8 |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
use Text::Unaccent 1.02; # 1.01 won't compile on my platform, |
9 |
use Text::Iconv; |
use Text::Iconv; |
10 |
use Config::IniFiles; |
use Config::IniFiles; |
11 |
|
use Encode; |
12 |
|
|
13 |
$|=1; |
$|=1; |
14 |
|
|
34 |
|
|
35 |
my $db_dir; |
my $db_dir; |
36 |
|
|
|
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
|
|
|
|
|
#print Dumper($config->{indexer}); |
|
|
#print "-" x 70,"\n"; |
|
|
|
|
37 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
38 |
|
|
39 |
my $isis_codepage; |
# this is encoding of all files on disk, including import_xml/*.xml file and |
40 |
my $index_codepage; |
# filter/*.pm files! It will be used to store strings in perl internally! |
41 |
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
my $codepage = 'ISO-8859-2'; |
42 |
my $xml_codepage; |
|
43 |
|
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
44 |
|
# this function will convert data from XML files to local encoding |
45 |
|
sub x { |
46 |
|
return $utf2cp->convert($_[0]); |
47 |
|
} |
48 |
|
|
49 |
my $XML_CHARSET = 'UTF8'; |
# decode isis import codepage |
50 |
|
my $isis2cp; |
51 |
|
|
52 |
|
# outgoing xml must be in UTF-8 |
53 |
|
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
54 |
|
|
55 |
sub isis2xml { |
sub isis2xml { |
56 |
|
|
78 |
|
|
79 |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
80 |
|
|
81 |
|
$field=x($field); |
82 |
|
|
83 |
$field_usage{$field}++; |
$field_usage{$field}++; |
84 |
|
|
85 |
my $swish_data = ""; |
my $swish_data = ""; |
90 |
|
|
91 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
92 |
|
|
93 |
my $format = $x->{content}; |
my $format = x($x->{content}); |
94 |
my $delimiter = $x->{delimiter} || ' '; |
my $delimiter = x($x->{delimiter}) || ' '; |
|
|
|
|
# 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); |
|
95 |
|
|
96 |
my $isis_i = 0; # isis repeatable offset |
my $isis_i = 0; # isis repeatable offset |
97 |
|
|
99 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
100 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
101 |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
($s,$d,$i) = (0,0,1) if (lc($x->{type}) eq "index"); |
|
#print STDERR "## s: $s d: $d i: $i ## $format ##\n"; |
|
102 |
|
|
103 |
# what will separate last line from this one? |
# what will separate last line from this one? |
104 |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
111 |
($swish,$display) = (1,1); |
($swish,$display) = (1,1); |
112 |
|
|
113 |
while ($swish || $display) { |
while ($swish || $display) { |
114 |
($swish,$display) = parse_format($format,$row,$isis_i++); |
($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp); |
|
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
|
|
|
|
|
#print STDERR "swish: $swish<-- display: $display<--\n"; |
|
115 |
|
|
116 |
# filter="name" ; filter this field through |
# filter="name" ; filter this field through |
117 |
# filter/[name].pm |
# filter/[name].pm |
122 |
# type="swish" ; field for swish |
# type="swish" ; field for swish |
123 |
if ($s && $swish) { |
if ($s && $swish) { |
124 |
if ($filter) { |
if ($filter) { |
|
#print STDERR "using filter '$filter'\n"; |
|
125 |
no strict 'refs'; |
no strict 'refs'; |
126 |
$swish_data .= join(" ",&$filter($swish)); |
$swish_data .= join(" ",&$filter($swish)); |
127 |
} else { |
} else { |
149 |
|
|
150 |
# type="index" ; insert into index |
# type="index" ; insert into index |
151 |
if ($i && $display) { |
if ($i && $display) { |
152 |
my $index_data = $index_codepage->convert($display) || $display; |
my $index_data = $display; |
153 |
if ($filter) { |
if ($filter) { |
154 |
no strict 'refs'; |
no strict 'refs'; |
155 |
foreach my $d (&$filter($index_data)) { |
foreach my $d (&$filter($index_data)) { |
163 |
} |
} |
164 |
|
|
165 |
|
|
|
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
|
166 |
if ($display_data) { |
if ($display_data) { |
167 |
|
|
|
# $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", $display_data); |
|
|
|
|
168 |
if ($field eq "headline") { |
if ($field eq "headline") { |
169 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
170 |
} else { |
} else { |
181 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
182 |
} |
} |
183 |
if ($field_name) { |
if ($field_name) { |
184 |
$html .= $xml_codepage->convert($field_name); |
$html .= x($field_name); |
185 |
} |
} |
186 |
$html .= $display_data."###\n"; |
$html .= $display_data."###\n"; |
187 |
} |
} |
188 |
} |
} |
189 |
if ($swish_data) { |
if ($swish_data) { |
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
|
190 |
# remove extra spaces |
# remove extra spaces |
191 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
192 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
193 |
|
|
194 |
$swish_data = $i->convert($swish_data); |
$xml .= xmlify($field."_swish", unac_string($codepage,$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},$swish_data)); |
|
195 |
} |
} |
196 |
|
|
197 |
|
|
203 |
} |
} |
204 |
|
|
205 |
if ($xml) { |
if ($xml) { |
|
#print STDERR "x: $xml\n"; |
|
206 |
$xml .= $add_xml if ($add_xml); |
$xml .= $add_xml if ($add_xml); |
207 |
return "<xml>\n$xml</xml>\n"; |
return "<xml>\n$xml</xml>\n"; |
208 |
} else { |
} else { |
216 |
|
|
217 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
218 |
|
|
219 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
220 |
my $type = $cfg -> val($database, 'type'); |
my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined"; |
221 |
my $add_xml = $cfg -> val($database, 'xml'); |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
222 |
|
|
|
# read configuration for this type |
|
223 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
224 |
$isis_codepage = Text::Iconv->new($config->{isis_codepage},$XML_CHARSET); |
|
225 |
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
$isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
|
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),$XML_CHARSET); |
|
226 |
|
|
227 |
my $db = OpenIsis::open( $isis_db ); |
my $db = OpenIsis::open( $isis_db ); |
228 |
if (0) { |
if (0) { |
245 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
246 |
my $row = OpenIsis::read( $db, $row_id ); |
my $row = OpenIsis::read( $db, $row_id ); |
247 |
if ($row && $row->{mfn}) { |
if ($row && $row->{mfn}) { |
|
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
248 |
# output current process indicator |
# output current process indicator |
249 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
my $p = int($row->{mfn} * 100 / $max_rowid); |
250 |
if ($p != $last_p) { |
if ($p != $last_p) { |
252 |
$last_p = $p; |
$last_p = $p; |
253 |
} |
} |
254 |
|
|
255 |
if (my $xml = isis2xml($row,$add_xml)) { |
if (my $xml = $cp2utf->convert(isis2xml($row,$add_xml))) { |
|
#print STDERR "--ret-->$xml\n"; |
|
256 |
use bytes; # as opposed to chars |
use bytes; # as opposed to chars |
257 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
258 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |