git.rot13.org Git - webpac/blob - all2xml.pl

   1 #!/usr/bin/perl -w
   2
   3 use strict;
   4 use OpenIsis;
   5 use Getopt::Std;
   6 use Data::Dumper;
   7 use XML::Simple;
   8 use Text::Unaccent 1.02;        # 1.01 won't compile on my platform,
   9 use Text::Iconv;
  10 use Config::IniFiles;
  11 use Encode;
  12
  13 $|=1;
  14
  15 my $config_file = $0;
  16 $config_file =~ s/\.pl$/.conf/;
  17 die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
  18
  19 my $config;
  20
  21 #use index_DBI;         # default DBI module for index
  22 use index_DBI_cache;    # faster DBI module using memory cache
  23 my $index;
  24
  25 my %opts;
  26
  27 # usage:
  28 #       -d directory name
  29 #       -m multiple directories
  30 #       -q quiet
  31 #       -s run swish
  32
  33 getopts('d:m:qs', \%opts);
  34
  35 my $path;       # this is name of database
  36
  37 Text::Iconv->raise_error(0);     # Conversion errors don't raise exceptions
  38
  39 # this is encoding of all files on disk, including import_xml/*.xml file and
  40 # filter/*.pm files! It will be used to store strings in perl internally!
  41 my $codepage = 'ISO-8859-2';
  42
  43 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
  44 # this function will convert data from XML files to local encoding
  45 sub x {
  46         return $utf2cp->convert($_[0]);
  47 }
  48
  49 # decode isis/excel or other import codepage
  50 my $import2cp;
  51
  52 # outgoing xml must be in UTF-8
  53 my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
  54
  55 # mapping between data type and tag which specify
  56 # format in XML file
  57 my %type2tag = (
  58         'isis' => 'isis',
  59         'excel' => 'column',
  60         'marc' => 'marc',
  61         'feed' => 'feed'
  62 );
  63
  64 sub data2xml {
  65
  66         use xmlify;
  67
  68         my $type = shift @_;
  69         my $row = shift @_;
  70         my $add_xml = shift @_;
  71         # needed to read values from configuration file
  72         my $cfg = shift @_;
  73         my $database = shift @_;
  74
  75         my $xml;
  76
  77         use parse_format;
  78
  79         my $html = "";          # html formatted display output
  80
  81         my %field_usage;        # counter for usage of each field
  82
  83         # sort subrouting using order="" attribute
  84         sub by_order {
  85                 my $va = $config->{indexer}->{$a}->{order} ||
  86                         $config->{indexer}->{$a};
  87                 my $vb = $config->{indexer}->{$b}->{order} ||
  88                         $config->{indexer}->{$b};
  89
  90                 return $va <=> $vb;
  91         }
  92
  93         foreach my $field (sort by_order keys %{$config->{indexer}}) {
  94
  95                 $field=x($field);
  96                 $field_usage{$field}++;
  97
  98                 my $swish_data = "";
  99                 my $display_data = "";
 100                 my $line_delimiter;
 101
 102                 my ($swish,$display);
 103
 104                 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
 105                 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
 106
 107                         my $format = x($x->{content});
 108                         my $delimiter = x($x->{delimiter}) || ' ';
 109
 110                         my $repeat_off = 0;             # repeatable offset
 111
 112                         my ($s,$d,$i) = (1,1,0);        # swish, display default
 113                         $s = 0 if (lc($x->{type}) eq "display");
 114                         $d = 0 if (lc($x->{type}) eq "swish");
 115                         ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
 116
 117                         # what will separate last line from this one?
 118                         if ($display_data && $x->{append} && $x->{append} eq "1") {
 119                                 $line_delimiter = ' ';
 120                         } elsif ($display_data) {
 121                                 $line_delimiter = '<br/>';
 122                         }
 123
 124                         # init vars so that we go into while...
 125                         ($swish,$display) = (1,1);
 126
 127                         # placeholder for all repeatable entries for index
 128                         my @index_data;
 129                         my $index_filter;
 130
 131                         # while because of repeatable fields
 132                         while ($swish || $display) {
 133                                 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
 134                                 if ($repeat_off > 1000) {
 135                                         print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
 136                                         last;
 137                                 }
 138
 139                                 # filter="name" ; filter this field through
 140                                 # filter/[name].pm
 141                                 my $filter = $x->{filter};
 142                                 if ($filter) {
 143                                         require "filter/".$filter.".pm";
 144                                 }
 145                                 # type="swish" ; field for swish
 146                                 if ($s && $swish) {
 147                                         if ($filter) {
 148                                                 no strict 'refs';
 149                                                 $swish_data .= join(" ",&$filter($swish));
 150                                         } else {
 151                                                 $swish_data .= $swish;
 152                                         }
 153                                 }
 154
 155                                 # type="display" ; field for display
 156                                 if ($d && $display) {
 157                                         if ($line_delimiter && $display_data) {
 158                                                 $display_data .= $line_delimiter;
 159                                                 undef $line_delimiter;
 160                                         }
 161                                         if ($filter) {
 162                                                 no strict 'refs';
 163                                                 $display_data .= $delimiter.&$filter($display);
 164                                         } else {
 165                                                 $display_data .= $delimiter.$display;
 166                                         }
 167                                 }
 168
 169                                 # type="index" ; insert into index
 170                                 if ($i && $display) {
 171                                         push @index_data, $display;
 172                                         $index_filter = $filter if ($filter);
 173                                 }
 174                         }
 175
 176                         # fill data in index
 177                         if (@index_data) {
 178                                 if ($index_filter) {
 179                                         no strict 'refs';
 180                                         foreach my $d (@index_data) {
 181                                                 $index->insert($field, &$index_filter($d), $path);
 182                                         }
 183                                 } else {
 184                                         foreach my $d (@index_data) {
 185                                                 $index->insert($field, $d, $path);
 186                                         }
 187                                 }
 188                         }
 189                 }
 190
 191                 # now try to parse variables from configuration file
 192                 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
 193
 194                         my $delimiter = x($x->{delimiter}) || ' ';
 195                         my $val = $cfg->val($database, x($x->{content}));
 196
 197                         my ($s,$d,$i) = (1,1,0);        # swish, display default
 198                         $s = 0 if (lc($x->{type}) eq "display");
 199                         $d = 0 if (lc($x->{type}) eq "swish");
 200                         ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
 201
 202                         if ($val) {
 203                                 $display_data .= $delimiter.$val if ($d);
 204                                 $swish_data .= $val if ($s);
 205                                 $index->insert($field, $val, $path) if ($i);
 206                         }
 207
 208                 }
 209
 210
 211                 if ($display_data) {
 212
 213                         if ($field eq "headline") {
 214                                 $xml .= xmlify("headline", $display_data);
 215                         } else {
 216
 217                                 # find field name (signular, plural)
 218                                 my $field_name = "";
 219                                 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
 220                                         $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
 221                                 } elsif ($config->{indexer}->{$field}->{name_plural}) {
 222                                         $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
 223                                 } elsif ($config->{indexer}->{$field}->{name}) {
 224                                         $field_name = $config->{indexer}->{$field}->{name}."#-#";
 225                                 } else {
 226                                         print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
 227                                 }
 228                                 if ($field_name) {
 229                                         $html .= x($field_name);
 230                                 }
 231                                 $html .= $display_data."###\n";
 232                         }
 233                 }
 234                 if ($swish_data) {
 235                         # remove extra spaces
 236                         $swish_data =~ s/ +/ /g;
 237                         $swish_data =~ s/ +$//g;
 238
 239                         $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
 240                 }
 241
 242
 243         }
 244
 245         # dump formatted output in <html>
 246         if ($html) {
 247                 #$xml .= xmlify("html",$html);
 248                 $xml .= "<html><![CDATA[ $html ]]></html>";
 249         }
 250
 251         if ($xml) {
 252                 $xml .= $add_xml if ($add_xml);
 253                 return "<xml>\n$xml</xml>\n";
 254         } else {
 255                 return;
 256         }
 257 }
 258
 259 ##########################################################################
 260
 261 # read configuration for this script
 262 my $cfg = new Config::IniFiles( -file => $config_file );
 263
 264 # read global.conf configuration
 265 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
 266
 267 # open index
 268 $index = new index_DBI(
 269                 $cfg_global->val('global', 'dbi_dbd'),
 270                 $cfg_global->val('global', 'dbi_dsn'),
 271                 $cfg_global->val('global', 'dbi_user'),
 272                 $cfg_global->val('global', 'dbi_passwd') || '',
 273         );
 274
 275 my $show_progress = $cfg_global->val('global', 'show_progress');
 276
 277 foreach my $database ($cfg->Sections) {
 278
 279         my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
 280         my $add_xml = $cfg -> val($database, 'xml');    # optional
 281
 282 print STDERR "reading ./import_xml/$type.xml\n";
 283
 284         # extract just type basic
 285         my $type_base = $type;
 286         $type_base =~ s/_.+$//g;
 287
 288         $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
 289
 290         # output current progress indicator
 291         my $last_p = 0;
 292         sub progress {
 293                 return if (! $show_progress);
 294                 my $current = shift;
 295                 my $total = shift || 1;
 296                 my $p = int($current * 100 / $total);
 297                 if ($p != $last_p) {
 298                         printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
 299                         $last_p = $p;
 300                 }
 301         }
 302
 303         my $fake_dir = 1;
 304         sub fakeprogress {
 305                 return if (! $show_progress);
 306                 my $current = shift @_;
 307
 308                 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
 309
 310                 $last_p += $fake_dir;
 311                 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
 312                 if ($last_p % 10 == 0) {
 313                         printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
 314                 }
 315         }
 316
 317         # now read database
 318 print STDERR "using: $type...\n";
 319
 320         if ($type_base eq "isis") {
 321
 322                 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
 323
 324                 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
 325                 my $db = OpenIsis::open( $isis_db );
 326
 327                 # check if .txt database for OpenIsis is zero length,
 328                 # if so, erase it and re-open database
 329                 sub check_txt_db {
 330                         my $isis_db = shift || die "need isis database name";
 331                         my $reopen = 0;
 332
 333                         if (-e $isis_db.".TXT") {
 334                                 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
 335                                 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
 336                                 $reopen++;
 337                         }
 338                         if (-e $isis_db.".PTR") {
 339                                 print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
 340                                 unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
 341                                 $reopen++;
 342                         }
 343                         return OpenIsis::open( $isis_db ) if ($reopen);
 344                 }
 345
 346                 # EOF error
 347                 if ($db == -1) {
 348                         $db = check_txt_db($isis_db);
 349                         if ($db == -1) {
 350                                 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
 351                                 next;
 352                         }
 353                 }
 354
 355                 # OpenIsis::ERR_BADF
 356                 if ($db == -4) {
 357                         print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
 358                         next;
 359                 # OpenIsis::ERR_IO
 360                 } elsif ($db == -5) {
 361                         print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
 362                         next;
 363                 } elsif ($db < 0) {
 364                         print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
 365                         next;
 366                 }
 367
 368                 my $max_rowid = OpenIsis::maxRowid( $db );
 369
 370                 # if 0 records, try to rease isis .txt database
 371                 if ($max_rowid == 0) {
 372                         # force removal of database
 373                         $db = check_txt_db($isis_db);
 374                         $max_rowid = OpenIsis::maxRowid( $db );
 375                 }
 376
 377                 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
 378
 379                 my $path = $database;
 380
 381                 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
 382                         my $row = OpenIsis::read( $db, $row_id );
 383                         if ($row && $row->{mfn}) {
 384
 385                                 progress($row->{mfn}, $max_rowid);
 386
 387                                 my $swishpath = $path."#".int($row->{mfn});
 388
 389                                 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
 390                                         $xml = $cp2utf->convert($xml);
 391                                         use bytes;      # as opposed to chars
 392                                         print "Path-Name: $swishpath\n";
 393                                         print "Content-Length: ".(length($xml)+1)."\n";
 394                                         print "Document-Type: XML\n\n$xml\n";
 395                                 }
 396                         }
 397                 }
 398                 # for this to work with current version of OpenIsis (0.9.0)
 399                 # you might need my patch from
 400                 # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
 401                 OpenIsis::close($db);
 402                 print STDERR "\n";
 403
 404         } elsif ($type_base eq "excel") {
 405                 use Spreadsheet::ParseExcel;
 406                 use Spreadsheet::ParseExcel::Utility qw(int2col);
 407
 408                 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
 409                 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
 410
 411                 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
 412                 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
 413
 414                 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
 415
 416                 my $sheet_nr = 0;
 417                 foreach my $oWks (@{$oBook->{Worksheet}}) {
 418                         #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
 419                         last if ($oWks->{Name} eq $sheet);
 420                         $sheet_nr++;
 421                 }
 422
 423                 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
 424
 425                 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
 426                 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
 427                 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
 428
 429                 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
 430                         my $row;
 431                         for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
 432                                 my $cell = $oWorksheet->{Cells}[$iR][$iC];
 433                                 if ($cell) {
 434                                         $row->{int2col($iC)} = $cell->Value;
 435                                 }
 436                         }
 437
 438                         progress($iR, $end_row);
 439
 440 #                       print "row[$iR/$end_row] ";
 441 #                       foreach (keys %{$row}) {
 442 #                               print "$_: ",$row->{$_},"\t";
 443 #                       }
 444 #                       print "\n";
 445
 446                         my $swishpath = $database."#".$iR;
 447
 448                         next if (! $row);
 449
 450                         if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
 451                                 $xml = $cp2utf->convert($xml);
 452                                 use bytes;      # as opposed to chars
 453                                 print "Path-Name: $swishpath\n";
 454                                 print "Content-Length: ".(length($xml)+1)."\n";
 455                                 print "Document-Type: XML\n\n$xml\n";
 456                         }
 457                 }
 458         } elsif ($type_base eq "marc") {
 459
 460                 use MARC;
 461
 462                 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
 463                 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
 464
 465                 # optional argument is format
 466                 my $format = x($config->{format}) || 'usmarc';
 467
 468                 print STDERR "Reading MARC file '$marc_file'\n";
 469
 470                 my $marc = new MARC;
 471                 my $nr = $marc->openmarc({
 472                                 file=>$marc_file, format=>$format
 473                         }) || die "Can't open MARC file '$marc_file'";
 474
 475                 my $i=0;        # record nr.
 476
 477                 my $rec;
 478
 479                 while ($marc->nextmarc(1)) {
 480
 481                         # XXX
 482                         fakeprogress($i++);
 483
 484                         my $swishpath = $database."#".$i;
 485
 486                         if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
 487                                 $xml = $cp2utf->convert($xml);
 488                                 use bytes;      # as opposed to chars
 489                                 print "Path-Name: $swishpath\n";
 490                                 print "Content-Length: ".(length($xml)+1)."\n";
 491                                 print "Document-Type: XML\n\n$xml\n";
 492                         }
 493                 }
 494         } elsif ($type_base eq "feed") {
 495
 496                 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
 497                 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
 498
 499                 print STDERR "Reading feed from program '$prog'\n";
 500
 501                 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
 502
 503                 my $i=1;        # record nr.
 504
 505                 my $data;
 506                 my $line=1;
 507
 508                 while (<FEED>) {
 509                         chomp;
 510
 511                         if (/^$/) {
 512                                 my $swishpath = $database."#".$i++;
 513
 514                                 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
 515                                         $xml = $cp2utf->convert($xml);
 516                                         use bytes;      # as opposed to chars
 517                                         print "Path-Name: $swishpath\n";
 518                                         print "Content-Length: ".(length($xml)+1)."\n";
 519                                         print "Document-Type: XML\n\n$xml\n";
 520                                 }
 521                                 $line = 1;
 522                                 $data = {};
 523                                 next;
 524                         }
 525
 526                         $line = $1 if (s/^(\d+):\s*//);
 527                         $data->{$line++} = $_;
 528
 529                         fakeprogress($i);
 530
 531                 }
 532         }
 533 }
 534
 535 # call this to commit index
 536 $index->close;
 537
 538 1;
 539 __END__
 540 ##########################################################################
 541
 542 =head1 NAME
 543
 544 all2xml.pl - read various file formats and dump XML for SWISH-E
 545
 546 =head1 DESCRIPTION
 547
 548 This command will read ISIS data file using OpenIsis perl module, MARC
 549 records using MARC module and optionally Micro$oft Excel files to
 550 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
 551 this script B<isn't general xml generator> from isis files (isis allready
 552 has something like that). Output of this script is tailor-made for SWISH-E.
 553
 554 =head1 BUGS
 555
 556 Documentation is really lacking. However, in true Open Source spirit, source
 557 is best documentation. I even made considerable effort to comment parts
 558 which are not intuitively clear, so...
 559
 560 =head1 AUTHOR
 561
 562 Dobrica Pavlinusic <dpavlin@rot13.org>
 563
 564 =head1 COPYRIGHT
 565
 566 GNU Public License (GPL) v2 or later
 567
 568 =head1 SEE ALSO
 569
 570 SWISH-E web site at http://www.swish-e.org
 571
 572 =cut