7 |
use XML::Simple; |
use XML::Simple; |
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; |
11 |
|
use Encode; |
12 |
|
|
13 |
$|=1; |
$|=1; |
14 |
|
|
15 |
my $config; |
my $config_file = $0; |
16 |
|
$config_file =~ s/\.pl$/.conf/; |
17 |
|
die "FATAL: can't find configuration file '$config_file'" if (! -e $config_file); |
18 |
|
|
19 |
$config=XMLin(undef, forcearray => [ 'isis' ], forcecontent => 1); |
my $config; |
20 |
|
|
21 |
use index_DBI; # there is no other, right now ;-) |
use index_DBI; # there is no other, right now ;-) |
22 |
|
|
32 |
|
|
33 |
getopts('d:m:qs', \%opts); |
getopts('d:m:qs', \%opts); |
34 |
|
|
35 |
my $db_dir = $opts{d} || "ps"; # FIX |
my $db_dir; |
36 |
|
|
37 |
#die "usage: $0 -d [database_dir] -m [database1,database2] " if (! %opts); |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
38 |
|
|
39 |
#print Dumper($config->{indexer}); |
# this is encoding of all files on disk, including import_xml/*.xml file and |
40 |
#print "-" x 70,"\n"; |
# filter/*.pm files! It will be used to store strings in perl internally! |
41 |
|
my $codepage = 'ISO-8859-2'; |
42 |
|
|
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 |
Text::Iconv->raise_error(1); # Conversion errors raise exceptions |
# decode isis import codepage |
50 |
|
my $isis2cp; |
51 |
|
|
52 |
my $isis_codepage = Text::Iconv->new($config->{isis_codepage},'UTF8'); |
# outgoing xml must be in UTF-8 |
53 |
my $index_codepage = Text::Iconv->new($config->{isis_codepage},$config->{index_codepage}); |
my $cp2utf = Text::Iconv->new($codepage,'UTF-8'); |
|
my $cludge_codepage = Text::Iconv->new('UTF8','ISO8859-1'); |
|
54 |
|
|
55 |
sub isis2xml { |
sub isis2xml { |
56 |
|
|
57 |
use xmlify; |
use xmlify; |
58 |
|
|
59 |
my $row = shift @_; |
my $row = shift @_; |
60 |
|
my $add_xml = shift @_; |
61 |
|
|
62 |
my $xml; |
my $xml; |
|
$xml .= xmlify('db_dir',$db_dir); # FIX remove? |
|
63 |
|
|
64 |
use parse_format; |
use parse_format; |
65 |
|
|
66 |
|
my $html = ""; # html formatted display output |
67 |
|
|
68 |
|
my %field_usage; # counter for usage of each field |
69 |
|
|
70 |
|
# sort subrouting using order="" attribute |
71 |
|
sub by_order { |
72 |
|
return 0 if (! $config->{indexer}->{$a}->{order}); |
73 |
|
return 0 if (! $config->{indexer}->{$b}->{order}); |
74 |
|
|
75 |
|
return $config->{indexer}->{$a}->{order} <=> |
76 |
|
$config->{indexer}->{$b}->{order} ; |
77 |
|
} |
78 |
|
|
79 |
foreach my $field (keys %{$config->{indexer}}) { |
foreach my $field (sort by_order keys %{$config->{indexer}}) { |
80 |
|
|
81 |
|
$field=x($field); |
82 |
|
|
83 |
|
$field_usage{$field}++; |
84 |
|
|
85 |
my $swish_data = ""; |
my $swish_data = ""; |
86 |
my $display_data = ""; |
my $display_data = ""; |
87 |
my $index_data = ""; |
my $line_delimiter; |
88 |
|
|
89 |
|
my ($swish,$display); |
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 $delimiter = x($x->{delimiter}) || ' '; |
95 |
|
|
96 |
|
my $isis_i = 0; # isis repeatable offset |
97 |
|
|
98 |
my ($s,$d,$i) = (1,1,0); # swish, display default |
my ($s,$d,$i) = (1,1,0); # swish, display default |
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 |
$format = $cludge_codepage->convert($format); |
# what will separate last line from this one? |
104 |
my ($swish,$display) = parse_format($format,$row); |
if ($display_data && $x->{append} && $x->{append} eq "1") { |
105 |
#print STDERR "s: $swish\nd: $display\n" if ($swish); |
$line_delimiter = ' '; |
106 |
|
} elsif ($display_data) { |
107 |
#print STDERR "swish: $swish<-- display: $display<--\n"; |
$line_delimiter = '<br/>'; |
108 |
# FIX: this is ugly, UGLY, cludge: OpenIsis return |
} |
109 |
# UTF8 encoding of strings, but as if source charset |
|
110 |
# is ISO8859-1 and not some other. This breaks our |
# init vars so that we go into while... |
111 |
# isis character encoding, so we convert it first |
($swish,$display) = (1,1); |
112 |
# back to ISO8859-1 (which can actually be different |
|
113 |
# encoding in isis) |
while ($swish || $display) { |
114 |
|
($swish,$display) = parse_format($format,$row,$isis_i++,$isis2cp); |
115 |
$swish_data .= $swish if ($s && $swish); |
|
116 |
$display_data .= $display if ($d && $display); |
# filter="name" ; filter this field through |
117 |
$index_data .= $display if ($i && $display); |
# filter/[name].pm |
118 |
|
my $filter = $x->{filter}; |
119 |
|
if ($filter) { |
120 |
|
require "filter/".$filter.".pm"; |
121 |
|
} |
122 |
|
# type="swish" ; field for swish |
123 |
|
if ($s && $swish) { |
124 |
|
if ($filter) { |
125 |
|
no strict 'refs'; |
126 |
|
$swish_data .= join(" ",&$filter($swish)); |
127 |
|
print STDERR "#### $swish_data\n"; |
128 |
|
} else { |
129 |
|
$swish_data .= $swish; |
130 |
|
} |
131 |
|
} |
132 |
|
|
133 |
|
# type="display" ; field for display |
134 |
|
if ($d && $display) { |
135 |
|
if ($line_delimiter && $display_data) { |
136 |
|
$display_data .= $line_delimiter; |
137 |
|
undef $line_delimiter; |
138 |
|
} |
139 |
|
if ($filter) { |
140 |
|
no strict 'refs'; |
141 |
|
$display_data .= join($delimiter,&$filter($display)); |
142 |
|
} else { |
143 |
|
if ($display_data) { |
144 |
|
$display_data .= $delimiter.$display; |
145 |
|
} else { |
146 |
|
$display_data .= $display; |
147 |
|
} |
148 |
|
} |
149 |
|
} |
150 |
|
|
151 |
|
# type="index" ; insert into index |
152 |
|
if ($i && $display) { |
153 |
|
my $index_data = $display; |
154 |
|
if ($filter) { |
155 |
|
no strict 'refs'; |
156 |
|
foreach my $d (&$filter($index_data)) { |
157 |
|
$index->insert($field, $d, $db_dir); |
158 |
|
} |
159 |
|
} else { |
160 |
|
$index->insert($field, $index_data, $db_dir); |
161 |
|
} |
162 |
|
} |
163 |
|
} |
164 |
} |
} |
165 |
|
|
166 |
#print STDERR "s_d: $swish_data\nd_d: $display_data\n" if ($swish_data); |
|
167 |
if ($display_data) { |
if ($display_data) { |
168 |
$display_data = $isis_codepage->convert($display_data)."##" || $display_data; |
|
169 |
$xml .= xmlify($field."_display", $display_data); |
if ($field eq "headline") { |
170 |
|
$xml .= xmlify("headline", $display_data); |
171 |
|
} else { |
172 |
|
|
173 |
|
# find field name (signular, plural) |
174 |
|
my $field_name = ""; |
175 |
|
if ($config->{indexer}->{$field}->{name_singular} && $field_usage{$field} == 1) { |
176 |
|
$field_name = $config->{indexer}->{$field}->{name_singular}."#-#"; |
177 |
|
} elsif ($config->{indexer}->{$field}->{name_plural}) { |
178 |
|
$field_name = $config->{indexer}->{$field}->{name_plural}."#-#"; |
179 |
|
} elsif ($config->{indexer}->{$field}->{name}) { |
180 |
|
$field_name = $config->{indexer}->{$field}->{name}."#-#"; |
181 |
|
} else { |
182 |
|
print STDERR "WARNING: field '$field' doesn't have 'name' attribute!"; |
183 |
|
} |
184 |
|
if ($field_name) { |
185 |
|
$html .= x($field_name); |
186 |
|
} |
187 |
|
$html .= $display_data."###\n"; |
188 |
|
} |
189 |
} |
} |
190 |
if ($swish_data) { |
if ($swish_data) { |
191 |
my $i = Text::Iconv->new($config->{isis_codepage},'ISO8859-2'); |
# remove extra spaces |
192 |
$swish_data = $i->convert($swish_data); |
$swish_data =~ s/ +/ /g; |
193 |
$xml .= xmlify($field."_swish",unac_string('ISO8859-2',$swish_data)); |
$swish_data =~ s/ +$//g; |
|
#$swish_data = $isis_codepage->convert($swish_data)."##" || $swish_data; |
|
|
#$xml .= xmlify($field."_swish",unac_string($config->{isis_codepage},$swish_data)); |
|
|
} |
|
194 |
|
|
195 |
# index |
$xml .= xmlify($field."_swish", unac_string($codepage,$swish_data)); |
|
if ($index_data && $index_data ne "") { |
|
|
$index_data = $index_codepage->convert($index_data) || $index_data; |
|
|
$index->insert($field, $index_data, $db_dir); |
|
196 |
} |
} |
197 |
|
|
198 |
|
|
199 |
|
} |
200 |
|
|
201 |
|
# dump formatted output in <html> |
202 |
|
if ($html) { |
203 |
|
$xml .= xmlify("html",$html); |
204 |
} |
} |
205 |
|
|
206 |
if ($xml) { |
if ($xml) { |
207 |
#print STDERR "x: $xml\n"; |
$xml .= $add_xml if ($add_xml); |
208 |
return "<xml>\n$xml</xml>\n"; |
return "<xml>\n$xml</xml>\n"; |
209 |
} else { |
} else { |
210 |
return; |
return; |
213 |
|
|
214 |
########################################################################## |
########################################################################## |
215 |
|
|
216 |
my $last_tell=0; |
my $cfg = new Config::IniFiles( -file => $config_file ); |
|
|
|
|
my @isis_dirs = ( '.' ); # use dirname as database name |
|
|
|
|
|
if ($opts{m}) { |
|
|
@isis_dirs = split(/,/,$opts{m}); |
|
|
} |
|
|
|
|
|
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"; |
|
|
} |
|
|
if (-e $config->{isis_data}."/$db_dir/$_/ARTI") { |
|
|
# push @isis_dbs,$config->{isis_data}."/$db_dir/$_/ARTI/ARTI"; |
|
|
} |
|
|
} |
|
217 |
|
|
218 |
print STDERR "FATAL: Can't find isis database.\nPerhaps isis_data (".$config->{isis_data}.") has wrong value?\n" if (! @isis_dbs); |
foreach my $database ($cfg->Sections) { |
219 |
|
|
220 |
my $db; |
my $isis_db = $cfg -> val($database, 'isis_db') || die "$database doesn't have 'isis_db' defined!"; |
221 |
|
my $type = $cfg -> val($database, 'type') || die "$database doesn't have 'type' defined"; |
222 |
|
my $add_xml = $cfg -> val($database, 'xml'); # optional |
223 |
|
|
224 |
foreach my $isis_db (@isis_dbs) { |
$config=XMLin("./import_xml/$type.xml", forcearray => [ 'isis' ], forcecontent => 1); |
225 |
|
|
226 |
|
$isis2cp = Text::Iconv->new($config->{isis_codepage},$codepage); |
227 |
|
|
228 |
my $db = OpenIsis::open( $isis_db ); |
my $db = OpenIsis::open( $isis_db ); |
229 |
if (0) { |
if (0) { |
237 |
|
|
238 |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
print STDERR "Reading database: $isis_db [$max_rowid rows]\n"; |
239 |
|
|
240 |
my $path = $isis_db; |
my $path = $database; # was $isis_db |
|
$path =~ s#$config->{isis_data}/*##g; |
|
241 |
|
|
242 |
my $last_p = 0; |
my $last_p = 0; |
243 |
|
|
246 |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
for (my $row_id = 1; $row_id <= $max_rowid; $row_id++ ) { |
247 |
my $row = OpenIsis::read( $db, $row_id ); |
my $row = OpenIsis::read( $db, $row_id ); |
248 |
if ($row && $row->{mfn}) { |
if ($row && $row->{mfn}) { |
|
#print STDERR "mfn: ",$row->{mfn},"\n"; |
|
249 |
# output current process indicator |
# output current process indicator |
250 |
my $p = int($row->{mfn} * 100 / $max_rowid); |
my $p = int($row->{mfn} * 100 / $max_rowid); |
251 |
if ($p != $last_p) { |
if ($p != $last_p) { |
253 |
$last_p = $p; |
$last_p = $p; |
254 |
} |
} |
255 |
|
|
256 |
if (my $xml = isis2xml($row)) { |
if (my $xml = $cp2utf->convert(isis2xml($row,$add_xml))) { |
257 |
#print STDERR "--ret-->$xml\n"; |
use bytes; # as opposed to chars |
258 |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
print "Path-Name: $path#".int($row->{mfn})."\n"; |
259 |
print "Content-Length: ".(length($xml)+1)."\n"; |
print "Content-Length: ".(length($xml)+1)."\n"; |
260 |
print "Document-Type: XML\n\n$xml\n"; |
print "Document-Type: XML\n\n$xml\n"; |