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 ($cache->{tags_by_order}) {
106 @sorted_tags = @{$cache->{tags_by_order}};
108 @sorted_tags = sort by_order keys %{$config->{indexer}};
109 $cache->{tags_by_order} = \@sorted_tags;
115 # cache for field in pages
116 delete $cache->{display_data};
117 delete $cache->{swish_data};
118 delete $cache->{swish_exact_data};
119 delete $cache->{index_data};
120 my @page_fields; # names of fields
123 # subs used to produce output
125 sub get_field_name($$$) {
126 my ($config,$field,$field_usage) = @_;
128 # find field name (signular, plural)
130 if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
131 $field_name = $config->{indexer}->{$field}->{name_singular};
132 } elsif ($config->{indexer}->{$field}->{name_plural}) {
133 $field_name = $config->{indexer}->{$field}->{name_plural};
134 } elsif ($config->{indexer}->{$field}->{name}) {
135 $field_name = $config->{indexer}->{$field}->{name};
137 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
140 return x($field_name);
145 # begin real work: go field by field
146 foreach my $field (@sorted_tags) {
149 $field_usage{$field}++;
152 my $swish_exact_data = "";
153 my $display_data = "";
157 my ($swish,$display);
159 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
161 # is this field page-by-page?
162 my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
163 push @page_fields,$field if ($iterate_by_page);
165 # default line_delimiter if using
166 my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
168 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
170 my $format = x($x->{content});
171 my $delimiter = x($x->{delimiter}) || ' ';
173 my $repeat_off = 0; # init repeatable offset
175 # swish, swish_exact, display, index, index_lookup
176 # swish and display defaults
177 my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
178 $s = 0 if (lc($x->{type}) eq "display");
179 $d = 0 if (lc($x->{type}) eq "swish");
180 ($s,$se,$d,$i) = (0,0,0,1) if (lc($x->{type}) eq "index");
181 ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
182 ($s,$se,$d,$i,$il) = (0,1,0,0,1) if (lc($x->{type}) =~ /^lookup/);
184 # what will separate last line from this one?
185 if ($display_data && $x->{append}) {
186 $line_delimiter = ' ';
187 } elsif ($display_data) {
188 $line_delimiter = '<br/>';
191 # init vars so that we go into while...
192 ($swish,$display) = (1,1);
194 # placeholder for all repeatable entries for index
197 my $x = shift || die "mkformat needs tag reference";
198 my $data = shift || return;
199 my $format_name = x($x->{format_name}) || return $data;
200 my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
201 my $format_delimiter = x($x->{format_delimiter});
203 if ($format_delimiter) {
204 @data = split(/$format_delimiter/,$data);
210 my $nr = scalar $fmt =~ s/%s/%s/g;
211 if (($#data+1) == $nr) {
212 return sprintf($fmt,@data);
214 print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
218 print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
222 # while because of repeatable fields
223 while ($swish || $display) {
224 my $page = $repeat_off;
225 $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
226 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
227 if ($repeat_off > 1000) {
228 print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
232 # is this field is lookup?
233 if ($display && $x->{lookup}) {
234 my $null = "<!-- null -->";
235 if ($use_lhash_cache) {
236 if (!defined($cache->{lhash}->{$display})) {
237 my $new_display = $lhash{$display};
238 if (defined($new_display)) {
239 #print STDERR "lookup cache store '$display' = '$new_display'\n";
240 $display = $new_display;
241 $cache->{lhash}->{$display} = $new_display;
243 print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
245 $cache->{lhash}->{$display} = $null;
248 $display = $cache->{lhash}->{$display};
251 $display = $lhash{$display} || $null;
255 # filter="name" ; filter this field through
257 my $filter = $x->{filter};
258 if ($filter && !$cache->{filter_loaded}->{$filter}) {
259 require "filter/".$filter.".pm";
260 $cache->{filter_loaded}->{$filter}++;
262 # type="swish" ; field for swish
264 if ($filter && ($s || $se)) {
266 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
267 $swish_data .= $tmp if ($s);
268 $swish_exact_data .= $tmp if ($se);
271 $swish_data .= $swish if ($s);
272 $swish_exact_data .= $swish if ($se);
276 # type="display" ; field for display
277 if ($d && $display) {
278 if ($line_delimiter && $display_data) {
279 $display_data .= $line_delimiter;
284 $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
286 $display_data = join($delimiter,mkformat($x,&$filter($display)));
290 $display_data .= $delimiter.mkformat($x,$display);
292 $display_data = mkformat($x,$display);
297 # type="index" ; insert into index
299 if ($i && $display) {
300 $idisplay = $display;
303 $idisplay = &$filter($idisplay);
305 push @index_data, $idisplay if (! $iterate_by_page);
308 # store fields in lookup
309 if ($il && $display) {
310 if (lc($x->{type}) eq "lookup_key") {
312 print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
314 $lookup_key = $display;
316 } elsif (lc($x->{type}) eq "lookup_val") {
318 $lhash{$lookup_key} = $display;
320 print STDERR "WARNING: no lookup_key defined for '$display'?";
326 # store data for page-by-page repeatable fields
327 if ($iterate_by_page) {
328 sub iterate_fld($$$$$$) {
329 my ($cache,$what,$field,$page,$data,$append) = @_;
332 my $ldel = $page_line_delimiter;
333 $ldel = " " if ($append);
334 #print STDERR "line delimiter: ",Dumper($ldel) if ($ldel);
335 if (! $cache->{$what}->{$field}->[$page]) {
336 $cache->{$what}->{$field}->[$page] = $data;
338 $cache->{$what}->{$field}->[$page] .= $ldel.$data;
343 iterate_fld($cache,'display_data',$field,$page,$display_data,$x->{append});
347 iterate_fld($cache,'swish_data',$field,$page,$swish_data,$x->{append});
350 if ($swish_exact_data) {
351 iterate_fld($cache,'swish_exact_data',$field,$page,$swish_exact_data,$x->{append});
352 $swish_exact_data = "";
356 my $ldel=$page_line_delimiter;
358 if ($cache->{index_data}->{$field}->[$page]) {
360 @index_data = @{$cache->{index_data}->{$field}->[$page]};
363 $index_data[$#index_data] .= $idisplay;
365 push @index_data, $idisplay;
368 @{$cache->{index_data}->{$field}->[$page]} = @index_data;
373 if (! $iterate_by_page) {
375 foreach my $d (@index_data) {
376 $index->insert($field, $d, $path);
382 # now try to parse variables from configuration file
383 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
385 my $delimiter = x($x->{delimiter}) || ' ';
386 my $val = $cfg->val($database, x($x->{content}));
388 my ($s,$d,$i) = (1,1,0); # swish, display default
389 $s = 0 if (lc($x->{type}) eq "display");
390 $d = 0 if (lc($x->{type}) eq "swish");
391 # no support for swish exact in config.
393 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
396 $display_data .= $delimiter.$val if ($d);
397 $swish_data .= $val if ($s);
398 $index->insert($field, $val, $path) if ($i);
401 if ($iterate_by_page) {
402 # FIXME data from config tag will appear just
406 $cache->{display_data}->{$field}->[$page] = $display_data;
410 $cache->{swish_data}->{$field}->[$page] = $swish_data;
413 if ($swish_exact_data) {
414 $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
415 $swish_exact_data = "";
420 # save data page-by-page
421 foreach my $field (@page_fields) {
422 my $nr_pages = $page_max{$field} || next;
423 #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
424 #print STDERR Dumper($cache->{display_data});
425 for (my $page=0; $page <= $nr_pages; $page++) {
427 my $display_data = $cache->{display_data}->{$field}->[$page];
428 if ($display_data) { # default
429 if ($field eq "headline") {
430 $xml .= xmlify("headline", $display_data);
433 # fallback to empty field name if needed
434 $html .= get_field_name($config,$field,$field_usage{$field}) || '';
435 $html .= "#-#".$display_data."###\n";
439 my $swish_data = $cache->{swish_data}->{$field}->[$page];
441 # remove extra spaces
442 $swish_data =~ s/ +/ /g;
443 $swish_data =~ s/ +$//g;
445 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
448 my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
449 if ($swish_exact_data) {
450 $swish_exact_data =~ s/ +/ /g;
451 $swish_exact_data =~ s/ +$//g;
453 # add delimiters before and after word.
454 # That is required to produce exact match
455 $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
458 foreach my $d (@{$cache->{index_data}->{$field}->[$page]}) {
459 $index->insert($field, $d, $path);
460 #print STDERR "index $field: $d [$path]\n";
466 if (! $iterate_by_page) {
468 if ($field eq "headline") {
469 $xml .= xmlify("headline", $display_data);
472 # fallback to empty field name if needed
473 $html .= get_field_name($config,$field,$field_usage{$field}) || '';
474 $html .= "#-#".$display_data."###\n";
478 # remove extra spaces
479 $swish_data =~ s/ +/ /g;
480 $swish_data =~ s/ +$//g;
482 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
485 if ($swish_exact_data) {
486 $swish_exact_data =~ s/ +/ /g;
487 $swish_exact_data =~ s/ +$//g;
489 # add delimiters before and after word.
490 # That is required to produce exact match
491 $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
496 # dump formatted output in <html>
498 #$xml .= xmlify("html",$html);
499 $xml .= "<html><![CDATA[ $html ]]></html>";
503 $xml .= $add_xml if ($add_xml);
504 return "<xml>\n$xml</xml>\n";
510 ##########################################################################
512 # read configuration for this script
513 my $cfg = new Config::IniFiles( -file => $config_file );
515 # read global.conf configuration
516 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
519 $index = new index_DBI(
520 $cfg_global->val('global', 'dbi_dbd'),
521 $cfg_global->val('global', 'dbi_dsn'),
522 $cfg_global->val('global', 'dbi_user'),
523 $cfg_global->val('global', 'dbi_passwd') || '',
526 my $show_progress = $cfg_global->val('global', 'show_progress');
528 my $unac_filter = $cfg_global->val('global', 'unac_filter');
530 require $unac_filter;
533 foreach my $database ($cfg->Sections) {
535 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
536 my $add_xml = $cfg -> val($database, 'xml'); # optional
538 # create new lookup file
539 my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
541 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
542 tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
543 print STDERR "creating lookup file '$lookup_file'\n";
546 # open existing lookup file
547 $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
549 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
550 tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
551 print STDERR "opening lookup file '$lookup_file'\n";
554 print STDERR "reading ./import_xml/$type.xml\n";
556 # extract just type basic
557 my $type_base = $type;
558 $type_base =~ s/_.+$//g;
560 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
562 # output current progress indicator
565 return if (! $show_progress);
567 my $total = shift || 1;
568 my $p = int($current * 100 / $total);
570 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
577 return if (! $show_progress);
578 my $current = shift @_;
580 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
582 $last_p += $fake_dir;
583 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
584 if ($last_p % 10 == 0) {
585 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
590 print STDERR "using: $type...\n";
592 # erase cache for tags by order in this database
593 delete $cache->{tags_by_order};
595 if ($type_base eq "isis") {
597 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
599 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
600 my $db = OpenIsis::open( $isis_db );
602 # check if .txt database for OpenIsis is zero length,
603 # if so, erase it and re-open database
605 my $isis_db = shift || die "need isis database name";
608 if (-e $isis_db.".TXT") {
609 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
610 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
613 if (-e $isis_db.".PTR") {
614 print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
615 unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
618 return OpenIsis::open( $isis_db ) if ($reopen);
623 $db = check_txt_db($isis_db);
625 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
632 print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
635 } elsif ($db == -5) {
636 print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
639 print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
643 my $max_rowid = OpenIsis::maxRowid( $db );
645 # if 0 records, try to rease isis .txt database
646 if ($max_rowid == 0) {
647 # force removal of database
648 $db = check_txt_db($isis_db);
649 $max_rowid = OpenIsis::maxRowid( $db );
652 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
654 my $path = $database;
656 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
657 my $row = OpenIsis::read( $db, $row_id );
658 if ($row && $row->{mfn}) {
660 progress($row->{mfn}, $max_rowid);
662 my $swishpath = $path."#".int($row->{mfn});
664 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
665 $xml = $cp2utf->convert($xml);
666 use bytes; # as opposed to chars
667 print "Path-Name: $swishpath\n";
668 print "Content-Length: ".(length($xml)+1)."\n";
669 print "Document-Type: XML\n\n$xml\n";
673 # for this to work with current version of OpenIsis (0.9.0)
674 # you might need my patch from
675 # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
676 OpenIsis::close($db);
679 } elsif ($type_base eq "excel") {
680 use Spreadsheet::ParseExcel;
681 use Spreadsheet::ParseExcel::Utility qw(int2col);
683 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
684 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
686 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
687 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
689 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
692 foreach my $oWks (@{$oBook->{Worksheet}}) {
693 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
694 last if ($oWks->{Name} eq $sheet);
698 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
700 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
701 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
702 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
704 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
706 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
707 my $cell = $oWorksheet->{Cells}[$iR][$iC];
709 $row->{int2col($iC)} = $cell->Value;
713 progress($iR, $end_row);
715 # print "row[$iR/$end_row] ";
716 # foreach (keys %{$row}) {
717 # print "$_: ",$row->{$_},"\t";
721 my $swishpath = $database."#".$iR;
725 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
726 $xml = $cp2utf->convert($xml);
727 use bytes; # as opposed to chars
728 print "Path-Name: $swishpath\n";
729 print "Content-Length: ".(length($xml)+1)."\n";
730 print "Document-Type: XML\n\n$xml\n";
733 } elsif ($type_base eq "marc") {
737 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
738 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
740 # optional argument is format
741 my $format = x($config->{format}) || 'usmarc';
743 print STDERR "Reading MARC file '$marc_file'\n";
746 my $nr = $marc->openmarc({
747 file=>$marc_file, format=>$format
748 }) || die "Can't open MARC file '$marc_file'";
750 my $i=0; # record nr.
754 while ($marc->nextmarc(1)) {
759 my $swishpath = $database."#".$i;
761 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
762 $xml = $cp2utf->convert($xml);
763 use bytes; # as opposed to chars
764 print "Path-Name: $swishpath\n";
765 print "Content-Length: ".(length($xml)+1)."\n";
766 print "Document-Type: XML\n\n$xml\n";
769 } elsif ($type_base eq "feed") {
771 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
772 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
774 print STDERR "Reading feed from program '$prog'\n";
776 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
778 my $i=1; # record nr.
787 my $swishpath = $database."#".$i++;
789 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
790 $xml = $cp2utf->convert($xml);
791 use bytes; # as opposed to chars
792 print "Path-Name: $swishpath\n";
793 print "Content-Length: ".(length($xml)+1)."\n";
794 print "Document-Type: XML\n\n$xml\n";
801 $line = $1 if (s/^(\d+):\s*//);
802 $data->{$line++} = $_;
808 untie %lhash if (%lhash);
812 # call this to commit index
817 ##########################################################################
821 all2xml.pl - read various file formats and dump XML for SWISH-E
825 This command will read ISIS data file using OpenIsis perl module, MARC
826 records using MARC module and optionally Micro$oft Excel files to
827 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
828 this script B<isn't general xml generator> from isis files (isis allready
829 has something like that). Output of this script is tailor-made for SWISH-E.
833 Documentation is really lacking. However, in true Open Source spirit, source
834 is best documentation. I even made considerable effort to comment parts
835 which are not intuitively clear, so...
839 Dobrica Pavlinusic <dpavlin@rot13.org>
843 GNU Public License (GPL) v2 or later
847 SWISH-E web site at http://www.swish-e.org