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 OpenSRF::Utils::Logger qw/$logger/;
  11 use Data::Dumper;
  12
  13 my $parser              = XML::LibXML->new();
  14 my $xslt                        = XML::LibXSLT->new();
  15 my $mods_sheet;
  16
  17 # ----------------------------------------------------------------------------------------
  18 # XPATH for extracting info from a MODS doc
  19 my $isbn_xpath                  = "//mods:mods/mods:identifier[\@type='isbn']";
  20 my $resource_xpath      = "//mods:mods/mods:typeOfResource";
  21 my $pub_xpath                   = "//mods:mods/mods:originInfo//mods:dateIssued[\@encoding='marc']|" .
  22                                                                 "//mods:mods/mods:originInfo//mods:dateIssued[1]";
  23 my $tcn_xpath                   = "//mods:mods/mods:recordInfo/mods:recordIdentifier";
  24 my $publisher_xpath     = "//mods:mods/mods:originInfo//mods:publisher[1]";
  25 my $edition_xpath               = "//mods:mods/mods:originInfo//mods:edition[1]";
  26 my $abstract_xpath      = "//mods:mods/mods:abstract";
  27 my $related_xpath               = "";
  28 my $online_loc_xpath = "(//mods:location/mods:url|//mods:location/mods:url/\@displayLabel)";
  29 my $physical_desc               = "(//mods:physicalDescription/mods:form|//mods:physicalDescription/mods:extent|".
  30         "//mods:physicalDescription/mods:reformattingQuality|//mods:physicalDescription/mods:internetMediaType|".
  31         "//mods:physicalDescription/mods:digitalOrigin)";
  32 my $toc_xpath                   = "//mods:tableOfContents";
  33
  34 my $xpathset = {
  35
  36         title => {
  37                 abbreviated =>
  38                         "//mods:mods/mods:titleInfo[mods:title and (\@type='abreviated')]",
  39                 translated =>
  40                         "//mods:mods/mods:titleInfo[mods:title and (\@type='translated')]",
  41                 uniform =>
  42                         "//mods:mods/mods:titleInfo[mods:title and (\@type='uniform')]",
  43                 proper =>
  44                         "//mods:mods/mods:titleInfo[mods:title and not (\@type)]",
  45                 any =>
  46                         "//mods:mods/mods:titleInfo",
  47         },
  48
  49         author => {
  50                 corporate =>
  51                         "//mods:mods/mods:name[\@type='corporate']/*[local-name()='namePart']".
  52                                 "[../mods:role/mods:text[text()='creator']".
  53                                 " or ../mods:role/mods:roleTerm[".
  54                                 "        \@type='text'".
  55                                 "        and \@authority='marcrelator'".
  56                                 "        and text()='creator']".
  57                                 "][1]",
  58                 personal =>
  59                         "//mods:mods/mods:name[\@type='personal']/*[local-name()='namePart']".
  60                                 "[../mods:role/mods:text[text()='creator']".
  61                                 " or ../mods:role/mods:roleTerm[".
  62                                 "        \@type='text'".
  63                                 "        and \@authority='marcrelator'".
  64                                 "        and text()='creator']".
  65                                 "][1]",
  66                 conference =>
  67                         "//mods:mods/mods:name[\@type='conference']/*[local-name()='namePart']".
  68                                 "[../mods:role/mods:text[text()='creator']".
  69                                 " or ../mods:role/mods:roleTerm[".
  70                                 "        \@type='text'".
  71                                 "        and \@authority='marcrelator'".
  72                                 "        and text()='creator']".
  73                                 "][1]",
  74                 other =>
  75                         "//mods:mods/mods:name[\@type='personal']/*[local-name()='namePart']",
  76                 any =>
  77                         "//mods:mods/mods:name/*[local-name()='namePart'][1]",
  78         },
  79
  80         subject => {
  81
  82                 topic =>
  83                         "//mods:mods/mods:subject/*[".
  84                         "   local-name()='geographic'".
  85                         "   or local-name()='name'".
  86                         "   or local-name()='temporal'".
  87                         "   or local-name()='topic'".
  88                         "]/parent::mods:subject",
  89
  90 #               geographic =>
  91 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='geographic']",
  92 #               name =>
  93 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='name']",
  94 #               temporal =>
  95 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='temporal']",
  96 #               topic =>
  97 #                       "//mods:mods/*[local-name()='subject']/*[local-name()='topic']",
  98         },
  99         #keyword => { keyword => "//mods:mods/*[not(local-name()='originInfo')]", },
 100
 101         series => {
 102                 series => "//mods:mods/mods:relatedItem[\@type='series']/mods:titleInfo"
 103         }
 104 };
 105 # ----------------------------------------------------------------------------------------
 106
 107
 108
 109 sub new { return bless( {}, shift() ); }
 110
 111 sub get_field_value {
 112
 113         my( $self, $mods, $xpath ) = @_;
 114
 115         my @string;
 116
 117         my $root = $mods->documentElement;
 118         $root->setNamespace( "http://www.loc.gov/mods/v3", "mods", 1 );
 119
 120         try {
 121                 # grab the set of matching nodes
 122                 my @nodes = $root->findnodes( $xpath );
 123                 for my $value (@nodes) {
 124
 125                         # grab all children of the node
 126                         my @children = $value->childNodes();
 127                         my @child_text;
 128                         for my $child (@children) {
 129                                 next unless( $child->nodeType != 3 );
 130
 131                                 if($child->childNodes) {
 132                                         my @a;
 133                                         for my $c (@{$child->childNodes}){
 134                                                 push @a, $c->textContent;
 135                                         }
 136                                         push(@child_text, join(' ', @a));
 137
 138                                 } else {
 139                                         push(@child_text, $child->textContent);
 140                                 }
 141
 142                         }
 143                         if(@child_text) {
 144                                 push(@string, \@child_text);
 145                         }
 146
 147                         if( !@child_text  ) {
 148                                 push(@string, $value->textContent );
 149                         }
 150                 }
 151         } otherwise {
 152                 $logger->info("MODS-izing failure: ".shift());
 153                 $logger->info("Failed MODS xml: ".$root->toString);
 154                 $logger->info("Failed MODS xpath: $xpath");
 155         };
 156         return @string;
 157 }
 158
 159 =head
 160 sub _modsdoc_to_values {
 161         my( $self, $mods ) = @_;
 162         my $data = {};
 163         for my $class (keys %$xpathset) {
 164                 $data->{$class} = {};
 165                 for my $type(keys %{$xpathset->{$class}}) {
 166                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 167                         if( $class eq "subject" ) {
 168                                 push( @{$data->{$class}->{$type}},  @value );
 169                         } else {
 170                                 $data->{$class}->{$type} = $value[0];
 171                         }
 172                 }
 173         }
 174         return $data;
 175 }
 176 =cut
 177
 178 sub modsdoc_to_values {
 179         my( $self, $mods ) = @_;
 180         my $data = {};
 181
 182         {
 183                 my $class = "subject";
 184                 $data->{$class} = {};
 185                 for my $type(keys %{$xpathset->{$class}}) {
 186                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 187                         for my $arr (@value) {
 188                                 push( @{$data->{$class}->{$type}},  $arr);
 189                         }
 190                 }
 191         }
 192
 193         {
 194                 my $class = "title";
 195                 $data->{$class} = {};
 196                 for my $type(keys %{$xpathset->{$class}}) {
 197                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 198                         for my $arr (@value) {
 199                                 if( ref($arr) ) {
 200                                         $data->{$class}->{$type} = shift @$arr;
 201
 202                                         my $t = lc($data->{$class}->{$type});
 203                                         if($t and $t =~ /^l[eoa]s|l[ae]|el|the|un[ae]?|an?\s?$/o ) {
 204                                                 my $val = shift @$arr || "";
 205                                                 $data->{$class}->{$type} .= " $val" if $data->{$class}->{$type};
 206                                                 $data->{$class}->{$type} = " $val" unless $data->{$class}->{$type};
 207                                         }
 208
 209                                         for my $t (@$arr) {
 210                                                 $data->{$class}->{$type} .= ' : ' if ($data->{$class}->{$type} =~ /\w\s*$/o);
 211                                                 $data->{$class}->{$type} .= " $t";
 212                                         }
 213                                 } else {
 214                                         $data->{$class}->{$type} = $arr;
 215                                 }
 216                         }
 217                         $data->{$class}->{$type} =~ s/\s+/ /go if ($data->{$class}->{$type});
 218                 }
 219         }
 220
 221         {
 222                 my $class = "author";
 223                 $data->{$class} = {};
 224                 for my $type(keys %{$xpathset->{$class}}) {
 225                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 226                         $data->{$class}->{$type} = $value[0];
 227                 }
 228         }
 229
 230         {
 231                 my $class = "series";
 232                 $data->{$class} = {};
 233                 for my $type(keys %{$xpathset->{$class}}) {
 234                         my @value = $self->get_field_value( $mods, $xpathset->{$class}->{$type} );
 235                         for my $arr (@value) {
 236                                 if( ref($arr) ) {
 237                                         push(@{$data->{$class}->{$type}}, join(" ", @$arr));
 238                                 } else {
 239                                         push( @{$data->{$class}->{$type}}, $arr );
 240                                 }
 241                         }
 242                 }
 243
 244         }
 245
 246         return $data;
 247 }
 248
 249
 250
 251
 252 # ---------------------------------------------------------------------------
 253 # Grabs the data 'we want' from the MODS doc and returns it in hash form
 254 # ---------------------------------------------------------------------------
 255 sub mods_values_to_mods_slim {
 256         my( $self, $modsperl ) = @_;
 257
 258         my $title = "";
 259         my $author = "";
 260         my $subject = [];
 261         my $series      = [];
 262
 263         my $tmp = $modsperl->{title};
 264
 265
 266         if(!$tmp) { $title = ""; }
 267         else {
 268                 ($title = $tmp->{proper}) ||
 269                 ($title = $tmp->{translated}) ||
 270                 ($title = $tmp->{abbreviated}) ||
 271                 ($title = $tmp->{uniform}) ||
 272                 ($title = $tmp->{any});
 273         }
 274
 275         $tmp = $modsperl->{author};
 276         if(!$tmp) { $author = ""; }
 277         else {
 278                 ($author = $tmp->{personal}) ||
 279                 ($author = $tmp->{corporate}) ||
 280                 ($author = $tmp->{conference}) ||
 281                 ($author = $tmp->{other}) ||
 282                 ($author = $tmp->{any});
 283         }
 284
 285         $tmp = $modsperl->{subject};
 286         if(!$tmp) { $subject = {}; }
 287         else {
 288                 for my $key( keys %{$tmp}) {
 289                         push(@$subject, @{$tmp->{$key}}) if ($tmp->{$key});
 290                 }
 291                 my $subh = {};
 292                 for my $s (@$subject) {
 293                         if(defined($subh->{$s})) { $subh->{$s->[0]}++ } else { $subh->{$s->[0]} = 1;}
 294                 }
 295                 $subject = $subh
 296         }
 297
 298         $tmp = $modsperl->{'series'};
 299         if(!$tmp) { $series = []; }
 300         else { $series = $tmp->{'series'}; }
 301
 302
 303         return { series => $series, title => $title,
 304                         author => $author, subject => $subject };
 305 }
 306
 307
 308
 309 # ---------------------------------------------------------------------------
 310 # Initializes a MARC -> Unified MODS batch process
 311 # ---------------------------------------------------------------------------
 312
 313 sub start_mods_batch {
 314
 315         my( $self, $master_doc ) = @_;
 316
 317         if(!$master_doc) {
 318                 $self->{master_doc} = undef;
 319                 return;
 320         }
 321
 322         if(!$mods_sheet) {
 323                  my $xslt_doc = $parser->parse_file(
 324                         OpenSRF::Utils::SettingsClient->new->config_value(dirs => 'xsl') .  "/MARC21slim2MODS3.xsl");
 325                 $mods_sheet = $xslt->parse_stylesheet( $xslt_doc );
 326         }
 327
 328
 329         my $xmldoc = $parser->parse_string($master_doc);
 330         my $mods = $mods_sheet->transform($xmldoc);
 331
 332         $self->{master_doc} = $self->modsdoc_to_values( $mods );
 333         $self->{master_doc} = $self->mods_values_to_mods_slim( $self->{master_doc} );
 334
 335         ($self->{master_doc}->{isbn}) =
 336                 $self->get_field_value( $mods, $isbn_xpath );
 337
 338         $self->{master_doc}->{type_of_resource} =
 339                 [ $self->get_field_value( $mods, $resource_xpath ) ];
 340
 341         ($self->{master_doc}->{tcn}) =
 342                 $self->get_field_value( $mods, $tcn_xpath );
 343
 344         ($self->{master_doc}->{pubdate}) =
 345                 $self->get_field_value( $mods, $pub_xpath );
 346
 347         ($self->{master_doc}->{publisher}) =
 348                 $self->get_field_value( $mods, $publisher_xpath );
 349
 350         ($self->{master_doc}->{edition}) =
 351                 $self->get_field_value( $mods, $edition_xpath );
 352
 353
 354
 355 # ------------------------------
 356         # holds an array of [ link, title, link, title, ... ]
 357         $self->{master_doc}->{online_loc} = [];
 358         push(@{$self->{master_doc}->{online_loc}},
 359                 $self->get_field_value( $mods, $online_loc_xpath ));
 360
 361         ($self->{master_doc}->{synopsis}) =
 362                 $self->get_field_value( $mods, $abstract_xpath );
 363
 364         $self->{master_doc}->{physical_description} = [];
 365         push(@{$self->{master_doc}->{physical_description}},
 366                 $self->get_field_value( $mods, $physical_desc ) );
 367         $self->{master_doc}->{physical_description} =
 368                 join( ' ', @{$self->{master_doc}->{physical_description}});
 369
 370         ($self->{master_doc}->{toc}) = $self->get_field_value($mods, $toc_xpath);
 371
 372 }
 373
 374
 375
 376 # ---------------------------------------------------------------------------
 377 # Takes a MARCXML string and adds it to the growing MODS doc
 378 # ---------------------------------------------------------------------------
 379 sub push_mods_batch {
 380         my( $self, $marcxml ) = @_;
 381
 382         my $xmldoc = $parser->parse_string($marcxml);
 383         my $mods = $mods_sheet->transform($xmldoc);
 384
 385         my $xmlperl = $self->modsdoc_to_values( $mods );
 386         $xmlperl = $self->mods_values_to_mods_slim( $xmlperl );
 387
 388         # for backwards compatibility, remove the array part when all is decided
 389         if(ref($xmlperl->{subject}) eq 'ARRAY' ) {
 390                 for my $subject( @{$xmlperl->{subject}} ) {
 391                         push @{$self->{master_doc}->{subject}}, $subject;
 392                 }
 393         } else {
 394                 for my $subject ( keys %{$xmlperl->{subject}} ) {
 395                         my $s = $self->{master_doc}->{subject};
 396                         if(defined($s->{$subject})) { $s->{$subject}++; } else { $s->{$subject} = 1; }
 397                 }
 398         }
 399
 400         push( @{$self->{master_doc}->{type_of_resource}},
 401                 $self->get_field_value( $mods, $resource_xpath ));
 402
 403         if(!($self->{master_doc}->{isbn}) ) {
 404                 ($self->{master_doc}->{isbn}) =
 405                         $self->get_field_value( $mods, $isbn_xpath );
 406         }
 407 }
 408
 409
 410 # ---------------------------------------------------------------------------
 411 # Completes a MARC -> Unified MODS batch process and returns the perl hash
 412 # ---------------------------------------------------------------------------
 413 sub init_virtual_record {
 414         my $record = Fieldmapper::metabib::virtual_record->new;
 415         $record->subject([]);
 416         $record->types_of_resource([]);
 417         $record->call_numbers([]);
 418         return $record;
 419 }
 420
 421 sub finish_mods_batch {
 422         my $self = shift;
 423
 424         return undef unless $self->{master_doc};
 425
 426         my $perl = $self->{master_doc};
 427         my $record = init_virtual_record();
 428
 429         # turn the hash into a fieldmapper object
 430         #(my $title = $perl->{title}) =~ s/\[.*?\]//og;
 431         #(my $author = $perl->{author}) =~ s/\(.*?\)//og;
 432         my $title = $perl->{title};
 433         my $author = $perl->{author};
 434
 435         my @series;
 436         for my $s (@{$perl->{series}}) {
 437                 push @series, (split( /\s*;/, $s ))[0];
 438         }
 439
 440         # uniquify the types of resource
 441         my $rtypes = $perl->{type_of_resource};
 442         my %hash = map { ($_ => 1) } @$rtypes;
 443         $rtypes = [ keys %hash ];
 444
 445         $record->title($title);
 446         $record->author($author);
 447
 448         $record->doc_id($perl->{doc_id});
 449         $record->isbn($perl->{isbn});
 450         $record->pubdate($perl->{pubdate});
 451         $record->publisher($perl->{publisher});
 452         $record->tcn($perl->{tcn});
 453
 454         $record->edition($perl->{edition});
 455
 456         $record->subject($perl->{subject});
 457         $record->types_of_resource($rtypes);
 458         $record->series(\@series);
 459
 460         $record->online_loc($perl->{online_loc});
 461         $record->synopsis($perl->{synopsis});
 462         $record->physical_description($perl->{physical_description});
 463         $record->toc($perl->{toc});
 464
 465         $self->{master_doc} = undef;
 466         return $record;
 467 }
 468
 469
 470