title = ""; }
 219         else {
 220                 ($title = $tmp->{proper}) ||
 221                 ($title = $tmp->{translated}) ||
 222                 ($title = $tmp->{abbreviated}) ||
 223                 ($title = $tmp->{uniform});
 224         }
 225
 226         $tmp = $modsperl->{author};
 227         if(!$tmp) { $author = ""; }
 228         else {
 229                 ($author = $tmp->{personal}) ||
 230                 ($author = $tmp->{other}) ||
 231                 ($author = $tmp->{corporate}) ||
 232                 ($author = $tmp->{conference});
 233         }
 234
 235         $tmp = $modsperl->{subject};
 236         if(!$tmp) { $subject = {}; }
 237         else {
 238                 for my $key( keys %{$tmp}) {
 239                         push(@$subject, @{$tmp->{$key}}) if ($tmp->{$key});
 240                 }
 241                 my $subh = {};
 242                 for my $s (@$subject) {
 243                         if(defined($subh->{$s})) { $subh->{$s->[0]}++ } else { $subh->{$s->[0]} = 1;}
 244                 }
 245                 $subject = $subh
 246         }
 247
 248         $tmp = $modsperl->{'series'};
 249         if(!$tmp) { $series = []; }
 250         else { $series = $tmp->{'series'}; }
 251
 252
 253         return { series => $series, title => $title,
 254                         author => $author, subject => $subject };
 255 }
 256
 257
 258
 259 # ---------------------------------------------------------------------------
 260 # Initializes a MARC -> Unified MODS batch process
 261 # ---------------------------------------------------------------------------
 262
 263 sub start_mods_batch {
 264
 265         my( $self, $master_doc ) = @_;
 266
 267
 268         if(!$mods_sheet) {
 269                  my $xslt_doc = $parser->parse_file(
 270                         OpenSRF::Utils::SettingsClient->new->config_value(dirs => 'xsl') .  "/MARC21slim2MODS3.xsl");
 271                 $mods_sheet = $xslt->parse_stylesheet( $xslt_doc );
 272         }
 273
 274
 275         my $xmldoc = $parser->parse_string($master_doc);
 276         my $mods = $mods_sheet->transform($xmldoc);
 277
 278 #       warn "-" x 100 . "\n";
 279 #       warn "MODS " . $mods->toString(1) . "\n";
 280 #       warn "-" x 100 . "\n";
 281
 282         $self->{master_doc} = $self->modsdoc_to_values( $mods );
 283         $self->{master_doc} = $self->mods_values_to_mods_slim( $self->{master_doc} );
 284
 285         ($self->{master_doc}->{isbn}) =
 286                 $self->get_field_value( $mods, $isbn_xpath );
 287
 288         $self->{master_doc}->{type_of_resource} =
 289                 [ $self->get_field_value( $mods, $resource_xpath ) ];
 290
 291         ($self->{master_doc}->{tcn}) =
 292                 $self->get_field_value( $mods, $tcn_xpath );
 293
 294         ($self->{master_doc}->{pubdate}) =
 295                 $self->get_field_value( $mods, $pub_xpath );
 296
 297         ($self->{master_doc}->{publisher}) =
 298                 $self->get_field_value( $mods, $publisher_xpath );
 299
 300         ($self->{master_doc}->{edition}) =
 301                 $self->get_field_value( $mods, $edition_xpath );
 302
 303
 304
 305 # ------------------------------
 306         # holds an array of [ link, title, link, title, ... ]
 307         $self->{master_doc}->{online_loc} = [];
 308         push(@{$self->{master_doc}->{online_loc}},
 309                 $self->get_field_value( $mods, $online_loc_xpath ));
 310
 311         ($self->{master_doc}->{synopsis}) =
 312                 $self->get_field_value( $mods, $abstract_xpath );
 313
 314 }
 315
 316
 317
 318 # ---------------------------------------------------------------------------
 319 # Takes a MARCXML string and adds it to the growing MODS doc
 320 # ---------------------------------------------------------------------------
 321 sub push_mods_batch {
 322         my( $self, $marcxml ) = @_;
 323
 324         my $xmldoc = $parser->parse_string($marcxml);
 325         my $mods = $mods_sheet->transform($xmldoc);
 326
 327         my $xmlperl = $self->modsdoc_to_values( $mods );
 328         $xmlperl = $self->mods_values_to_mods_slim( $xmlperl );
 329
 330         # for backwards compatibility, remove the array part when all is decided
 331         if(ref($xmlperl->{subject}) eq 'ARRAY' ) {
 332                 for my $subject( @{$xmlperl->{subject}} ) {
 333                         push @{$self->{master_doc}->{subject}}, $subject;
 334                 }
 335         } else {
 336                 for my $subject ( keys %{$xmlperl->{subject}} ) {
 337                         my $s = $self->{master_doc}->{subject};
 338                         if(defined($s->{$subject})) { $s->{$subject}++; } else { $s->{$subject} = 1; }
 339                 }
 340         }
 341
 342         push( @{$self->{master_doc}->{type_of_resource}},
 343                 $self->get_field_value( $mods, $resource_xpath ));
 344
 345         if(!($self->{master_doc}->{isbn}) ) {
 346                 ($self->{master_doc}->{isbn}) =
 347                         $self->get_field_value( $mods, $isbn_xpath );
 348         }
 349 }
 350
 351
 352 # ---------------------------------------------------------------------------
 353 # Completes a MARC -> Unified MODS batch process and returns the perl hash
 354 # ---------------------------------------------------------------------------
 355 sub init_virtual_record {
 356         my $record = new Fieldmapper::metabib::virtual_record;
 357         $record->subject([]);
 358         $record->types_of_resource([]);
 359         $record->call_numbers([]);
 360         return $record;
 361 }
 362
 363 sub finish_mods_batch {
 364         my $self = shift;
 365         my $perl = $self->{master_doc};
 366         my $record = init_virtual_record();
 367
 368         # turn the hash into a fieldmapper object
 369         (my $title = $perl->{title}) =~ s/\[.*?\]//og;
 370         (my $author = $perl->{author}) =~ s/\(.*?\)//og;
 371
 372         my @series;
 373         for my $s (@{$perl->{series}}) {
 374                 push @series, (split( /\s*;/, $s ))[0];
 375         }
 376
 377         # uniquify the types of resource
 378         my $rtypes = $perl->{type_of_resource};
 379         my %hash = map { ($_ => 1) } @$rtypes;
 380         $rtypes = [ keys %hash ];
 381
 382         $record->title($title);
 383         $record->author($author);
 384
 385         $record->doc_id($perl->{doc_id});
 386         $record->isbn($perl->{isbn});
 387         $record->pubdate($perl->{pubdate});
 388         $record->publisher($perl->{publisher});
 389         $record->tcn($perl->{tcn});
 390
 391         $record->edition($perl->{edition});
 392
 393         $record->subject($perl->{subject});
 394         $record->types_of_resource($rtypes);
 395         $record->series(\@series);
 396
 397         $record->online_loc($perl->{online_loc});
 398         $record->synopsis($perl->{synopsis});
 399
 400         $self->{master_doc} = undef;
 401         return $record;
 402 }
 403
 404
 405