--- trunk/all2xml.pl 2003/07/04 20:11:48 62
+++ trunk2/all2all.pl 2004/06/13 20:39:44 349
@@ -1,451 +1,126 @@
#!/usr/bin/perl -w
-use strict;
-use OpenIsis;
-use Getopt::Std;
-use Data::Dumper;
-use XML::Simple;
-use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
-use Text::Iconv;
-use Config::IniFiles;
-use Encode;
-
-$|=1;
-
-my $config_file = $0;
-$config_file =~ s/\.pl$/.conf/;
-die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
-
-my $config;
-
-#use index_DBI; # default DBI module for index
-use index_DBI_cache; # faster DBI module using memory cache
-my $index;
-
-my %opts;
+=head1 NAME
-# usage:
-# -d directory name
-# -m multiple directories
-# -q quiet
-# -s run swish
+webpac.pl - basic script for all WebPac needs
-getopts('d:m:qs', \%opts);
+=cut
-my $path; # this is name of database
+use strict;
+use OpenIsis;
+use Text::Iconv;
+use Data::Dumper;
-Text::Iconv->raise_error(0); # Conversion errors don't raise exceptions
+my $cp = Text::Iconv->new("852", "iso-8859-2");
-# this is encoding of all files on disk, including import_xml/*.xml file and
-# filter/*.pm files! It will be used to store strings in perl internally!
-my $codepage = 'ISO-8859-2';
+$|=1;
-my $utf2cp = Text::Iconv->new('UTF-8',$codepage);
-# this function will convert data from XML files to local encoding
-sub x {
- return $utf2cp->convert($_[0]);
-}
+my $db = OpenIsis::open( shift @ARGV || '/data/hidra/THS-500/THS' );
+my $maxmfn = OpenIsis::maxRowid( $db ) || 1;
-# decode isis/excel or other import codepage
-my $import2cp;
+print "rows: $maxmfn\n\n";
-# outgoing xml must be in UTF-8
-my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
+my $data;
-# mapping between data type and tag which specify
-# format in XML file
-my %type2tag = (
- 'isis' => 'isis',
- 'excel' => 'column',
- 'marc' => 'marc',
+my @lookup = (
+ { 'key' => 'd:v900', 'val' => 'v250^a' },
+ { 'eval' => '"v901^a" eq "Područje"', 'key' => 'pa:v561^4:v562^4:v461^1', 'val' => 'v900' },
+ { 'eval '=> '"v901^a" eq "Mikrotezaurus"', 'key' => 'ma:v561^4:v562^4:v461^1', 'val' => 'v900' },
+ { 'eval' => '"v901^a" eq "Deskriptor"', 'key' => 'da:v561^4:v562^4:v461^1', 'val' => 'v900' },
);
-sub data2xml {
-
- use xmlify;
-
- my $type = shift @_;
- my $row = shift @_;
- my $add_xml = shift @_;
- # needed to read values from configuration file
- my $cfg = shift @_;
- my $database = shift @_;
-
- my $xml;
-
- use parse_format;
-
- my $html = ""; # html formatted display output
-
- my %field_usage; # counter for usage of each field
-
- # sort subrouting using order="" attribute
- sub by_order {
- return 0 if (! $config->{indexer}->{$a}->{order});
- return 0 if (! $config->{indexer}->{$b}->{order});
-
- return $config->{indexer}->{$a}->{order} <=>
- $config->{indexer}->{$b}->{order} ;
- }
-
- foreach my $field (sort by_order keys %{$config->{indexer}}) {
-
- $field=x($field);
-
- $field_usage{$field}++;
-
- my $swish_data = "";
- my $display_data = "";
- my $line_delimiter;
-
- my ($swish,$display);
-
- my $tag = $type2tag{$type} || die "can't find which tag to use for type $type";
- foreach my $x (@{$config->{indexer}->{$field}->{$tag}}) {
-
- my $format = x($x->{content});
- my $delimiter = x($x->{delimiter}) || ' ';
-
- my $repeat_off = 0; # repeatable offset
-
- my ($s,$d,$i) = (1,1,0); # swish, display default
- $s = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "swish");
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
-
- # what will separate last line from this one?
- if ($display_data && $x->{append} && $x->{append} eq "1") {
- $line_delimiter = ' ';
- } elsif ($display_data) {
- $line_delimiter = '
';
- }
-
- # init vars so that we go into while...
- ($swish,$display) = (1,1);
-
- if ($swish || $display) {
- ($swish,$display) = parse_format($type, $format,$row,$repeat_off++,$import2cp);
- # filter="name" ; filter this field through
- # filter/[name].pm
- my $filter = $x->{filter};
- if ($filter) {
- require "filter/".$filter.".pm";
- }
- # type="swish" ; field for swish
- if ($s && $swish) {
- if ($filter) {
- no strict 'refs';
- $swish_data .= join(" ",&$filter($swish));
- } else {
- $swish_data .= $swish;
- }
- }
-
- # type="display" ; field for display
- if ($d && $display) {
- if ($line_delimiter && $display_data) {
- $display_data .= $line_delimiter;
- undef $line_delimiter;
- }
- if ($filter) {
- no strict 'refs';
- $display_data .= join($delimiter,&$filter($display));
- } else {
- if ($display_data) {
- $display_data .= $delimiter.$display;
- } else {
- $display_data .= $display;
- }
- }
- }
-
- # type="index" ; insert into index
- if ($i && $display) {
- my $index_data = $display;
- if ($filter) {
- no strict 'refs';
- foreach my $d (&$filter($index_data)) {
- $index->insert($field, $d, $path);
- }
- } else {
- $index->insert($field, $index_data, $path);
- }
- }
- }
- }
-
- # now try to parse variables from configuration file
- foreach my $x (@{$config->{indexer}->{$field}->{'config'}}) {
+sub fill_in {
+ my $tmp = shift || return;
+ my $data = shift || return;
+ my $i = shift || 0;
- my $delimiter = x($x->{delimiter}) || ' ';
- my $val = $cfg->val($database, x($x->{content}));
-
- my ($s,$d,$i) = (1,1,0); # swish, display default
- $s = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "swish");
- ($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index");
-
- if ($val) {
- $display_data .= $delimiter.$val if ($d);
- $swish_data .= $val if ($s);
- $index->insert($field, $val, $path) if ($i);
- }
+ my $found = 0;
+ sub get_sf {
+ my ($found,$data,$f,$sf,$i) = @_;
+ if ($$data->{$f} && $$data->{$f}->[$i]->{$sf}) {
+ $$found++;
+ return $$data->{$f}->[$i]->{$sf};
+ } else {
+ return '';
}
+ }
-
- if ($display_data) {
-
- if ($field eq "headline") {
- $xml .= xmlify("headline", $display_data);
- } else {
-
- # find field name (signular, plural)
- my $field_name = "";
- if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) {
- $field_name = $config->{indexer}->{$field}->{name_singular}."#-#";
- } elsif ($config->{indexer}->{$field}->{name_plural}) {
- $field_name = $config->{indexer}->{$field}->{name_plural}."#-#";
- } elsif ($config->{indexer}->{$field}->{name}) {
- $field_name = $config->{indexer}->{$field}->{name}."#-#";
- } else {
- print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
- }
- if ($field_name) {
- $html .= x($field_name);
- }
- $html .= $display_data."###\n";
- }
- }
- if ($swish_data) {
- # remove extra spaces
- $swish_data =~ s/ +/ /g;
- $swish_data =~ s/ +$//g;
-
- $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
+ sub get_nosf {
+ my ($found,$data,$f,$i) = @_;
+ if ($$data->{$f} && $$data->{$f}->[$i]) {
+ $$found++;
+ return $$data->{$f}->[$i];
+ } else {
+ return '';
}
-
-
}
- # dump formatted output in
- if ($html) {
- $xml .= xmlify("html",$html);
- }
-
- if ($xml) {
- $xml .= $add_xml if ($add_xml);
- return "\n$xml\n";
+ $tmp =~ s/v(\d+)\^(\w)/get_sf(\$found,\$data,$1,$2,$i)/ges;
+ $tmp =~ s/v(\d+)/get_nosf(\$found,\$data,$1,$i)/ges;
+
+ if ($found) {
+ return $tmp;
} else {
return;
}
}
-##########################################################################
-
-# read configuration for this script
-my $cfg = new Config::IniFiles( -file => $config_file );
-
-# read global.conf configuration
-my $cfg_global = new Config::IniFiles( -file => 'global.conf' );
-
-# open index
-$index = new index_DBI(
- $cfg_global->val('global', 'dbi_dbd'),
- $cfg_global->val('global', 'dbi_dsn'),
- $cfg_global->val('global', 'dbi_user'),
- $cfg_global->val('global', 'dbi_passwd') || '',
- );
-
-foreach my $database ($cfg->Sections) {
-
- my $type = lc($cfg -> val($database, 'type')) || die "$database doesn't have 'type' defined";
- my $add_xml = $cfg -> val($database, 'xml'); # optional
-
-print STDERR "reading ./import_xml/$type.xml\n";
-
- # extract just type basic
- my $type_base = $type;
- $type_base =~ s/_.+$//g;
-
- $config=XMLin("./import_xml/$type.xml", forcearray => [ $type2tag{$type_base}, 'config' ], forcecontent => 1);
-
- # output current progress indicator
- my $last_p = 0;
- sub progress {
- #return if (! $opts{q}); # FIXME
- my $current = shift;
- my $total = shift || 1;
- my $p = int($current * 100 / $total);
- if ($p != $last_p) {
- printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$current,$total,"=" x ($p/2).">", $p );
- $last_p = $p;
- }
- }
-
- # now read database
-print STDERR "using: $type...\n";
-
- if ($type_base eq "isis") {
-
- my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
-
- $import2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
- my $db = OpenIsis::open( $isis_db );
-
- my $max_rowid = OpenIsis::maxRowid( $db );
-
- print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
+print "reading database...\n";
- my $path = $database;
+for (my $mfn = 1; $mfn <= $maxmfn; $mfn++) {
- for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
- my $row = OpenIsis::read( $db, $row_id );
- if ($row && $row->{mfn}) {
-
- progress($row->{mfn}, $max_rowid);
-
- my $swishpath = $path."#".int($row->{mfn});
-
- if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
- $xml = $cp2utf->convert($xml);
- use bytes; # as opposed to chars
- print "Path-Name: $swishpath\n";
- print "Content-Length: ".(length($xml)+1)."\n";
- print "Document-Type: XML\n\n$xml\n";
+ my $row = OpenIsis::read( $db, $mfn );
+ foreach my $k (keys %{$row}) {
+ if ($k ne "mfn") {
+ foreach my $l (@{$row->{$k}}) {
+ $l = $cp->convert($l);
+ # has subfields?
+ my $val;
+ if ($l =~ m/\^/) {
+ foreach my $t (split(/\^/,$l)) {
+ next if (! $t);
+ $val->{substr($t,0,1)} = substr($t,1);
+ }
+ } else {
+ $val = $l;
}
- }
- }
- print STDERR "\n";
- } elsif ($type_base eq "excel") {
- use Spreadsheet::ParseExcel;
- use Spreadsheet::ParseExcel::Utility qw(int2col);
-
- $import2cp = Text::Iconv->new($config->{excel_codepage},$codepage);
- my $excel_file = $cfg -> val($database, 'excel_file') || die "$database doesn't have 'excel_file' defined!";
-
- my $sheet = x($config->{sheet}) || die "no sheet in $type.xml";
- my $start_row = x($config->{start_row}) - 1 || die "no start_row in $type.xml";
-
- my $oBook = Spreadsheet::ParseExcel::Workbook->Parse($excel_file) || die "can't open Excel file '$excel_file'";
-
- my $sheet_nr = 0;
- foreach my $oWks (@{$oBook->{Worksheet}}) {
- #print STDERR "-- SHEET $sheet_nr:", $oWks->{Name}, "\n";
- last if ($oWks->{Name} eq $sheet);
- $sheet_nr++;
+ push @{$data->{$mfn}->{$k}}, $val;
+ }
}
- my $oWorksheet = $oBook->{Worksheet}[$sheet_nr];
-
- print STDERR "using sheet: ",$oWorksheet->{Name},"\n";
- defined ($oWorksheet) || die "can't find sheet '$sheet' in $excel_file";
- my $end_row = x($config->{end_row}) || $oWorksheet->{MaxRow};
-
- for(my $iR = $start_row ; defined $end_row && $iR <= $end_row ; $iR++) {
- my $row;
- for(my $iC = $oWorksheet->{MinCol} ; defined $oWorksheet->{MaxCol} && $iC <= $oWorksheet->{MaxCol} ; $iC++) {
- my $cell = $oWorksheet->{Cells}[$iR][$iC];
- if ($cell) {
- $row->{int2col($iC)} = $cell->Value;
- }
- }
+ }
- progress($iR, $end_row);
-# print "row[$iR/$end_row] ";
-# foreach (keys %{$row}) {
-# print "$_: ",$row->{$_},"\t";
-# }
-# print "\n";
-
- my $swishpath = $database."#".$iR;
-
- next if (! $row);
-
- if (my $xml = data2xml($type_base,$row,$add_xml,$cfg,$database)) {
- $xml = $cp2utf->convert($xml);
- use bytes; # as opposed to chars
- print "Path-Name: $swishpath\n";
- print "Content-Length: ".(length($xml)+1)."\n";
- print "Document-Type: XML\n\n$xml\n";
+ foreach my $i (@lookup) {
+ my $d = $data->{$mfn};
+ if ($i->{'eval'}) {
+ my $eval = fill_in($i->{'eval'},$d);
+ my $key = fill_in($i->{'key'}, $d);
+ my $val = fill_in($i->{'val'}, $d);
+ if ($key && $val && eval $eval) {
+ push @{$data->{lookup}->{$key}}, $val;
}
- }
- } elsif ($type_base eq "marc") {
- ## XXX
- use MARC;
-
- $import2cp = Text::Iconv->new($config->{marc_codepage},$codepage);
- my $marc_file = $cfg -> val($database, 'marc_file') || die "$database doesn't have 'marc_file' defined!";
-
- # optional argument is format
- my $format = x($config->{format}) || 'usmarc';
-
- my %id_stored; # to aviod duplicates
-
- print STDERR "Reading MARC file '$marc_file'\n";
-
- my $marc = new MARC;
- my $nr = $marc->openmarc({
- file=>$marc_file, format=>$format
- }) || die "Can't open MARC file '$marc_file'";
-
- my $i=0; # record nr.
- my $inc=1;
- my $max_i=1000;
-
- my $rec;
-
- while ($marc->nextmarc(1)) {
-
- # XXX
- progress($i, $max_i);
- $i += $inc;
- $inc = -$inc if ($i > $max_i || $i < 0);
-
- my $swishpath = $database."#".$i;
-
- if (my $xml = data2xml($type_base,$marc,$add_xml,$cfg,$database)) {
- $xml = $cp2utf->convert($xml);
- use bytes; # as opposed to chars
- print "Path-Name: $swishpath\n";
- print "Content-Length: ".(length($xml)+1)."\n";
- print "Document-Type: XML\n\n$xml\n";
+ } else {
+ my $key = fill_in($i->{'key'}, $d);
+ my $val = fill_in($i->{'val'}, $d);
+ if ($key && $val) {
+ push @{$data->{lookup}->{$key}}, $val;
}
}
}
-}
-
-# call this to commit index
-$index->close;
-1;
-__END__
-##########################################################################
-
-=head1 NAME
-
-isis2xml.pl - read isis file and dump XML
-
-=head1 DESCRIPTION
-
-This command will read ISIS data file using OpenIsis perl module and
-create XML file for usage with I
-indexer. Dispite it's name, this script B
-from isis files (isis allready has something like that). Output of this
-script is tailor-made for SWISH-E.
-
-=head1 AUTHOR
-
-Dobrica Pavlinusic
-
-=head1 COPYRIGHT
-
-GNU Public License (GPL) v2 or later
-
-=head1 SEE ALSO
+}
+print "\n";
-SWISH-E web site at http://www.swish-e.org
+for (my $mfn = 1; $mfn <= $maxmfn; $mfn++) {
+ my $d = $data->{$mfn};
+ print $d->{'250'}->[0]->{'a'},"\n";
+ my $t = fill_in('v553^1;;v553^a', $d);
+ print " Uži pojam: $t\n" if ($t);
+ $t = fill_in('[a:v251];;[d:[a:v251]]', $d);
+ print " Uži pojam: $t\n" if ($t);
+}
-=cut
+print Dumper($data);