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                                                 $display_data .= join($delimiter,&$filter($display));
 164                                         } else {
 165                                                 if ($display_data) {
 166                                                         $display_data .= $delimiter.$display;
 167                                                 } else {
 168                                                         $display_data .= $display;
 169                                                 }
 170                                         }
 171                                 }
 172
 173                                 # type="index" ; insert into index
 174                                 if ($i && $display) {
 175                                         push @index_data, $display;
 176                                         $index_filter = $filter if ($filter);
 177                                 }
 178                         }
 179
 180                         # fill data in index
 181                         if (@index_data) {
 182                                 if ($index_filter) {
 183                                         no strict 'refs';
 184                                         foreach my $d (&$index_filter(@index_data)) {
 185                                                 $index->insert($field, $d, $path);
 186                                         }
 187                                 } else {
 188                                         foreach my $d (@index_data) {
 189                                                 $index->insert($field, $d, $path);
 190                                         }
 191                                 }
 192                         }
 193                 }
 194
 195                 # now try to parse variables from configuration file
 196                 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
 197
 198                         my $delimiter = x($x->{delimiter}) || ' ';
 199                         my $val = $cfg->val($database, x($x->{content}));
 200
 201                         my ($s,$d,$i) = (1,1,0);        # swish, display default
 202                         $s = 0 if (lc($x->{type}) eq "display");
 203                         $d = 0 if (lc($x->{type}) eq "swish");
 204                         ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
 205
 206                         if ($val) {
 207                                 $display_data .= $delimiter.$val if ($d);
 208                                 $swish_data .= $val if ($s);
 209                                 $index->insert($field, $val, $path) if ($i);
 210                         }
 211
 212                 }
 213
 214
 215                 if ($display_data) {
 216
 217                         if ($field eq "headline") {
 218                                 $xml .= xmlify("headline", $display_data);
 219                         } else {
 220
 221                                 # find field name (signular, plural)
 222                                 my $field_name = "";
 223                                 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
 224                                         $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
 225                                 } elsif ($config->{indexer}->{$field}->{name_plural}) {
 226                                         $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
 227                                 } elsif ($config->{indexer}->{$field}->{name}) {
 228                                         $field_name = $config->{indexer}->{$field}->{name}."#-#";
 229                                 } else {
 230                                         print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
 231                                 }
 232                                 if ($field_name) {
 233                                         $html .= x($field_name);
 234                                 }
 235                                 $html .= $display_data."###\n";
 236                         }
 237                 }
 238                 if ($swish_data) {
 239                         # remove extra spaces
 240                         $swish_data =~ s/ +/ /g;
 241                         $swish_data =~ s/ +$//g;
 242
 243                         $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
 244                 }
 245
 246
 247         }
 248
 249         # dump formatted output in <html>
 250         if ($html) {
 251                 #$xml .= xmlify("html",$html);
 252                 $xml .= "<html><![CDATA[ $html ]]></html>";
 253         }
 254
 255         if ($xml) {
 256                 $xml .= $add_xml if ($add_xml);
 257                 return "<xml>\n$xml</xml>\n";
 258         } else {
 259                 return;
 260         }
 261 }
 262
 263 ##########################################################################
 264
 265 # read configuration for this script
 266 my $cfg = new Config::IniFiles( -file => $config_file );
 267
 268 # read global.conf configuration
 269 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
 270
 271 # open index
 272 $index = new index_DBI(
 273                 $cfg_global->val('global', 'dbi_dbd'),
 274                 $cfg_global->val('global', 'dbi_dsn'),
 275                 $cfg_global->val('global', 'dbi_user'),
 276                 $cfg_global->val('global', 'dbi_passwd') || '',
 277         );
 278
 279 my $show_progress = $cfg_global->val('global', 'show_progress');
 280
 281 foreach my $database ($cfg->Sections) {
 282
 283         my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
 284         my $add_xml = $cfg -> val($database, 'xml');    # optional
 285
 286 print STDERR "reading ./import_xml/$type.xml\n";
 287
 288         # extract just type basic
 289         my $type_base = $type;
 290         $type_base =~ s/_.+$//g;
 291
 292         $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
 293
 294         # output current progress indicator
 295         my $last_p = 0;
 296         sub progress {
 297                 return if (! $show_progress);
 298                 my $current = shift;
 299                 my $total = shift || 1;
 300                 my $p = int($current * 100 / $total);
 301                 if ($p != $last_p) {
 302                         printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
 303                         $last_p = $p;
 304                 }
 305         }
 306
 307         my $fake_dir = 1;
 308         sub fakeprogress {
 309                 return if (! $show_progress);
 310                 my $current = shift @_;
 311
 312                 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
 313
 314                 $last_p += $fake_dir;
 315                 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
 316                 if ($last_p % 10 == 0) {
 317                         printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
 318                 }
 319         }
 320
 321         # now read database
 322 print STDERR "using: $type...\n";
 323
 324         if ($type_base eq "isis") {
 325
 326                 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
 327
 328                 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
 329                 my $db = OpenIsis::open( $isis_db );
 330
 331                 # check if .txt database for OpenIsis is zero length,
 332                 # if so, erase it and re-open database
 333                 sub check_txt_db {
 334                         my $isis_db = shift || die "need isis database name";
 335                         if (-e $isis_db.".TXT") {
 336                                 print STDERR "WARNING: removing .txt OpenIsis database...\n";
 337                                 unlink $isis_db.".TXT" || warn "unlink error on '$isis_db.TXT': $!";
 338                                 my $db = OpenIsis::open( $isis_db );
 339                                 return $db;
 340                         }
 341                 }
 342
 343                 # EOF error
 344                 if ($db == -1) {
 345                         $db = check_txt_db($isis_db);
 346                         if (! $db) {
 347                                 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
 348                                 next;
 349                         }
 350                 }
 351
 352                 # OpenIsis::ERR_BADF
 353                 if ($db == -4) {
 354                         print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
 355                         next;
 356                 # OpenIsis::ERR_IO
 357                 } elsif ($db == -5) {
 358                         print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
 359                         next;
 360                 } elsif ($db < 0) {
 361                         print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
 362                         next;
 363                 }
 364
 365                 my $max_rowid = OpenIsis::maxRowid( $db );
 366
 367                 # if 0 records, try to rease isis .txt database
 368                 if ($max_rowid == 0) {
 369                         # force removal of database
 370                         $db = check_txt_db($isis_db);
 371                         $max_rowid = OpenIsis::maxRowid( $db );
 372                 }
 373
 374                 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
 375
 376                 my $path = $database;
 377
 378                 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
 379                         my $row = OpenIsis::read( $db, $row_id );
 380                         if ($row && $row->{mfn}) {
 381
 382                                 progress($row->{mfn}, $max_rowid);
 383
 384                                 my $swishpath = $path."#".int($row->{mfn});
 385
 386                                 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
 387                                         $xml = $cp2utf->convert($xml);
 388                                         use bytes;      # as opposed to chars
 389                                         print "Path-Name: $swishpath\n";
 390                                         print "Content-Length: ".(length($xml)+1)."\n";
 391                                         print "Document-Type: XML\n\n$xml\n";
 392                                 }
 393                         }
 394                 }
 395                 print STDERR "\n";
 396
 397         } elsif ($type_base eq "excel") {
 398                 use Spreadsheet::ParseExcel;
 399                 use Spreadsheet::ParseExcel::Utility qw(int2col);
 400
 401                 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
 402                 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
 403
 404                 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
 405                 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
 406
 407                 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
 408
 409                 my $sheet_nr = 0;
 410                 foreach my $oWks (@{$oBook->{Worksheet}}) {
 411                         #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
 412                         last if ($oWks->{Name} eq $sheet);
 413                         $sheet_nr++;
 414                 }
 415
 416                 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
 417
 418                 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
 419                 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
 420                 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
 421
 422                 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
 423                         my $row;
 424                         for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
 425                                 my $cell = $oWorksheet->{Cells}[$iR][$iC];
 426                                 if ($cell) {
 427                                         $row->{int2col($iC)} = $cell->Value;
 428                                 }
 429                         }
 430
 431                         progress($iR, $end_row);
 432
 433 #                       print "row[$iR/$end_row] ";
 434 #                       foreach (keys %{$row}) {
 435 #                               print "$_: ",$row->{$_},"\t";
 436 #                       }
 437 #                       print "\n";
 438
 439                         my $swishpath = $database."#".$iR;
 440
 441                         next if (! $row);
 442
 443                         if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
 444                                 $xml = $cp2utf->convert($xml);
 445                                 use bytes;      # as opposed to chars
 446                                 print "Path-Name: $swishpath\n";
 447                                 print "Content-Length: ".(length($xml)+1)."\n";
 448                                 print "Document-Type: XML\n\n$xml\n";
 449                         }
 450                 }
 451         } elsif ($type_base eq "marc") {
 452
 453                 use MARC;
 454
 455                 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
 456                 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
 457
 458                 # optional argument is format
 459                 my $format = x($config->{format}) || 'usmarc';
 460
 461                 print STDERR "Reading MARC file '$marc_file'\n";
 462
 463                 my $marc = new MARC;
 464                 my $nr = $marc->openmarc({
 465                                 file=>$marc_file, format=>$format
 466                         }) || die "Can't open MARC file '$marc_file'";
 467
 468                 my $i=0;        # record nr.
 469
 470                 my $rec;
 471
 472                 while ($marc->nextmarc(1)) {
 473
 474                         # XXX
 475                         fakeprogress($i++);
 476
 477                         my $swishpath = $database."#".$i;
 478
 479                         if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
 480                                 $xml = $cp2utf->convert($xml);
 481                                 use bytes;      # as opposed to chars
 482                                 print "Path-Name: $swishpath\n";
 483                                 print "Content-Length: ".(length($xml)+1)."\n";
 484                                 print "Document-Type: XML\n\n$xml\n";
 485                         }
 486                 }
 487         } elsif ($type_base eq "feed") {
 488
 489                 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
 490                 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
 491
 492                 print STDERR "Reading feed from program '$prog'\n";
 493
 494                 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
 495
 496                 my $i=1;        # record nr.
 497
 498                 my $data;
 499                 my $line=1;
 500
 501                 while (<FEED>) {
 502                         chomp;
 503
 504                         if (/^$/) {
 505                                 my $swishpath = $database."#".$i++;
 506
 507                                 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
 508                                         $xml = $cp2utf->convert($xml);
 509                                         use bytes;      # as opposed to chars
 510                                         print "Path-Name: $swishpath\n";
 511                                         print "Content-Length: ".(length($xml)+1)."\n";
 512                                         print "Document-Type: XML\n\n$xml\n";
 513                                 }
 514                                 $line = 1;
 515                                 $data = {};
 516                                 next;
 517                         }
 518
 519                         $line = $1 if (s/^(\d+):\s*//);
 520                         $data->{$line++} = $_;
 521
 522                         fakeprogress($i);
 523
 524                 }
 525         }
 526 }
 527
 528 # call this to commit index
 529 $index->close;
 530
 531 1;
 532 __END__
 533 ##########################################################################
 534
 535 =head1 NAME
 536
 537 all2xml.pl - read various file formats and dump XML for SWISH-E
 538
 539 =head1 DESCRIPTION
 540
 541 This command will read ISIS data file using OpenIsis perl module, MARC
 542 records using MARC module and optionally Micro$oft Excel files to
 543 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
 544 this script B<isn't general xml generator> from isis files (isis allready
 545 has something like that). Output of this script is tailor-made for SWISH-E.
 546
 547 =head1 BUGS
 548
 549 Documentation is really lacking. However, in true Open Source spirit, source
 550 is best documentation. I even made considerable effort to comment parts
 551 which are not intuitively clear, so...
 552
 553 =head1 AUTHOR
 554
 555 Dobrica Pavlinusic <dpavlin@rot13.org>
 556
 557 =head1 COPYRIGHT
 558
 559 GNU Public License (GPL) v2 or later
 560
 561 =head1 SEE ALSO
 562
 563 SWISH-E web site at http://www.swish-e.org
 564
 565 =cut