--- trunk/all2xml.pl 2002/11/24 20:52:11 1
+++ trunk/all2xml.pl 2003/02/22 23:49:22 20
@@ -5,62 +5,271 @@
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;
-my $config=XMLin();
+$|=1;
-print Dumper($config);
+my $config_file = $0;
+$config_file =~ s/\.pl$/.conf/;
+die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file);
-my %opts;
-
-getopts('d:m:q', \%opts);
+my $config;
-my $db_dir = $opts{d};
+use index_DBI; # there is no other, right now ;-)
-die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+my $index = new index_DBI(); # open index
-#--------------------------------------------------------------------
+my %opts;
-my $last_tell=0;
+# usage:
+# -d directory name
+# -m multiple directories
+# -q quiet
+# -s run swish
+
+getopts('d:m:qs', \%opts);
+
+my $db_dir;
+
+#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+
+#print Dumper($config->{indexer});
+#print "-" x 70,"\n";
+
+Text::Iconv->raise_error(1); # Conversion errors raise exceptions
+
+#my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8');
+#my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage});
+my $isis_codepage;
+my $index_codepage;
+my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1');
+my $xml_codepage;
+
+sub isis2xml {
+
+ use xmlify;
+
+ my $row = shift @_;
+ my $add_xml = shift @_;
+
+ my $xml;
+
+ use parse_format;
+
+ my $html = ""; # html formatted display output
+
+ my %field_usage; # counter for usage of each field
+
+ foreach my $field (keys %{$config->{indexer}}) {
+
+ $field_usage{$field}++;
+
+ my $swish_data = "";
+ my $display_data = "";
+
+ foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+
+ my $format = $x->{content};
+ my ($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");
+#print STDERR "## s: $s d: $d i: $i ## $format ##\n";
+
+ # FIX: this is ugly, UGLY, cludge string is returned
+ # in UTF8 encoding , but as if source charset
+ # is ISO8859-1 and not some other. This break other
+ # character encodings, so we convert it first
+ # back to ISO8859-1
+ $format = $cludge_codepage->convert($format);
+
+ my ($swish,$display) = parse_format($format,$row);
+#print STDERR "s: $swish\nd: $display\n" if ($swish);
+
+#print STDERR "swish: $swish<-- display: $display<--\n";
+
+ # filter="name" ; filter this field through
+ # filter/[name].pm
+ my $filter;
+ if ($x->{filter}) {
+ $filter = "filter/".$x->{filter}.".pm";
+ require $filter;
+ }
+ # type="swish" ; field for swish
+ if ($s && $swish) {
+ if ($filter) {
+ $swish_data .= join(" ",&filter($swish));
+ } else {
+ $swish_data .= $swish if ($s && $swish);
+ }
+ }
+
+ # type="display" ; field for display
+ if ($d && $display) {
+ if ($filter) {
+ $display_data .= join(" ",&filter($display));
+ } else {
+ $display_data .= $display if ($s && $display);
+ }
+ }
+
+ # type="index" ; insert into index
+ if ($i && $display) {
+ my $index_data = $index_codepage->convert($display) || $display;
+ if ($filter) {
+ foreach my $d (&filter($index_data)) {
+ $index->insert($field, $d, $db_dir);
+ }
+ } else {
+ $index->insert($field, $index_data, $db_dir);
+ }
+ }
+ }
-my @isis_dirs = ( '.' ); # use dirname as database name
-if ($opts{m}) {
- @isis_dirs = split(/,/,$opts{m});
-}
+#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data);
+ if ($display_data) {
+ $display_data = $isis_codepage->convert($display_data) || die "Can't convert '$display_data' !";
+ # FIX: this is removed and replaced by html tag.
+ #$xml .= xmlify($field."_display", $display_data);
+
+ if ($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 .= $xml_codepage->convert($field_name);
+ }
+ $html .= $display_data."###\n";
+ }
+ }
+ if ($swish_data) {
+ my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2');
+ # remove extra spaces
+ $swish_data =~ s/ +/ /g;
+ $swish_data =~ s/ +$//g;
+
+ $swish_data = $i->convert($swish_data);
+ $xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data));
+ #$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data;
+ #$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data));
+ }
-my @isis_dbs;
-foreach (@isis_dirs) {
- if (-e "$common::isis_data/$db_dir/$_/LIBRI") {
- push @isis_dbs,"$common::isis_data/$db_dir/$_/LIBRI/LIBRI";
}
- if (-e "$common::isis_data/$db_dir/$_/PERI") {
- push @isis_dbs,"$common::isis_data/$db_dir/$_/PERI/PERI";
- }
- if (-e "$common::isis_data/$db_dir/$_/AMS") {
- push @isis_dbs,"$common::isis_data/$db_dir/$_/AMS/AMS";
+
+ # dump formatted output in
+ if ($html) {
+ $xml .= xmlify("html",$html);
}
- if (-e "$common::isis_data/$db_dir/$_/ARTI") {
-# push @isis_dbs,"$common::isis_data/$db_dir/$_/ARTI/ARTI";
+
+ if ($xml) {
+#print STDERR "x: $xml\n";
+ $xml .= $add_xml if ($add_xml);
+ return "\n$xml\n";
+ } else {
+ return;
}
}
-foreach my $isis_db (@isis_dbs) {
+##########################################################################
- my $db = OpenIsis::open( "$isis_db" ) || warn "can't open '$isis_db'";
+my $cfg = new Config::IniFiles( -file => $config_file );
+
+foreach my $database ($cfg->Sections) {
+
+ my $isis_db = $cfg -> val($database, 'isis_db');
+ my $type = $cfg -> val($database, 'type');
+ my $add_xml = $cfg -> val($database, 'xml');
+
+ # read configuration for this type
+ $config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1);
+ $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8');
+ $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage});
+ $xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),'UTF8');
+
+ my $db = OpenIsis::open( $isis_db );
+ if (0) {
+# # FIX
+# if (! $db ) {
+ print STDERR "WARNING: can't open '$isis_db'\n";
+ next ;
+ }
my $max_rowid = OpenIsis::maxRowid( $db );
- my $last_pcnt = 0;
+ print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
+
+ my $path = $database; # was $isis_db
+
+ my $last_p = 0;
+# { my $row_id = 4514;
+# FIX
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
my $row = OpenIsis::read( $db, $row_id );
-
- # output current process indicator
- my $pcnt = int($row->{mfn} * 100 / $max_rowid);
- if ($pcnt != $last_pcnt) {
- printf STDERR ("%5d / %5d -- %-2d %%\n",$row->{mfn},$max_rowid,$pcnt) if (! $opts{q});
- $last_pcnt = $pcnt;
+ if ($row && $row->{mfn}) {
+#print STDERR "mfn: ",$row->{mfn},"\n";
+ # output current process indicator
+ my $p = int($row->{mfn} * 100 / $max_rowid);
+ if ($p != $last_p) {
+ printf STDERR ("%5d / %5d [%-51s] %-2d %% \r",$row->{mfn},$max_rowid,"=" x ($p/2).">", $p ) if (! $opts{q});
+ $last_p = $p;
+ }
+
+ if (my $xml = isis2xml($row,$add_xml)) {
+#print STDERR "--ret-->$xml\n";
+ print "Path-Name: $path#".int($row->{mfn})."\n";
+ print "Content-Length: ".(length($xml)+1)."\n";
+ print "Document-Type: XML\n\n$xml\n";
+ }
}
}
+ print STDERR "\n";
}
+
+# 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
+
+SWISH-E web site at http://www.swish-e.org
+
+=cut