--- trunk/all2xml.pl 2002/11/24 20:52:11 1
+++ trunk/all2xml.pl 2003/06/01 13:46:42 50
@@ -5,62 +5,306 @@
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;
-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 $config;
+
+use index_DBI; # there is no other, right now ;-)
+my $index;
my %opts;
-getopts('d:m:q', \%opts);
+# usage:
+# -d directory name
+# -m multiple directories
+# -q quiet
+# -s run swish
+
+getopts('d:m:qs', \%opts);
+
+my $path; # this is name of database
+
+Text::Iconv->raise_error(1); # Conversion errors raise exceptions
+
+# 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';
+
+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]);
+}
+
+# decode isis import codepage
+my $isis2cp;
-my $db_dir = $opts{d};
+# outgoing xml must be in UTF-8
+my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
-die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
+sub isis2xml {
-#--------------------------------------------------------------------
+ use xmlify;
-my $last_tell=0;
+ my $row = shift @_;
+ my $add_xml = shift @_;
-my @isis_dirs = ( '.' ); # use dirname as database name
+ my $xml;
-if ($opts{m}) {
- @isis_dirs = split(/,/,$opts{m});
-}
+ use parse_format;
+
+ my $html = ""; # html formatted display output
-my @isis_dbs;
+ my %field_usage; # counter for usage of each field
-foreach (@isis_dirs) {
- if (-e "$common::isis_data/$db_dir/$_/LIBRI") {
- push @isis_dbs,"$common::isis_data/$db_dir/$_/LIBRI/LIBRI";
+ # 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} ;
}
- if (-e "$common::isis_data/$db_dir/$_/PERI") {
- push @isis_dbs,"$common::isis_data/$db_dir/$_/PERI/PERI";
+
+ 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);
+
+ foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+
+ my $format = x($x->{content});
+ my $delimiter = x($x->{delimiter}) || ' ';
+
+ my $isis_i = 0; # isis repeatable offset
+
+ my ($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,$display) = (1,1);
+
+ while ($swish || $display) {
+ ($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp);
+
+ # 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, $path);
+ }
+ }
+ }
+ }
+
+
+ if ($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));
+ }
+
+
}
- 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) {
+ $xml .= $add_xml if ($add_xml);
+ return "\n$xml\n";
+ } else {
+ return;
}
}
-foreach my $isis_db (@isis_dbs) {
+##########################################################################
+
+my $cfg = new Config::IniFiles( -file => $config_file );
+
+# open index
+$index = new index_DBI(
+ $cfg->val('global', 'dbi_dbd'),
+ $cfg->val('global', 'dbi_dsn'),
+ $cfg->val('global', 'dbi_user'),
+ $cfg->val('global', 'dbi_passwd') || '',
+ );
- my $db = OpenIsis::open( "$isis_db" ) || warn "can't open '$isis_db'";
+# delete [global] section to leave just databases sections
+$cfg->DeleteSection('global');
+
+foreach my $database ($cfg->Sections) {
+
+ my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!";
+ my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined";
+ my $add_xml = $cfg -> val($database, 'xml'); # optional
+
+ $config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1);
+
+ $isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
+
+ 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}) {
+ # 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;
+ }
+
+ my $swishpath = $path."#".int($row->{mfn});
+
+ if (my $xml = isis2xml($row,$add_xml)) {
+ $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$xml\n";
+ }
}
}
+ print STDERR "\n";
}
+
+# call this to commit index
+$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