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 my @page_fields; # names of fields
122 # subs used to produce output
124 sub get_field_name($$$) {
125 my ($config,$field,$field_usage) = @_;
127 # find field name (signular, plural)
129 if ($config->{indexer}->{$field}->{name_singular} && $field_usage == 1) {
130 $field_name = $config->{indexer}->{$field}->{name_singular};
131 } elsif ($config->{indexer}->{$field}->{name_plural}) {
132 $field_name = $config->{indexer}->{$field}->{name_plural};
133 } elsif ($config->{indexer}->{$field}->{name}) {
134 $field_name = $config->{indexer}->{$field}->{name};
136 print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
139 return x($field_name);
144 # begin real work: go field by field
145 foreach my $field (@sorted_tags) {
148 $field_usage{$field}++;
151 my $swish_exact_data = "";
152 my $display_data = "";
155 my ($swish,$display);
157 my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
159 # is this field page-by-page?
160 my $iterate_by_page = $config->{indexer}->{$field}->{iterate_by_page};
161 push @page_fields,$field if ($iterate_by_page);
163 # default line_delimiter if using
164 my $page_line_delimiter = $config->{indexer}->{$field}->{page_line_delimiter} || '<br/>';
166 foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
168 my $format = x($x->{content});
169 my $delimiter = x($x->{delimiter}) || ' ';
171 my $repeat_off = 0; # init repeatable offset
173 # swish, swish_exact, display, index, index_lookup
174 # swish and display defaults
175 my ($s,$se,$d,$i,$il) = (1,0,1,0,0);
176 $s = 0 if (lc($x->{type}) eq "display");
177 $d = 0 if (lc($x->{type}) eq "swish");
178 ($s,$se,$d,$i) = (0,0,0,1) if (lc($x->{type}) eq "index");
179 ($s,$se,$d,$i) = (0,1,0,0) if (lc($x->{type}) eq "swish_exact");
180 $il = 1 if (lc($x->{type}) =~ /^lookup/);
182 # what will separate last line from this one?
183 if ($display_data && $x->{append} && $x->{append} eq "1") {
184 $line_delimiter = ' ';
185 } elsif ($display_data) {
186 $line_delimiter = '<br/>';
189 # init vars so that we go into while...
190 ($swish,$display) = (1,1);
192 # placeholder for all repeatable entries for index
196 my $x = shift || die "mkformat needs tag reference";
197 my $data = shift || return;
198 my $format_name = x($x->{format_name}) || return $data;
199 my $fmt = x($config->{format}->{$format_name}->{content}) || die "<format name=\"$format_name\"> is not defined!";
200 my $format_delimiter = x($x->{format_delimiter});
202 if ($format_delimiter) {
203 @data = split(/$format_delimiter/,$data);
209 my $nr = scalar $fmt =~ s/%s/%s/g;
210 if (($#data+1) == $nr) {
211 return sprintf($fmt,@data);
213 print STDERR "mkformat: [$data] can't be split on [$format_delimiter] to $nr fields!\n";
217 print STDERR "usage of link '$format_name' without defined format (<link> tag)\n";
221 # while because of repeatable fields
222 while ($swish || $display) {
223 my $page = $repeat_off;
224 $page_max{$field} = $page if ($iterate_by_page && $page > ($page_max{$field} || 0));
225 ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
226 if ($repeat_off > 1000) {
227 print STDERR "loop (more than 1000 repeatable fields) deteced in $row, $format\n";
231 # is this field is lookup?
232 if ($display && $x->{lookup}) {
233 my $null = "<!-- null -->";
234 if ($use_lhash_cache) {
235 if (!defined($cache->{lhash}->{$display})) {
236 my $new_display = $lhash{$display};
237 if (defined($new_display)) {
238 #print STDERR "lookup cache store '$display' = '$new_display'\n";
239 $display = $new_display;
240 $cache->{lhash}->{$display} = $new_display;
242 print STDERR "WARNING: lookup for '$display' didn't find anything.\n";
244 $cache->{lhash}->{$display} = $null;
247 $display = $cache->{lhash}->{$display};
250 $display = $lhash{$display} || $null;
254 # filter="name" ; filter this field through
256 my $filter = $x->{filter};
257 if ($filter && !$cache->{filter_loaded}->{$filter}) {
258 require "filter/".$filter.".pm";
259 $cache->{filter_loaded}->{$filter}++;
261 # type="swish" ; field for swish
263 if ($filter && ($s || $se)) {
265 my $tmp = join(" ",&$filter($swish)) if ($s || $se);
266 $swish_data .= $tmp if ($s);
267 $swish_exact_data .= $tmp if ($se);
270 $swish_data .= $swish if ($s);
271 $swish_exact_data .= $swish if ($se);
275 # type="display" ; field for display
276 if ($d && $display) {
277 if ($line_delimiter && $display_data) {
278 $display_data .= $line_delimiter;
283 $display_data .= $delimiter.join($delimiter,mkformat($x,&$filter($display)));
285 $display_data = join($delimiter,mkformat($x,&$filter($display)));
289 $display_data .= $delimiter.mkformat($x,$display);
291 $display_data = mkformat($x,$display);
296 # type="index" ; insert into index
297 if ($i && $display) {
300 $display = &$filter($display);
302 if ($x->{append} && @index_data) {
303 $index_data[$#index_data].=$display;
305 push @index_data, $display;
309 # store fields in lookup
310 if ($il && $display) {
311 if (lc($x->{type}) eq "lookup_key") {
313 print STDERR "WARNING: try to redefine lookup_key (keys shouldn't be repeatable fields!)";
315 $lookup_key = $display;
317 } elsif (lc($x->{type}) eq "lookup_val") {
319 $lhash{$lookup_key} = $display;
321 print STDERR "WARNING: no lookup_key defined for '$display'?";
327 # store data for page-by-page repeatable fields
328 if ($iterate_by_page) {
329 sub iterate_fld($$$$$$) {
330 my ($cache,$what,$field,$page,$data,$append) = @_;
332 my $line_delimiter = $page_line_delimiter;
333 $line_delimiter = " " if ($append);
334 if (! $cache->{$what}->{$field}->[$page]) {
335 $cache->{$what}->{$field}->[$page] = $data;
337 $cache->{$what}->{$field}->[$page] .= $line_delimiter.$data;
342 print STDERR "line delimiter: ",Dumper($line_delimiter) if ($line_delimiter);
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 = "";
358 foreach my $d (@index_data) {
359 $index->insert($field, $d, $path);
363 # now try to parse variables from configuration file
364 foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
366 my $delimiter = x($x->{delimiter}) || ' ';
367 my $val = $cfg->val($database, x($x->{content}));
369 my ($s,$d,$i) = (1,1,0); # swish, display default
370 $s = 0 if (lc($x->{type}) eq "display");
371 $d = 0 if (lc($x->{type}) eq "swish");
372 # no support for swish exact in config.
374 ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
377 $display_data .= $delimiter.$val if ($d);
378 $swish_data .= $val if ($s);
379 $index->insert($field, $val, $path) if ($i);
382 if ($iterate_by_page) {
383 # FIXME data from config tag will appear just
387 $cache->{display_data}->{$field}->[$page] = $display_data;
391 $cache->{swish_data}->{$field}->[$page] = $swish_data;
394 if ($swish_exact_data) {
395 $cache->{swish_exact_data}->{$field}->[$page] = $swish_exact_data;
396 $swish_exact_data = "";
401 # save data page-by-page
402 foreach my $field (@page_fields) {
403 my $nr_pages = $page_max{$field} || next;
404 #print STDERR "field '$field' iterate over ",($nr_pages || 0)," pages...\n";
405 #print STDERR Dumper($cache->{display_data});
406 for (my $page=0; $page <= $nr_pages; $page++) {
408 my $display_data = $cache->{display_data}->{$field}->[$page];
409 if ($display_data) { # default
410 if ($field eq "headline") {
411 $xml .= xmlify("headline", $display_data);
414 # fallback to empty field name if needed
415 $html .= get_field_name($config,$field,$field_usage{$field}) || '';
416 $html .= "#-#".$display_data."###\n";
420 my $swish_data = $cache->{swish_data}->{$field}->[$page];
422 # remove extra spaces
423 $swish_data =~ s/ +/ /g;
424 $swish_data =~ s/ +$//g;
426 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
429 my $swish_exact_data = $cache->{swish_exact_data}->{$field}->[$page];
430 if ($swish_exact_data) {
431 $swish_exact_data =~ s/ +/ /g;
432 $swish_exact_data =~ s/ +$//g;
434 # add delimiters before and after word.
435 # That is required to produce exact match
436 $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
442 if (! $iterate_by_page) {
444 if ($field eq "headline") {
445 $xml .= xmlify("headline", $display_data);
448 # fallback to empty field name if needed
449 $html .= get_field_name($config,$field,$field_usage{$field}) || '';
450 $html .= "#-#".$display_data."###\n";
454 # remove extra spaces
455 $swish_data =~ s/ +/ /g;
456 $swish_data =~ s/ +$//g;
458 $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
461 if ($swish_exact_data) {
462 $swish_exact_data =~ s/ +/ /g;
463 $swish_exact_data =~ s/ +$//g;
465 # add delimiters before and after word.
466 # That is required to produce exact match
467 $xml .= xmlify($field."_swish_exact", unac_string($codepage,'xxbxx '.$swish_exact_data.' xxexx'));
472 # dump formatted output in <html>
474 #$xml .= xmlify("html",$html);
475 $xml .= "<html><![CDATA[ $html ]]></html>";
479 $xml .= $add_xml if ($add_xml);
480 return "<xml>\n$xml</xml>\n";
486 ##########################################################################
488 # read configuration for this script
489 my $cfg = new Config::IniFiles( -file => $config_file );
491 # read global.conf configuration
492 my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
495 $index = new index_DBI(
496 $cfg_global->val('global', 'dbi_dbd'),
497 $cfg_global->val('global', 'dbi_dsn'),
498 $cfg_global->val('global', 'dbi_user'),
499 $cfg_global->val('global', 'dbi_passwd') || '',
502 my $show_progress = $cfg_global->val('global', 'show_progress');
504 my $unac_filter = $cfg_global->val('global', 'unac_filter');
506 require $unac_filter;
509 foreach my $database ($cfg->Sections) {
511 my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
512 my $add_xml = $cfg -> val($database, 'xml'); # optional
514 # create new lookup file
515 my $lookup_file = $cfg -> val($database, 'lookup_newfile'); # optional
517 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_NEWDB, 0644;
518 tie %lhash, 'TDB_File', $lookup_file, TDB_CLEAR_IF_FIRST, O_RDWR, 0644;
519 print STDERR "creating lookup file '$lookup_file'\n";
522 # open existing lookup file
523 $lookup_file = $cfg -> val($database, 'lookup_open'); # optional
525 #tie %lhash, 'GDBM_File', $lookup_file, &GDBM_READER, 0644;
526 tie %lhash, 'TDB_File', $lookup_file, TDB_DEFAULT, O_RDWR, 0644;
527 print STDERR "opening lookup file '$lookup_file'\n";
530 print STDERR "reading ./import_xml/$type.xml\n";
532 # extract just type basic
533 my $type_base = $type;
534 $type_base =~ s/_.+$//g;
536 $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config', 'format' ], forcecontent => 1);
538 # output current progress indicator
541 return if (! $show_progress);
543 my $total = shift || 1;
544 my $p = int($current * 100 / $total);
546 printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
553 return if (! $show_progress);
554 my $current = shift @_;
556 my @ind = ('-','\\','|','/','-','\\','|','/', '-');
558 $last_p += $fake_dir;
559 $fake_dir = -$fake_dir if ($last_p > 1000 || $last_p < 0);
560 if ($last_p % 10 == 0) {
561 printf STDERR ("%5d / %5s [%-51s]\r",$current,"?"," " x ($last_p/20).$ind[($last_p/20) % $#ind]);
566 print STDERR "using: $type...\n";
568 # erase cache for tags by order in this database
569 delete $cache->{tags_by_order};
571 if ($type_base eq "isis") {
573 my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
575 $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
576 my $db = OpenIsis::open( $isis_db );
578 # check if .txt database for OpenIsis is zero length,
579 # if so, erase it and re-open database
581 my $isis_db = shift || die "need isis database name";
584 if (-e $isis_db.".TXT") {
585 print STDERR "WARNING: removing $isis_db.TXT OpenIsis database...\n";
586 unlink $isis_db.".TXT" || warn "FATAL: unlink error on '$isis_db.TXT': $!";
589 if (-e $isis_db.".PTR") {
590 print STDERR "WARNING: removing $isis_db.PTR OpenIsis database...\n";
591 unlink $isis_db.".PTR" || warn "FATAL: unlink error on '$isis_db.PTR': $!";
594 return OpenIsis::open( $isis_db ) if ($reopen);
599 $db = check_txt_db($isis_db);
601 print STDERR "FATAL: OpenIsis can't open zero size file $isis_db\n";
608 print STDERR "FATAL: OpenIsis can't find file $isis_db\n";
611 } elsif ($db == -5) {
612 print STDERR "FATAL: OpenIsis can't access file $isis_db\n";
615 print STDERR "FATAL: OpenIsis unknown error $db with file $isis_db\n";
619 my $max_rowid = OpenIsis::maxRowid( $db );
621 # if 0 records, try to rease isis .txt database
622 if ($max_rowid == 0) {
623 # force removal of database
624 $db = check_txt_db($isis_db);
625 $max_rowid = OpenIsis::maxRowid( $db );
628 print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
630 my $path = $database;
632 for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
633 my $row = OpenIsis::read( $db, $row_id );
634 if ($row && $row->{mfn}) {
636 progress($row->{mfn}, $max_rowid);
638 my $swishpath = $path."#".int($row->{mfn});
640 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
641 $xml = $cp2utf->convert($xml);
642 use bytes; # as opposed to chars
643 print "Path-Name: $swishpath\n";
644 print "Content-Length: ".(length($xml)+1)."\n";
645 print "Document-Type: XML\n\n$xml\n";
649 # for this to work with current version of OpenIsis (0.9.0)
650 # you might need my patch from
651 # http://www.rot13.org/~dpavlin/projects/openisis-0.9.0-perl_close.diff
652 OpenIsis::close($db);
655 } elsif ($type_base eq "excel") {
656 use Spreadsheet::ParseExcel;
657 use Spreadsheet::ParseExcel::Utility qw(int2col);
659 $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
660 my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
662 my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
663 my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
665 my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
668 foreach my $oWks (@{$oBook->{Worksheet}}) {
669 #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
670 last if ($oWks->{Name} eq $sheet);
674 my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
676 print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
677 defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
678 my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
680 for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
682 for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
683 my $cell = $oWorksheet->{Cells}[$iR][$iC];
685 $row->{int2col($iC)} = $cell->Value;
689 progress($iR, $end_row);
691 # print "row[$iR/$end_row] ";
692 # foreach (keys %{$row}) {
693 # print "$_: ",$row->{$_},"\t";
697 my $swishpath = $database."#".$iR;
701 if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
702 $xml = $cp2utf->convert($xml);
703 use bytes; # as opposed to chars
704 print "Path-Name: $swishpath\n";
705 print "Content-Length: ".(length($xml)+1)."\n";
706 print "Document-Type: XML\n\n$xml\n";
709 } elsif ($type_base eq "marc") {
713 $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
714 my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
716 # optional argument is format
717 my $format = x($config->{format}) || 'usmarc';
719 print STDERR "Reading MARC file '$marc_file'\n";
722 my $nr = $marc->openmarc({
723 file=>$marc_file, format=>$format
724 }) || die "Can't open MARC file '$marc_file'";
726 my $i=0; # record nr.
730 while ($marc->nextmarc(1)) {
735 my $swishpath = $database."#".$i;
737 if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
738 $xml = $cp2utf->convert($xml);
739 use bytes; # as opposed to chars
740 print "Path-Name: $swishpath\n";
741 print "Content-Length: ".(length($xml)+1)."\n";
742 print "Document-Type: XML\n\n$xml\n";
745 } elsif ($type_base eq "feed") {
747 $import2cp = Text::Iconv->new($config->{feed_codepage},$codepage);
748 my $prog = x($config->{prog}) || die "$database doesn't have 'prog' defined!";
750 print STDERR "Reading feed from program '$prog'\n";
752 open(FEED,"feeds/$prog |") || die "can't start $prog: $!";
754 my $i=1; # record nr.
763 my $swishpath = $database."#".$i++;
765 if (my $xml = data2xml($type_base,$data,$add_xml,$cfg,$database)) {
766 $xml = $cp2utf->convert($xml);
767 use bytes; # as opposed to chars
768 print "Path-Name: $swishpath\n";
769 print "Content-Length: ".(length($xml)+1)."\n";
770 print "Document-Type: XML\n\n$xml\n";
777 $line = $1 if (s/^(\d+):\s*//);
778 $data->{$line++} = $_;
784 untie %lhash if (%lhash);
788 # call this to commit index
793 ##########################################################################
797 all2xml.pl - read various file formats and dump XML for SWISH-E
801 This command will read ISIS data file using OpenIsis perl module, MARC
802 records using MARC module and optionally Micro$oft Excel files to
803 create one XML file for usage with I<SWISH-E> indexer. Dispite it's name,
804 this script B<isn't general xml generator> from isis files (isis allready
805 has something like that). Output of this script is tailor-made for SWISH-E.
809 Documentation is really lacking. However, in true Open Source spirit, source
810 is best documentation. I even made considerable effort to comment parts
811 which are not intuitively clear, so...
815 Dobrica Pavlinusic <dpavlin@rot13.org>
819 GNU Public License (GPL) v2 or later
823 SWISH-E web site at http://www.swish-e.org