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 );
  61
  62 sub data2xml {
  63
  64         use xmlify;
  65
  66         my $type = shift @_;
  67         my $row = shift @_;
  68         my $add_xml = shift @_;
  69
  70         my $xml;
  71
  72         use parse_format;
  73
  74         my $html = "";          # html formatted display output
  75
  76         my %field_usage;        # counter for usage of each field
  77
  78         # sort subrouting using order="" attribute
  79         sub by_order {
  80                 return 0 if (! $config->{indexer}->{$a}->{order});
  81                 return 0 if (! $config->{indexer}->{$b}->{order});
  82
  83                 return $config->{indexer}->{$a}->{order} <=>
  84                         $config->{indexer}->{$b}->{order} ;
  85         }
  86
  87         foreach my $field (sort by_order keys %{$config->{indexer}}) {
  88
  89                 $field=x($field);
  90
  91                 $field_usage{$field}++;
  92
  93                 my $swish_data = "";
  94                 my $display_data = "";
  95                 my $line_delimiter;
  96
  97                 my ($swish,$display);
  98
  99                 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
 100                 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
 101
 102                         my $format = x($x->{content});
 103                         my $delimiter = x($x->{delimiter}) || ' ';
 104
 105                         my $repeat_off = 0;             # repeatable offset
 106
 107                         my ($s,$d,$i) = (1,1,0);        # swish, display default
 108                         $s = 0 if (lc($x->{type}) eq "display");
 109                         $d = 0 if (lc($x->{type}) eq "swish");
 110                         ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
 111
 112                         # what will separate last line from this one?
 113                         if ($display_data && $x->{append} && $x->{append} eq "1") {
 114                                 $line_delimiter = ' ';
 115                         } elsif ($display_data) {
 116                                 $line_delimiter = '<br/>';
 117                         }
 118
 119                         # init vars so that we go into while...
 120                         ($swish,$display) = (1,1);
 121
 122                         if ($swish || $display) {
 123                                 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
 124                                 # filter="name" ; filter this field through
 125                                 # filter/[name].pm
 126                                 my $filter = $x->{filter};
 127                                 if ($filter) {
 128                                         require "filter/".$filter.".pm";
 129                                 }
 130                                 # type="swish" ; field for swish
 131                                 if ($s && $swish) {
 132                                         if ($filter) {
 133                                                 no strict 'refs';
 134                                                 $swish_data .= join(" ",&$filter($swish));
 135                                         } else {
 136                                                 $swish_data .= $swish;
 137                                         }
 138                                 }
 139
 140                                 # type="display" ; field for display
 141                                 if ($d && $display) {
 142                                         if ($line_delimiter && $display_data) {
 143                                                 $display_data .= $line_delimiter;
 144                                                 undef $line_delimiter;
 145                                         }
 146                                         if ($filter) {
 147                                                 no strict 'refs';
 148                                                 $display_data .= join($delimiter,&$filter($display));
 149                                         } else {
 150                                                 if ($display_data) {
 151                                                         $display_data .= $delimiter.$display;
 152                                                 } else {
 153                                                         $display_data .= $display;
 154                                                 }
 155                                         }
 156                                 }
 157
 158                                 # type="index" ; insert into index
 159                                 if ($i && $display) {
 160                                         my $index_data = $display;
 161                                         if ($filter) {
 162                                                 no strict 'refs';
 163                                                 foreach my $d (&$filter($index_data)) {
 164                                                         $index->insert($field, $d, $path);
 165                                                 }
 166                                         } else {
 167                                                 $index->insert($field, $index_data, $path);
 168                                         }
 169                                 }
 170                         }
 171                 }
 172
 173
 174                 if ($display_data) {
 175
 176                         if ($field eq "headline") {
 177                                 $xml .= xmlify("headline", $display_data);
 178                         } else {
 179
 180                                 # find field name (signular, plural)
 181                                 my $field_name = "";
 182                                 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
 183                                         $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
 184                                 } elsif ($config->{indexer}->{$field}->{name_plural}) {
 185                                         $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
 186                                 } elsif ($config->{indexer}->{$field}->{name}) {
 187                                         $field_name = $config->{indexer}->{$field}->{name}."#-#";
 188                                 } else {
 189                                         print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
 190                                 }
 191                                 if ($field_name) {
 192                                         $html .= x($field_name);
 193                                 }
 194                                 $html .= $display_data."###\n";
 195                         }
 196                 }
 197                 if ($swish_data) {
 198                         # remove extra spaces
 199                         $swish_data =~ s/ +/ /g;
 200                         $swish_data =~ s/ +$//g;
 201
 202                         $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
 203                 }
 204
 205
 206         }
 207
 208         # dump formatted output in <html>
 209         if ($html) {
 210                 $xml .= xmlify("html",$html);
 211         }
 212
 213         if ($xml) {
 214                 $xml .= $add_xml if ($add_xml);
 215                 return "<xml>\n$xml</xml>\n";
 216         } else {
 217                 return;
 218         }
 219 }
 220
 221 ##########################################################################
 222
 223 # read configuration for this script
 224 my $cfg = new Config::IniFiles( -file => $config_file );
 225
 226 # read global.conf configuration
 227 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
 228
 229 # open index
 230 $index = new index_DBI(
 231                 $cfg_global->val('global', 'dbi_dbd'),
 232                 $cfg_global->val('global', 'dbi_dsn'),
 233                 $cfg_global->val('global', 'dbi_user'),
 234                 $cfg_global->val('global', 'dbi_passwd') || '',
 235         );
 236
 237 foreach my $database ($cfg->Sections) {
 238
 239         my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
 240         my $add_xml = $cfg -> val($database, 'xml');    # optional
 241
 242 print STDERR "reading ./import_xml/$type.xml\n";
 243
 244         # extract just type basic
 245         my $type_base = $type;
 246         $type_base =~ s/_.+$//g;
 247
 248         $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base} ], forcecontent => 1);
 249
 250         # output current progress indicator
 251         my $last_p = 0;
 252         sub progress {
 253                 #return if (! $opts{q});        # FIXME
 254                 my $current = shift;
 255                 my $total = shift || 1;
 256                 my $p = int($current * 100 / $total);
 257                 if ($p != $last_p) {
 258                         printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
 259                         $last_p = $p;
 260                 }
 261         }
 262
 263         # now read database
 264 print STDERR "using: $type...\n";
 265
 266         if ($type_base eq "isis") {
 267
 268                 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
 269
 270                 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
 271                 my $db = OpenIsis::open( $isis_db );
 272
 273                 my $max_rowid = OpenIsis::maxRowid( $db );
 274
 275                 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
 276
 277                 my $path = $database;
 278
 279                 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
 280                         my $row = OpenIsis::read( $db, $row_id );
 281                         if ($row && $row->{mfn}) {
 282
 283                                 progress($row->{mfn}, $max_rowid);
 284
 285                                 my $swishpath = $path."#".int($row->{mfn});
 286
 287                                 if (my $xml = data2xml($type_base,$row,$add_xml)) {
 288                                         $xml = $cp2utf->convert($xml);
 289                                         use bytes;      # as opposed to chars
 290                                         print "Path-Name: $swishpath\n";
 291                                         print "Content-Length: ".(length($xml)+1)."\n";
 292                                         print "Document-Type: XML\n\n$xml\n";
 293                                 }
 294                         }
 295                 }
 296                 print STDERR "\n";
 297
 298         } elsif ($type_base eq "excel") {
 299                 use Spreadsheet::ParseExcel;
 300                 use Spreadsheet::ParseExcel::Utility qw(int2col);
 301
 302                 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
 303                 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
 304
 305                 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
 306                 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
 307
 308                 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
 309
 310                 my $sheet_nr = 0;
 311                 foreach my $oWks (@{$oBook->{Worksheet}}) {
 312                         #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
 313                         last if ($oWks->{Name} eq $sheet);
 314                         $sheet_nr++;
 315                 }
 316
 317                 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
 318
 319                 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
 320                 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
 321                 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
 322
 323                 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
 324                         my $row;
 325                         for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
 326                                 my $cell = $oWorksheet->{Cells}[$iR][$iC];
 327                                 if ($cell) {
 328                                         $row->{int2col($iC)} = $cell->Value;
 329                                 }
 330                         }
 331
 332                         progress($iR, $end_row);
 333
 334 #                       print "row[$iR/$end_row] ";
 335 #                       foreach (keys %{$row}) {
 336 #                               print "$_: ",$row->{$_},"\t";
 337 #                       }
 338 #                       print "\n";
 339
 340                         my $swishpath = $database."#".$iR;
 341
 342                         next if (! $row);
 343
 344                         if (my $xml = data2xml($type,$row,$add_xml)) {
 345                                 $xml = $cp2utf->convert($xml);
 346                                 use bytes;      # as opposed to chars
 347                                 print "Path-Name: $swishpath\n";
 348                                 print "Content-Length: ".(length($xml)+1)."\n";
 349                                 print "Document-Type: XML\n\n$xml\n";
 350                         }
 351                 }
 352         }
 353 }
 354
 355 # call this to commit index
 356 $index->close;
 357
 358 1;
 359 __END__
 360 ##########################################################################
 361
 362 =head1 NAME
 363
 364 isis2xml.pl - read isis file and dump XML
 365
 366 =head1 DESCRIPTION
 367
 368 This command will read ISIS data file using OpenIsis perl module and
 369 create XML file for usage with I<SWISH-E>
 370 indexer. Dispite it's name, this script B<isn't general xml generator>
 371 from isis files (isis allready has something like that). Output of this
 372 script is tailor-made for SWISH-E.
 373
 374 =head1 AUTHOR
 375
 376 Dobrica Pavlinusic <dpavlin@rot13.org>
 377
 378 =head1 COPYRIGHT
 379
 380 GNU Public License (GPL) v2 or later
 381
 382 =head1 SEE ALSO
 383
 384 SWISH-E web site at http://www.swish-e.org
 385
 386 =cut