--- trunk/all2xml.pl 2002/12/01 22:51:29 4
+++ trunk/all2xml.pl 2003/02/23 07:08:54 29
@@ -5,140 +5,232 @@
use Getopt::Std;
use Data::Dumper;
use XML::Simple;
-use Text::Unaccent;
-require Unicode::Map8;
+use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
+use Text::Iconv;
+use Config::IniFiles;
-my $config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1);
+$|=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; # there is no other, right now ;-)
+
+my $index = new index_DBI(); # open 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 $db_dir = $opts{d} || "ps"; # FIX
+my $db_dir;
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts);
#print Dumper($config->{indexer});
#print "-" x 70,"\n";
-# how to convert isis code page to UTF8?
-my $isis_map = Unicode::Map8->new($config->{isis_codepage}) || die;
+Text::Iconv->raise_error(1); # Conversion errors raise exceptions
+
+my $isis_codepage;
+my $index_codepage;
+my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1');
+my $xml_codepage;
+
+my $XML_CHARSET = 'UTF8';
+
sub isis2xml {
+ use xmlify;
+
my $row = shift @_;
+ my $add_xml = shift @_;
my $xml;
- sub isis_sf {
- my $row = shift @_;
- my $isis_id = shift @_;
- my $subfield = shift @_;
- if ($row->{$isis_id}->[0]) {
- my $sf = OpenIsis::subfields($row->{$isis_id}->[0]);
- if (! defined $subfield || length($subfield) == 0) {
- # subfield list undef, empty or no defined subfields for this record
- my $all_sf = $row->{$isis_id}->[0];
- $all_sf =~ s/\^./ /g; nuke definirions
- return $all_sf;
- } elsif ($sf->{$subfield}) {
- return $sf->{$subfield};
- }
- }
- }
+ 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 = "";
- my $index_data = "";
+ my $line_delimiter = "";
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
- my $display_tmp = "";
- my $index_tmp = "";
-
my $format = $x->{content};
- my $i = 1; # index only
- my $d = 1; # display only
- $i = 0 if (lc($x->{type}) eq "display");
- $d = 0 if (lc($x->{type}) eq "index");
-#print "## i: $i d: $d ## $format ##";
- # parse format
- my $prefix = "";
- if ($format =~ s/^([^\d]+)//) {
- $prefix = $1;
- }
- while ($format) {
- if ($format =~ s/^(\d\d\d)(\w?)//) {
- my $isis_tmp = isis_sf($row,$1,$2);
- if ($isis_tmp) {
-# $display_tmp .= $prefix . "/$1/$2/".$isis_tmp if ($d);
- $display_tmp .= $prefix . $isis_tmp if ($d);
- $index_tmp .= $isis_tmp." " if ($i);
-#print " $isis_tmp <--\n";
- }
- $prefix = "";
- } elsif ($format =~ s/^([^\d]+)//) {
- $prefix = $1;
- } else {
- print STDERR "WARNING: unparsed format '$format'\n";
- last;
- };
- }
- # add suffix
- $display_tmp .= $prefix if ($display_tmp);
+ my $delimiter = $x->{delimiter} || ' ';
-# $display_data .= $display_tmp if ($display_tmp ne "");
-# $index_data .= $index_tmp if ($index_tmp ne "");
- $display_data .= $display_tmp;
- $index_data .= $index_tmp;
+ # 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 = $xml_codepage->convert($format);
+ $delimiter = $xml_codepage->convert($delimiter) if ($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");
+#print STDERR "## s: $s d: $d i: $i ## $format ##\n";
+
+ $display_data .= $line_delimiter if ($display_data && $display_data !~ /$line_delimiter$/);
+
+ my ($swish,$display) = (1,1);
+
+ while ($swish || $display) {
+ ($swish,$display) = parse_format($format,$row,$isis_i++);
+#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 = $x->{filter};
+ if ($filter) {
+ require "filter/".$filter.".pm";
+ }
+ # type="swish" ; field for swish
+ if ($s && $swish) {
+ if ($filter) {
+print STDERR "using filter '$filter'\n";
+ no strict 'refs';
+ $swish_data .= join(" ",&$filter($swish));
+ } else {
+ $swish_data .= $swish;
+ }
+ }
+ # type="display" ; field for display
+ if ($d && $display) {
+ if ($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 = $index_codepage->convert($display) || $display;
+ if ($filter) {
+ no strict 'refs';
+ foreach my $d (&$filter($index_data)) {
+ $index->insert($field, $d, $db_dir);
+ }
+ } else {
+ $index->insert($field, $index_data, $db_dir);
+ }
+ }
+ }
+ if ($x->{append}) {
+ $line_delimiter = ' ';
+ } else {
+ $line_delimiter = '
';
+ }
}
-#print "--display:$display_data\n--index:$index_data\n";
- $xml->{$field}->{display} .= $isis_map->tou($display_data)->utf8 if ($display_data);
- $xml->{$field}->{index} .= unac_string($config->{isis_codepage},$index_data) if ($index_data);
-
- }
- if ($xml) {
- return XMLout($xml, rootname => 'xml', noattr => 1 );
- } else {
- return;
- }
-}
-##########################################################################
-my $last_tell=0;
+#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data);
+ if ($display_data) {
-my @isis_dirs = ( '.' ); # use dirname as database name
+ # remove last
+ $display_data =~ s/$line_delimiter$//;
-if ($opts{m}) {
- @isis_dirs = split(/,/,$opts{m});
-}
+ $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);
+# $html .= "-->".$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 $config->{isis_data}."/$db_dir/$_/LIBRI") {
- push @isis_dbs,$config->{isis_data}."/$db_dir/$_/LIBRI/LIBRI";
}
- if (-e $config->{isis_data}."/$db_dir/$_/PERI") {
- push @isis_dbs,$config->{isis_data}."/$db_dir/$_/PERI/PERI";
- }
- if (-e $config->{isis_data}."/$db_dir/$_/AMS") {
- push @isis_dbs,$config->{isis_data}."/$db_dir/$_/AMS/AMS";
+
+ # dump formatted output in
+ if ($html) {
+ $xml .= xmlify("html",$html);
}
- if (-e $config->{isis_data}."/$db_dir/$_/ARTI") {
-# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI";
+
+ if ($xml) {
+#print STDERR "x: $xml\n";
+ $xml .= $add_xml if ($add_xml);
+ return "\n$xml\n";
+ } else {
+ return;
}
}
-print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs);
+##########################################################################
-my $db;
+my $cfg = new Config::IniFiles( -file => $config_file );
-foreach my $isis_db (@isis_dbs) {
+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},$XML_CHARSET);
+ $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage});
+ $xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),$XML_CHARSET);
my $db = OpenIsis::open( $isis_db );
if (0) {
@@ -152,14 +244,16 @@
print STDERR "Reading database: $isis_db [$max_rowid rows]\n";
+ my $path = $database; # was $isis_db
+
my $last_p = 0;
-# { my $row_id = 1;
+# { my $row_id = 4514;
# FIX
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) {
my $row = OpenIsis::read( $db, $row_id );
if ($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) {
@@ -167,8 +261,9 @@
$last_p = $p;
}
- if (my $xml = isis2xml($row)) {
- print "Path-Name: $isis_db#".$row->{mfn}."\n";
+ 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";
}
@@ -177,6 +272,8 @@
print STDERR "\n";
}
+# call this to commit index
+$index->close;
1;
__END__