--- trunk/all2xml.pl 2003/02/23 07:08:54 29
+++ trunk/all2xml.pl 2003/06/01 13:46:42 50
@@ -8,6 +8,7 @@
use Text::Unaccent 1.02; # 1.01 won't compile on my platform,
use Text::Iconv;
use Config::IniFiles;
+use Encode;
$|=1;
@@ -18,8 +19,7 @@
my $config;
use index_DBI; # there is no other, right now ;-)
-
-my $index = new index_DBI(); # open index
+my $index;
my %opts;
@@ -31,22 +31,25 @@
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";
+my $path; # this is name of database
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;
+# 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]);
+}
-my $XML_CHARSET = 'UTF8';
+# decode isis import codepage
+my $isis2cp;
+# outgoing xml must be in UTF-8
+my $cp2utf = Text::Iconv->new($codepage,'UTF-8');
sub isis2xml {
@@ -63,27 +66,31 @@
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}}) {
- foreach my $field (keys %{$config->{indexer}}) {
+ $field=x($field);
$field_usage{$field}++;
my $swish_data = "";
my $display_data = "";
- my $line_delimiter = "";
+ my $line_delimiter;
- foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
+ my ($swish,$display);
- my $format = $x->{content};
- my $delimiter = $x->{delimiter} || ' ';
+ foreach my $x (@{$config->{indexer}->{$field}->{isis}}) {
- # 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 $format = x($x->{content});
+ my $delimiter = x($x->{delimiter}) || ' ';
my $isis_i = 0; # isis repeatable offset
@@ -91,17 +98,19 @@
$s = 0 if (lc($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$/);
+ # what will separate last line from this one?
+ if ($display_data && $x->{append} && $x->{append} eq "1") {
+ $line_delimiter = ' ';
+ } elsif ($display_data) {
+ $line_delimiter = '
';
+ }
- my ($swish,$display) = (1,1);
+ # init vars so that we go into while...
+ ($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: $display<--\n";
+ ($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp);
# filter="name" ; filter this field through
# filter/[name].pm
@@ -112,7 +121,6 @@
# type="swish" ; field for swish
if ($s && $swish) {
if ($filter) {
-print STDERR "using filter '$filter'\n";
no strict 'refs';
$swish_data .= join(" ",&$filter($swish));
} else {
@@ -122,6 +130,10 @@
# type="display" ; field for display
if ($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));
@@ -136,35 +148,22 @@
# type="index" ; insert into index
if ($i && $display) {
- my $index_data = $index_codepage->convert($display) || $display;
+ my $index_data = $display;
if ($filter) {
no strict 'refs';
foreach my $d (&$filter($index_data)) {
- $index->insert($field, $d, $db_dir);
+ $index->insert($field, $d, $path);
}
} else {
- $index->insert($field, $index_data, $db_dir);
+ $index->insert($field, $index_data, $path);
}
}
}
- if ($x->{append}) {
- $line_delimiter = ' ';
- } else {
- $line_delimiter = '
';
- }
}
-#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data);
if ($display_data) {
- # remove last
- $display_data =~ s/$line_delimiter$//;
-
- $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", $display_data);
} else {
@@ -181,22 +180,17 @@
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!";
}
if ($field_name) {
- $html .= $xml_codepage->convert($field_name);
-# $html .= "-->".$field_name."<--";
+ $html .= x($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));
+ $xml .= xmlify($field."_swish", unac_string($codepage,$swish_data));
}
@@ -208,7 +202,6 @@
}
if ($xml) {
-#print STDERR "x: $xml\n";
$xml .= $add_xml if ($add_xml);
return "\n$xml\n";
} else {
@@ -220,17 +213,26 @@
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'),
+ $cfg->val('global', 'dbi_passwd') || '',
+ );
+
+# delete [global] section to leave just databases sections
+$cfg->DeleteSection('global');
+
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');
+ 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
- # 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);
+
+ $isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage);
my $db = OpenIsis::open( $isis_db );
if (0) {
@@ -253,7 +255,6 @@
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) {
@@ -261,9 +262,12 @@
$last_p = $p;
}
+ my $swishpath = $path."#".int($row->{mfn});
+
if (my $xml = isis2xml($row,$add_xml)) {
-#print STDERR "--ret-->$xml\n";
- print "Path-Name: $path#".int($row->{mfn})."\n";
+ $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";
}