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 |
|
|
19 |
my $config; |
my $config; |
20 |
|
|
21 |
use index_DBI; # there is no other, right now ;-) |
use index_DBI; # there is no other, right now ;-) |
22 |
|
my $index; |
|
my $index = new index_DBI(); # open index |
|
23 |
|
|
24 |
my %opts; |
my %opts; |
25 |
|
|
31 |
|
|
32 |
getopts('d:m:qs', \%opts); |
getopts('d:m:qs', \%opts); |
33 |
|
|
34 |
my $db_dir; |
my $path; # this is name of database |
35 |
|
|
36 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
37 |
|
|
38 |
#print Dumper($config->{indexer}); |
# this is encoding of all files on disk, including import_xml/*.xml file and |
39 |
#print "-" x 70,"\n"; |
# filter/*.pm files! It will be used to store strings in perl internally! |
40 |
|
my $codepage = 'ISO-8859-2'; |
41 |
|
|
42 |
|
my $utf2cp = Text::Iconv->new('UTF-8',$codepage); |
43 |
|
# this function will convert data from XML files to local encoding |
44 |
|
sub x { |
45 |
|
return $utf2cp->convert($_[0]); |
46 |
|
} |
47 |
|
|
48 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
# decode isis import codepage |
49 |
|
my $isis2cp; |
50 |
|
|
51 |
#my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
# outgoing xml must be in UTF-8 |
52 |
#my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
|
my $isis_codepage; |
|
|
my $index_codepage; |
|
|
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
|
|
my $xml_codepage; |
|
53 |
|
|
54 |
sub isis2xml { |
sub isis2xml { |
55 |
|
|
66 |
|
|
67 |
my %field_usage; # counter for usage of each field |
my %field_usage; # counter for usage of each field |
68 |
|
|
69 |
foreach my $field (keys %{$config->{indexer}}) { |
# sort subrouting using order="" attribute |
70 |
|
sub by_order { |
71 |
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
72 |
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
73 |
|
|
74 |
|
return $config->{indexer}->{$a}->{order} <=> |
75 |
|
$config->{indexer}->{$b}->{order} ; |
76 |
|
} |
77 |
|
|
78 |
|
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
79 |
|
|
80 |
|
$field=x($field); |
81 |
|
|
82 |
$field_usage{$field}++; |
$field_usage{$field}++; |
83 |
|
|
84 |
my $swish_data = ""; |
my $swish_data = ""; |
85 |
my $display_data = ""; |
my $display_data = ""; |
86 |
|
my $line_delimiter; |
87 |
|
|
88 |
|
my ($swish,$display); |
89 |
|
|
90 |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
foreach my $x (@{$config->{indexer}->{$field}->{isis}}) { |
91 |
|
|
92 |
my $format = $x->{content}; |
my $format = x($x->{content}); |
93 |
|
my $delimiter = x($x->{delimiter}) || ' '; |
94 |
|
|
95 |
|
my $isis_i = 0; # isis repeatable offset |
96 |
|
|
97 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
my ($s,$d,$i) = (1,1,0); # swish, display default |
98 |
$s = 0 if (lc($x->{type}) eq "display"); |
$s = 0 if (lc($x->{type}) eq "display"); |
99 |
$d = 0 if (lc($x->{type}) eq "swish"); |
$d = 0 if (lc($x->{type}) eq "swish"); |
100 |
($s,$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 ## $format ##\n"; |
|
101 |
|
|
102 |
# FIX: this is ugly, UGLY, cludge string is returned |
# what will separate last line from this one? |
103 |
# in UTF8 encoding , but as if source charset |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
104 |
# is ISO8859-1 and not some other. This break other |
$line_delimiter = ' '; |
105 |
# character encodings, so we convert it first |
} elsif ($display_data) { |
106 |
# back to ISO8859-1 |
$line_delimiter = '<br/>'; |
|
$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 $filter; |
|
107 |
} |
} |
108 |
# type="swish" ; field for swish |
|
109 |
if ($s && $swish) { |
# init vars so that we go into while... |
110 |
|
($swish,$display) = (1,1); |
111 |
|
|
112 |
|
while ($swish || $display) { |
113 |
|
($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp); |
114 |
|
|
115 |
|
# filter="name" ; filter this field through |
116 |
|
# filter/[name].pm |
117 |
|
my $filter = $x->{filter}; |
118 |
if ($filter) { |
if ($filter) { |
119 |
$swish_data .= join(" ",&filter($swish)); |
require "filter/".$filter.".pm"; |
120 |
} else { |
} |
121 |
$swish_data .= $swish if ($s && $swish); |
# type="swish" ; field for swish |
122 |
|
if ($s && $swish) { |
123 |
|
if ($filter) { |
124 |
|
no strict 'refs'; |
125 |
|
$swish_data .= join(" ",&$filter($swish)); |
126 |
|
} else { |
127 |
|
$swish_data .= $swish; |
128 |
|
} |
129 |
} |
} |
|
} |
|
130 |
|
|
131 |
# type="display" ; field for display |
# type="display" ; field for display |
132 |
if ($d && $display) { |
if ($d && $display) { |
133 |
if ($filter) { |
if ($line_delimiter && $display_data) { |
134 |
$display_data .= join(" ",&filter($display)); |
$display_data .= $line_delimiter; |
135 |
} else { |
undef $line_delimiter; |
136 |
$display_data .= $display if ($s && $display); |
} |
137 |
|
if ($filter) { |
138 |
|
no strict 'refs'; |
139 |
|
$display_data .= join($delimiter,&$filter($display)); |
140 |
|
} else { |
141 |
|
if ($display_data) { |
142 |
|
$display_data .= $delimiter.$display; |
143 |
|
} else { |
144 |
|
$display_data .= $display; |
145 |
|
} |
146 |
|
} |
147 |
} |
} |
148 |
} |
|
149 |
|
# type="index" ; insert into index |
150 |
# type="index" ; insert into index |
if ($i && $display) { |
151 |
if ($i && $display) { |
my $index_data = $display; |
152 |
my $index_data = $index_codepage->convert($display) || $display; |
if ($filter) { |
153 |
if ($filter) { |
no strict 'refs'; |
154 |
foreach my $d (&filter($index_data)) { |
foreach my $d (&$filter($index_data)) { |
155 |
$index->insert($field, $d, $db_dir); |
$index->insert($field, $d, $path); |
156 |
|
} |
157 |
|
} else { |
158 |
|
$index->insert($field, $index_data, $path); |
159 |
} |
} |
|
} else { |
|
|
$index->insert($field, $index_data, $db_dir); |
|
160 |
} |
} |
161 |
} |
} |
162 |
} |
} |
163 |
|
|
164 |
|
|
|
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
|
165 |
if ($display_data) { |
if ($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", $display_data); |
|
166 |
|
|
167 |
if ($field eq "headline") { |
if ($field eq "headline") { |
168 |
$xml .= xmlify("headline", $display_data); |
$xml .= xmlify("headline", $display_data); |
180 |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
181 |
} |
} |
182 |
if ($field_name) { |
if ($field_name) { |
183 |
$html .= $xml_codepage->convert($field_name); |
$html .= x($field_name); |
184 |
} |
} |
185 |
$html .= $display_data."###\n"; |
$html .= $display_data."###\n"; |
186 |
} |
} |
187 |
} |
} |
188 |
if ($swish_data) { |
if ($swish_data) { |
|
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
|
189 |
# remove extra spaces |
# remove extra spaces |
190 |
$swish_data =~ s/ +/ /g; |
$swish_data =~ s/ +/ /g; |
191 |
$swish_data =~ s/ +$//g; |
$swish_data =~ s/ +$//g; |
192 |
|
|
193 |
$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},$swish_data)); |
|
194 |
} |
} |
195 |
|
|
196 |
|
|
202 |
} |
} |
203 |
|
|
204 |
if ($xml) { |
if ($xml) { |
|
#print STDERR "x: $xml\n"; |
|
205 |
$xml .= $add_xml if ($add_xml); |
$xml .= $add_xml if ($add_xml); |
206 |
return "<xml>\n$xml</xml>\n"; |
return "<xml>\n$xml</xml>\n"; |
207 |
} else { |
} else { |
213 |
|
|
214 |
my $cfg = new Config::IniFiles( -file => $config_file ); |
my $cfg = new Config::IniFiles( -file => $config_file ); |
215 |
|
|
216 |
|
# open index |
217 |
|
$index = new index_DBI( |
218 |
|
$cfg->val('global', 'dbi_dbd'), |
219 |
|
$cfg->val('global', 'dbi_dsn'), |
220 |
|
$cfg->val('global', 'dbi_user'), |
221 |
|
$cfg->val('global', 'dbi_passwd') || '', |
222 |
|
); |
223 |
|
|
224 |
|
# delete [global] section to leave just databases sections |
225 |
|
$cfg->DeleteSection('global'); |
226 |
|
|
227 |
foreach my $database ($cfg->Sections) { |
foreach my $database ($cfg->Sections) { |
228 |
|
|
229 |
my $isis_db = $cfg -> val($database, 'isis_db'); |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
230 |
my $type = $cfg -> val($database, 'type'); |
my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined"; |
231 |
my $add_xml = $cfg -> val($database, 'xml'); |
my $add_xml = $cfg -> val($database, 'xml'); # optional |
232 |
|
|
|
# read configuration for this type |
|
233 |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
234 |
$isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
|
235 |
$index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
$isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
|
$xml_codepage = Text::Iconv->new($cfg->val($database,'xml_codepage'),'UTF8'); |
|
236 |
|
|
237 |
my $db = OpenIsis::open( $isis_db ); |
my $db = OpenIsis::open( $isis_db ); |
238 |
if (0) { |
if (0) { |
255 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
256 |
my $row = OpenIsis::read( $db, $row_id ); |
my $row = OpenIsis::read( $db, $row_id ); |
257 |
if ($row && $row->{mfn}) { |
if ($row && $row->{mfn}) { |
|
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
258 |
# output current process indicator |
# output current process indicator |
259 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
my $p = int($row->{mfn} * 100 / $max_rowid); |
260 |
if ($p != $last_p) { |
if ($p != $last_p) { |
262 |
$last_p = $p; |
$last_p = $p; |
263 |
} |
} |
264 |
|
|
265 |
|
my $swishpath = $path."#".int($row->{mfn}); |
266 |
|
|
267 |
if (my $xml = isis2xml($row,$add_xml)) { |
if (my $xml = isis2xml($row,$add_xml)) { |
268 |
#print STDERR "--ret-->$xml\n"; |
$xml = $cp2utf->convert($xml); |
269 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
use bytes; # as opposed to chars |
270 |
|
print "Path-Name: $swishpath\n"; |
271 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |
272 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |
273 |
} |
} |