8 use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
16 $config_file =~ s/\.pl$/.conf/;
17 die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
21 #use index_DBI; # default DBI module for index
22 use index_DBI_cache; # faster DBI module using memory cache
29 # -m multiple directories
33 getopts('d:m:qs', \%opts);
35 my $path; # this is name of database
37 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
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';
43 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
44 # this function will convert data from XML files to local encoding
46 return $utf2cp->convert($_[0]);
49 # decode isis/excel or other import codepage
52 # outgoing xml must be in UTF-8
53 my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
55 # mapping between data type and tag which specify
70 my $add_xml = shift @_;
71 # needed to read values from configuration file
73 my $database = shift @_;
79 my $html = ""; # html formatted display output
81 my %field_usage; # counter for usage of each field
83 # sort subrouting using order="" attribute
85 return 0 if (! $config->{indexer}->{$a}->{order});
86 return 0 if (! $config->{indexer}->{$b}->{order});
88 return $config->{indexer}->{$a}->{order} <=>
89 $config->{indexer}->{$b}->{order} ;
92 foreach my $field (sort by_order keys %{$config->{indexer}}) {
96 $field_usage{$field}++;
99 my $display_data = "";
102 my ($swish,$display);
104 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
105 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
107 my $format = x($x->{content});
108 my $delimiter = x($x->{delimiter}) || ' ';
110 my $repeat_off = 0; # repeatable offset
112 my ($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");
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 $line_delimiter = '<br/>';
124 # init vars so that we go into while...
125 ($swish,$display) = (1,1);
127 if ($swish || $display) {
128 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
129 # filter="name" ; filter this field through
131 my $filter = $x->{filter};
133 require "filter/".$filter.".pm";
135 # type="swish" ; field for swish
139 $swish_data .= join(" ",&$filter($swish));
141 $swish_data .= $swish;
145 # type="display" ; field for display
146 if ($d && $display) {
147 if ($line_delimiter && $display_data) {
148 $display_data .= $line_delimiter;
149 undef $line_delimiter;
153 $display_data .= join($delimiter,&$filter($display));
156 $display_data .= $delimiter.$display;
158 $display_data .= $display;
163 # type="index" ; insert into index
164 if ($i && $display) {
165 my $index_data = $display;
168 foreach my $d (&$filter($index_data)) {
169 $index->insert($field, $d, $path);
172 $index->insert($field, $index_data, $path);
178 # now try to parse variables from configuration file
179 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
181 my $delimiter = x($x->{delimiter}) || ' ';
182 my $val = $cfg->val($database, x($x->{content}));
184 my ($s,$d,$i) = (1,1,0); # swish, display default
185 $s = 0 if (lc($x->{type}) eq "display");
186 $d = 0 if (lc($x->{type}) eq "swish");
187 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
190 $display_data .= $delimiter.$val if ($d);
191 $swish_data .= $val if ($s);
192 $index->insert($field, $val, $path) if ($i);
200 if ($field eq "headline") {
201 $xml .= xmlify("headline", $display_data);
204 # find field name (signular, plural)
206 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
207 $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
208 } elsif ($config->{indexer}->{$field}->{name_plural}) {
209 $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
210 } elsif ($config->{indexer}->{$field}->{name}) {
211 $field_name = $config->{indexer}->{$field}->{name}."#-#";
213 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
216 $html .= x($field_name);
218 $html .= $display_data."###\n";
222 # remove extra spaces
223 $swish_data =~ s/ +/ /g;
224 $swish_data =~ s/ +$//g;
226 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
232 # dump formatted output in <html>
234 #$xml .= xmlify("html",$html);
235 $xml .= "<html><![CDATA[ $html ]]></html>";
239 $xml .= $add_xml if ($add_xml);
240 return "<xml>\n$xml</xml>\n";
246 ##########################################################################
248 # read configuration for this script
249 my $cfg = new Config::IniFiles( -file => $config_file );
251 # read global.conf configuration
252 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
255 $index = new index_DBI(
256 $cfg_global->val('global', 'dbi_dbd'),
257 $cfg_global->val('global', 'dbi_dsn'),
258 $cfg_global->val('global', 'dbi_user'),
259 $cfg_global->val('global', 'dbi_passwd') || '',
262 foreach my $database ($cfg->Sections) {
264 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
265 my $add_xml = $cfg -> val($database, 'xml'); # optional
267 print STDERR "reading ./import_xml/$type.xml\n";
269 # extract just type basic
270 my $type_base = $type;
271 $type_base =~ s/_.+$//g;
273 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
275 # output current progress indicator
278 #return if (! $opts{q}); # FIXME
280 my $total = shift || 1;
281 my $p = int($current * 100 / $total);
283 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
290 my $current = shift @_;
292 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
294 $last_p += $fake_dir;
295 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
296 if ($last_p % 10 == 0) {
297 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
302 print STDERR "using: $type...\n";
304 if ($type_base eq "isis") {
306 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
308 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
309 my $db = OpenIsis::open( $isis_db );
311 my $max_rowid = OpenIsis::maxRowid( $db );
313 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
315 my $path = $database;
317 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
318 my $row = OpenIsis::read( $db, $row_id );
319 if ($row && $row->{mfn}) {
321 progress($row->{mfn}, $max_rowid);
323 my $swishpath = $path."#".int($row->{mfn});
325 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
326 $xml = $cp2utf->convert($xml);
327 use bytes; # as opposed to chars
328 print "Path-Name: $swishpath\n";
329 print "Content-Length: ".(length($xml)+1)."\n";
330 print "Document-Type: XML\n\n$xml\n";
336 } elsif ($type_base eq "excel") {
337 use Spreadsheet::ParseExcel;
338 use Spreadsheet::ParseExcel::Utility qw(int2col);
340 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
341 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
343 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
344 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
346 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
349 foreach my $oWks (@{$oBook->{Worksheet}}) {
350 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
351 last if ($oWks->{Name} eq $sheet);
355 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
357 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
358 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
359 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
361 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
363 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
364 my $cell = $oWorksheet->{Cells}[$iR][$iC];
366 $row->{int2col($iC)} = $cell->Value;
370 progress($iR, $end_row);
372 # print "row[$iR/$end_row] ";
373 # foreach (keys %{$row}) {
374 # print "$_: ",$row->{$_},"\t";
378 my $swishpath = $database."#".$iR;
382 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
383 $xml = $cp2utf->convert($xml);
384 use bytes; # as opposed to chars
385 print "Path-Name: $swishpath\n";
386 print "Content-Length: ".(length($xml)+1)."\n";
387 print "Document-Type: XML\n\n$xml\n";
390 } elsif ($type_base eq "marc") {
394 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
395 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
397 # optional argument is format
398 my $format = x($config->{format}) || 'usmarc';
400 print STDERR "Reading MARC file '$marc_file'\n";
403 my $nr = $marc->openmarc({
404 file=>$marc_file, format=>$format
405 }) || die "Can't open MARC file '$marc_file'";
407 my $i=0; # record nr.
411 while ($marc->nextmarc(1)) {
416 my $swishpath = $database."#".$i;
418 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
419 $xml = $cp2utf->convert($xml);
420 use bytes; # as opposed to chars
421 print "Path-Name: $swishpath\n";
422 print "Content-Length: ".(length($xml)+1)."\n";
423 print "Document-Type: XML\n\n$xml\n";
426 } elsif ($type_base eq "feed") {
428 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
429 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
431 print STDERR "Reading feed from program '$prog'\n";
433 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
435 my $i=1; # record nr.
444 my $swishpath = $database."#".$i++;
446 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
447 $xml = $cp2utf->convert($xml);
448 use bytes; # as opposed to chars
449 print "Path-Name: $swishpath\n";
450 print "Content-Length: ".(length($xml)+1)."\n";
451 print "Document-Type: XML\n\n$xml\n";
458 $line = $1 if (s/^(\d+):\s*//);
459 $data->{$line++} = $_;
467 # call this to commit index
472 ##########################################################################
476 all2xml.pl - read various file formats and dump XML for SWISH-E
480 This command will read ISIS data file using OpenIsis perl module, MARC
481 records using MARC module and optionally Micro$oft Excel files to
482 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
483 this script B<isn't general xml generator> from isis files (isis allready
484 has something like that). Output of this script is tailor-made for SWISH-E.
488 Documentation is really lacking. However, in true Open Source spirit, source
489 is best documentation. I even made considerable effort to comment parts
490 which are not intuitively clear, so...
494 Dobrica Pavlinusic <dpavlin@rot13.org>
498 GNU Public License (GPL) v2 or later
502 SWISH-E web site at http://www.swish-e.org