8 use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
13 use Fcntl; # for O_RDWR
19 $config_file =~ s/\.pl$/.conf/;
20 die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
24 #use index_DBI; # default DBI module for index
25 use index_DBI_cache; # faster DBI module using memory cache
32 # -m multiple directories
36 getopts('d:m:qs', \%opts);
38 my $path; # this is name of database
40 Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
42 # this is encoding of all files on disk, including import_xml/*.xml file and
43 # filter/*.pm files! It will be used to store strings in perl internally!
44 my $codepage = 'ISO-8859-2';
46 my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
47 # this function will convert data from XML files to local encoding
49 return $utf2cp->convert($_[0]);
52 # decode isis/excel or other import codepage
55 # outgoing xml must be in UTF-8
56 my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
58 # mapping between data type and tag which specify
67 my $cache; # for cacheing
69 # lookup hash (tied to file)
71 # this option will cache all lookup entries in memory.
72 # if you are tight on memory, turn this off
73 my $use_lhash_cache = 1;
81 my $add_xml = shift @_;
82 # needed to read values from configuration file
84 my $database = shift @_;
90 my $html = ""; # html formatted display output
92 my %field_usage; # counter for usage of each field
94 # sort subrouting using order="" attribute
96 my $va = $config->{indexer}->{$a}->{order} ||
97 $config->{indexer}->{$a};
98 my $vb = $config->{indexer}->{$b}->{order} ||
99 $config->{indexer}->{$b};
105 if ($cache->{tags_by_order}->{$type}) {
106 @sorted_tags = @{$cache->{tags_by_order}->{$type}};
108 @sorted_tags = sort by_order keys %{$config->{indexer}};
109 $cache->{tags_by_order}->{$type} = \@sorted_tags;
115 foreach my $field (@sorted_tags) {
118 $field_usage{$field}++;
121 my $swish_exact_data = "";
122 my $display_data = "";
125 my ($swish,$display);
127 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
128 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
130 my $format = x($x->{content});
131 my $delimiter = x($x->{delimiter}) || ' ';
133 my $repeat_off = 0; # repeatable offset
135 # swish, swish_exact, display, index, index_lookup
136 # swish and display defaults
137 my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
138 $s = 0 if (lc($x->{type}) eq "display");
139 $d = 0 if (lc($x->{type}) eq "swish");
140 $se = 1 if (lc($x->{type}) eq "swish_exact");
141 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
142 $il = 1 if (lc($x->{type}) =~ /^lookup/);
145 # what will separate last line from this one?
146 if ($display_data && $x->{append} && $x->{append} eq "1") {
147 $line_delimiter = ' ';
148 } elsif ($display_data) {
149 $line_delimiter = '<br/>';
152 # init vars so that we go into while...
153 ($swish,$display) = (1,1);
155 # placeholder for all repeatable entries for index
159 my $x = shift || die "mkformat needs tag reference";
160 my $data = shift || return;
161 my $format_name = x($x->{format_name}) || return $data;
162 my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
163 my $format_delimiter = x($x->{format_delimiter});
165 if ($format_delimiter) {
166 @data = split(/$format_delimiter/,$data);
172 my $nr = scalar $fmt =~ s/%s/%s/g;
173 if (($#data+1) == $nr) {
174 return sprintf($fmt,@data);
176 print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
180 print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
184 # while because of repeatable fields
185 while ($swish || $display) {
186 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
187 if ($repeat_off > 1000) {
188 print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
192 # is this field is lookup?
193 if ($display && $x->{lookup}) {
194 if ($use_lhash_cache) {
195 if (!defined($cache->{lhash}->{$display})) {
196 my $new_display = $lhash{$display};
198 #print STDERR "lookup cache store '$display' = '$new_display'\n";
199 $display = $new_display;
200 $cache->{lhash}->{$display} = $new_display;
202 print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
204 $cache->{lhash}->{$display} = "";
207 $display = $cache->{lhash}->{$display};
210 $display = $lhash{$display} || "";
214 # filter="name" ; filter this field through
216 my $filter = $x->{filter};
217 if ($filter && !$cache->{filter_loaded}->{$filter}) {
218 require "filter/".$filter.".pm";
219 $cache->{filter_loaded}->{$filter}++;
221 # type="swish" ; field for swish
223 if ($filter && ($s || $se)) {
225 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
226 $swish_data .= $tmp if ($s);
227 $swish_exact_data .= $tmp if ($se);
230 $swish_data .= $swish if ($s);
231 $swish_exact_data .= $swish if ($se);
235 # type="display" ; field for display
236 if ($d && $display) {
237 if ($line_delimiter && $display_data) {
238 $display_data .= $line_delimiter;
239 undef $line_delimiter;
244 $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
246 $display_data = join($delimiter,mkformat($x,&$filter($display)));
250 $display_data .= $delimiter.mkformat($x,$display);
252 $display_data = mkformat($x,$display);
257 # type="index" ; insert into index
258 if ($i && $display) {
261 $display = &$filter($display);
263 if ($x->{append} && @index_data) {
264 $index_data[$#index_data].=$display;
266 push @index_data, $display;
270 # store fields in lookup
271 if ($il && $display) {
272 if (lc($x->{type}) eq "lookup_key") {
274 print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
276 $lookup_key = $display;
278 } elsif (lc($x->{type}) eq "lookup_val") {
280 $lhash{$lookup_key} = $display;
282 print STDERR "WARNING: no lookup_key defined for '$display'?";
289 foreach my $d (@index_data) {
290 $index->insert($field, $d, $path);
294 # now try to parse variables from configuration file
295 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
297 my $delimiter = x($x->{delimiter}) || ' ';
298 my $val = $cfg->val($database, x($x->{content}));
300 my ($s,$d,$i) = (1,1,0); # swish, display default
301 $s = 0 if (lc($x->{type}) eq "display");
302 $d = 0 if (lc($x->{type}) eq "swish");
303 # no support for swish exact in config.
305 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
308 $display_data .= $delimiter.$val if ($d);
309 $swish_data .= $val if ($s);
310 $index->insert($field, $val, $path) if ($i);
318 if ($field eq "headline") {
319 $xml .= xmlify("headline", $display_data);
322 # find field name (signular, plural)
324 if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
325 $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
326 } elsif ($config->{indexer}->{$field}->{name_plural}) {
327 $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
328 } elsif ($config->{indexer}->{$field}->{name}) {
329 $field_name = $config->{indexer}->{$field}->{name}."#-#";
331 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
334 $html .= x($field_name);
336 $html .= $display_data."###\n";
340 # remove extra spaces
341 $swish_data =~ s/ +/ /g;
342 $swish_data =~ s/ +$//g;
344 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
347 if ($swish_exact_data) {
348 $swish_exact_data =~ s/ +/ /g;
349 $swish_exact_data =~ s/ +$//g;
351 # add delimiters before and after word.
352 # That is required to produce exact match
353 $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
359 # dump formatted output in <html>
361 #$xml .= xmlify("html",$html);
362 $xml .= "<html><![CDATA[ $html ]]></html>";
366 $xml .= $add_xml if ($add_xml);
367 return "<xml>\n$xml</xml>\n";
373 ##########################################################################
375 # read configuration for this script
376 my $cfg = new Config::IniFiles( -file => $config_file );
378 # read global.conf configuration
379 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
382 $index = new index_DBI(
383 $cfg_global->val('global', 'dbi_dbd'),
384 $cfg_global->val('global', 'dbi_dsn'),
385 $cfg_global->val('global', 'dbi_user'),
386 $cfg_global->val('global', 'dbi_passwd') || '',
389 my $show_progress = $cfg_global->val('global', 'show_progress');
391 my $unac_filter = $cfg_global->val('global', 'unac_filter');
393 require $unac_filter;
396 foreach my $database ($cfg->Sections) {
398 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
399 my $add_xml = $cfg -> val($database, 'xml'); # optional
401 # create new lookup file
402 my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
404 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
405 tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
406 print STDERR "creating lookup file '$lookup_file'\n";
409 # open existing lookup file
410 $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
412 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
413 tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
414 print STDERR "opening lookup file '$lookup_file'\n";
417 print STDERR "reading ./import_xml/$type.xml\n";
419 # extract just type basic
420 my $type_base = $type;
421 $type_base =~ s/_.+$//g;
423 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
425 # output current progress indicator
428 return if (! $show_progress);
430 my $total = shift || 1;
431 my $p = int($current * 100 / $total);
433 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
440 return if (! $show_progress);
441 my $current = shift @_;
443 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
445 $last_p += $fake_dir;
446 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
447 if ($last_p % 10 == 0) {
448 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
453 print STDERR "using: $type...\n";
455 if ($type_base eq "isis") {
457 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
459 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
460 my $db = OpenIsis::open( $isis_db );
462 # check if .txt database for OpenIsis is zero length,
463 # if so, erase it and re-open database
465 my $isis_db = shift || die "need isis database name";
468 if (-e $isis_db.".TXT") {
469 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
470 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
473 if (-e $isis_db.".PTR") {
474 print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
475 unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
478 return OpenIsis::open( $isis_db ) if ($reopen);
483 $db = check_txt_db($isis_db);
485 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
492 print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
495 } elsif ($db == -5) {
496 print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
499 print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
503 my $max_rowid = OpenIsis::maxRowid( $db );
505 # if 0 records, try to rease isis .txt database
506 if ($max_rowid == 0) {
507 # force removal of database
508 $db = check_txt_db($isis_db);
509 $max_rowid = OpenIsis::maxRowid( $db );
512 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
514 my $path = $database;
516 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
517 my $row = OpenIsis::read( $db, $row_id );
518 if ($row && $row->{mfn}) {
520 progress($row->{mfn}, $max_rowid);
522 my $swishpath = $path."#".int($row->{mfn});
524 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
525 $xml = $cp2utf->convert($xml);
526 use bytes; # as opposed to chars
527 print "Path-Name: $swishpath\n";
528 print "Content-Length: ".(length($xml)+1)."\n";
529 print "Document-Type: XML\n\n$xml\n";
533 # for this to work with current version of OpenIsis (0.9.0)
534 # you might need my patch from
535 # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
536 OpenIsis::close($db);
539 } elsif ($type_base eq "excel") {
540 use Spreadsheet::ParseExcel;
541 use Spreadsheet::ParseExcel::Utility qw(int2col);
543 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
544 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
546 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
547 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
549 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
552 foreach my $oWks (@{$oBook->{Worksheet}}) {
553 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
554 last if ($oWks->{Name} eq $sheet);
558 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
560 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
561 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
562 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
564 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
566 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
567 my $cell = $oWorksheet->{Cells}[$iR][$iC];
569 $row->{int2col($iC)} = $cell->Value;
573 progress($iR, $end_row);
575 # print "row[$iR/$end_row] ";
576 # foreach (keys %{$row}) {
577 # print "$_: ",$row->{$_},"\t";
581 my $swishpath = $database."#".$iR;
585 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
586 $xml = $cp2utf->convert($xml);
587 use bytes; # as opposed to chars
588 print "Path-Name: $swishpath\n";
589 print "Content-Length: ".(length($xml)+1)."\n";
590 print "Document-Type: XML\n\n$xml\n";
593 } elsif ($type_base eq "marc") {
597 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
598 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
600 # optional argument is format
601 my $format = x($config->{format}) || 'usmarc';
603 print STDERR "Reading MARC file '$marc_file'\n";
606 my $nr = $marc->openmarc({
607 file=>$marc_file, format=>$format
608 }) || die "Can't open MARC file '$marc_file'";
610 my $i=0; # record nr.
614 while ($marc->nextmarc(1)) {
619 my $swishpath = $database."#".$i;
621 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
622 $xml = $cp2utf->convert($xml);
623 use bytes; # as opposed to chars
624 print "Path-Name: $swishpath\n";
625 print "Content-Length: ".(length($xml)+1)."\n";
626 print "Document-Type: XML\n\n$xml\n";
629 } elsif ($type_base eq "feed") {
631 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
632 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
634 print STDERR "Reading feed from program '$prog'\n";
636 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
638 my $i=1; # record nr.
647 my $swishpath = $database."#".$i++;
649 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
650 $xml = $cp2utf->convert($xml);
651 use bytes; # as opposed to chars
652 print "Path-Name: $swishpath\n";
653 print "Content-Length: ".(length($xml)+1)."\n";
654 print "Document-Type: XML\n\n$xml\n";
661 $line = $1 if (s/^(\d+):\s*//);
662 $data->{$line++} = $_;
668 untie %lhash if (%lhash);
672 # call this to commit index
677 ##########################################################################
681 all2xml.pl - read various file formats and dump XML for SWISH-E
685 This command will read ISIS data file using OpenIsis perl module, MARC
686 records using MARC module and optionally Micro$oft Excel files to
687 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
688 this script B<isn't general xml generator> from isis files (isis allready
689 has something like that). Output of this script is tailor-made for SWISH-E.
693 Documentation is really lacking. However, in true Open Source spirit, source
694 is best documentation. I even made considerable effort to comment parts
695 which are not intuitively clear, so...
699 Dobrica Pavlinusic <dpavlin@rot13.org>
703 GNU Public License (GPL) v2 or later
707 SWISH-E web site at http://www.swish-e.org