line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package Alvis::Convert;
|
2
|
|
|
|
|
|
|
|
3
|
|
|
|
|
|
|
$Alvis::Convert::VERSION = '0.4';
|
4
|
|
|
|
|
|
|
|
5
|
|
|
|
|
|
|
########################################################################
|
6
|
|
|
|
|
|
|
#
|
7
|
|
|
|
|
|
|
# A general "set of document files in some format" ->
|
8
|
|
|
|
|
|
|
# "set of files in ALVIS format" converter.
|
9
|
|
|
|
|
|
|
#
|
10
|
|
|
|
|
|
|
# -- Kimmo Valtonen
|
11
|
|
|
|
|
|
|
#
|
12
|
|
|
|
|
|
|
########################################################################
|
13
|
|
|
|
|
|
|
|
14
|
1
|
|
|
1
|
|
24461
|
use strict;
|
|
1
|
|
|
|
|
3
|
|
|
1
|
|
|
|
|
37
|
|
15
|
1
|
|
|
1
|
|
5
|
use warnings;
|
|
1
|
|
|
|
|
2
|
|
|
1
|
|
|
|
|
29
|
|
16
|
|
|
|
|
|
|
|
17
|
1
|
|
|
1
|
|
6
|
use Carp;
|
|
1
|
|
|
|
|
6
|
|
|
1
|
|
|
|
|
92
|
|
18
|
1
|
|
|
1
|
|
1090
|
use Data::Dumper;
|
|
1
|
|
|
|
|
23561
|
|
|
1
|
|
|
|
|
226
|
|
19
|
1
|
|
|
1
|
|
3602
|
use Encode;
|
|
1
|
|
|
|
|
21492
|
|
|
1
|
|
|
|
|
106
|
|
20
|
1
|
|
|
1
|
|
487
|
use XML::LibXML;
|
|
0
|
|
|
|
|
|
|
|
0
|
|
|
|
|
|
|
21
|
|
|
|
|
|
|
|
22
|
|
|
|
|
|
|
use Alvis::Canonical;
|
23
|
|
|
|
|
|
|
use Alvis::Document;
|
24
|
|
|
|
|
|
|
use Alvis::Document::Encoding;
|
25
|
|
|
|
|
|
|
use Alvis::Document::Meta;
|
26
|
|
|
|
|
|
|
use Alvis::Document::Links;
|
27
|
|
|
|
|
|
|
use Alvis::Document::Type;
|
28
|
|
|
|
|
|
|
use Alvis::AinoDump;
|
29
|
|
|
|
|
|
|
use Alvis::Wikipedia::XMLDump;
|
30
|
|
|
|
|
|
|
|
31
|
|
|
|
|
|
|
|
32
|
|
|
|
|
|
|
############################################################################
|
33
|
|
|
|
|
|
|
#
|
34
|
|
|
|
|
|
|
# Global variables
|
35
|
|
|
|
|
|
|
#
|
36
|
|
|
|
|
|
|
############################################################################
|
37
|
|
|
|
|
|
|
|
38
|
|
|
|
|
|
|
# Types of documents handled
|
39
|
|
|
|
|
|
|
our ($UNKNOWN_FILE_TYPE,$DIR,$META,$HTML,$NEWS_XML,$AINODUMP,
|
40
|
|
|
|
|
|
|
$WIKIPEDIA_XML_DUMP)=(0..6);
|
41
|
|
|
|
|
|
|
my %RecognizedEntryTypeDescs=($UNKNOWN_FILE_TYPE=>"Guess the file type",
|
42
|
|
|
|
|
|
|
$DIR=>"Directory",
|
43
|
|
|
|
|
|
|
$META=>"Meta information",
|
44
|
|
|
|
|
|
|
$HTML=>"HTML",
|
45
|
|
|
|
|
|
|
$NEWS_XML=>
|
46
|
|
|
|
|
|
|
"XML information about a news article",
|
47
|
|
|
|
|
|
|
$AINODUMP=>"ainodump",
|
48
|
|
|
|
|
|
|
$WIKIPEDIA_XML_DUMP=>"Wikipedia XML dump");
|
49
|
|
|
|
|
|
|
|
50
|
|
|
|
|
|
|
############################################################################
|
51
|
|
|
|
|
|
|
#
|
52
|
|
|
|
|
|
|
# Error message stuff
|
53
|
|
|
|
|
|
|
#
|
54
|
|
|
|
|
|
|
############################################################################
|
55
|
|
|
|
|
|
|
|
56
|
|
|
|
|
|
|
my ($ERR_OK,
|
57
|
|
|
|
|
|
|
$ERR_CANONICAL,
|
58
|
|
|
|
|
|
|
$ERR_ASSEMBLER,
|
59
|
|
|
|
|
|
|
$ERR_CANDOC_CONV,
|
60
|
|
|
|
|
|
|
$ERR_META,
|
61
|
|
|
|
|
|
|
$ERR_LINKS,
|
62
|
|
|
|
|
|
|
$ERR_LINK_ADD,
|
63
|
|
|
|
|
|
|
$ERR_ASSEMBLE,
|
64
|
|
|
|
|
|
|
$ERR_NO_NEWS_XML_TEXT,
|
65
|
|
|
|
|
|
|
$ERR_XML_PARSER,
|
66
|
|
|
|
|
|
|
$ERR_XML_PARSE,
|
67
|
|
|
|
|
|
|
$ERR_NO_URL,
|
68
|
|
|
|
|
|
|
$ERR_ENCODING_WIZARD,
|
69
|
|
|
|
|
|
|
$ERR_UTF8_CONV,
|
70
|
|
|
|
|
|
|
$ERR_ENCODING_CONV,
|
71
|
|
|
|
|
|
|
$ERR_TYPE_SUFFIX,
|
72
|
|
|
|
|
|
|
$ERR_READ_HTML,
|
73
|
|
|
|
|
|
|
$ERR_READ_NEWS_XML,
|
74
|
|
|
|
|
|
|
$ERR_ALVIS_CONV,
|
75
|
|
|
|
|
|
|
$ERR_ALVIS_SUFFIX,
|
76
|
|
|
|
|
|
|
$ERR_NO_OUTPUT_ROOT_DIR,
|
77
|
|
|
|
|
|
|
$ERR_WRITING_OUTPUT,
|
78
|
|
|
|
|
|
|
$ERR_DIR_CONV,
|
79
|
|
|
|
|
|
|
$ERR_NO_HTML_F,
|
80
|
|
|
|
|
|
|
$ERR_META_F,
|
81
|
|
|
|
|
|
|
$ERR_HTML_F,
|
82
|
|
|
|
|
|
|
$ERR_NEWS_XML_F,
|
83
|
|
|
|
|
|
|
$ERR_DOC_ALVIS_CONV,
|
84
|
|
|
|
|
|
|
$ERR_NEWS_XML_PARSE,
|
85
|
|
|
|
|
|
|
$ERR_MULTIPLE_SUFFIX_MEANING,
|
86
|
|
|
|
|
|
|
$ERR_OUTPUT_ALVIS,
|
87
|
|
|
|
|
|
|
$ERR_OUTPUT_SET_OF_RECORDS,
|
88
|
|
|
|
|
|
|
$ERR_AINODUMP,
|
89
|
|
|
|
|
|
|
$ERR_OPEN_AINODUMP,
|
90
|
|
|
|
|
|
|
$ERR_AINODUMP_PROCESS,
|
91
|
|
|
|
|
|
|
$ERR_DOC_TYPE_WIZARD,
|
92
|
|
|
|
|
|
|
$ERR_TYPE_GUESS,
|
93
|
|
|
|
|
|
|
$ERR_UNK_FILE_TYPE,
|
94
|
|
|
|
|
|
|
$ERR_WIKIPEDIA,
|
95
|
|
|
|
|
|
|
$ERR_OPEN_WIKIPEDIA,
|
96
|
|
|
|
|
|
|
$ERR_WIKIPEDIA_CONV
|
97
|
|
|
|
|
|
|
)=(0..40);
|
98
|
|
|
|
|
|
|
|
99
|
|
|
|
|
|
|
my %ErrMsgs=($ERR_OK=>"",
|
100
|
|
|
|
|
|
|
$ERR_CANONICAL=>"Could not instantiate Alvis::Canonical.",
|
101
|
|
|
|
|
|
|
$ERR_ASSEMBLER=>"Could not instantiate Alvis::Document.",
|
102
|
|
|
|
|
|
|
$ERR_CANDOC_CONV=>"Conversion to canonicalDocument failed.",
|
103
|
|
|
|
|
|
|
$ERR_META=>"Could not instantiate Alvis::Document::Meta.",
|
104
|
|
|
|
|
|
|
$ERR_LINKS=>"Could not instantiate Alvis::Document::Links.",
|
105
|
|
|
|
|
|
|
$ERR_LINK_ADD=>"Adding a link failed.",
|
106
|
|
|
|
|
|
|
$ERR_ASSEMBLE=>"Assembling a document failed.",
|
107
|
|
|
|
|
|
|
$ERR_NO_NEWS_XML_TEXT=>"Unable to extract the content from News" .
|
108
|
|
|
|
|
|
|
" XML format.",
|
109
|
|
|
|
|
|
|
$ERR_XML_PARSER=>"Could not instantiate XML::LibXML.",
|
110
|
|
|
|
|
|
|
$ERR_XML_PARSE=>"Parsing the XML failed.",
|
111
|
|
|
|
|
|
|
$ERR_NO_URL=>"No URL.",
|
112
|
|
|
|
|
|
|
$ERR_ENCODING_WIZARD=>"Unable to instantiate " .
|
113
|
|
|
|
|
|
|
"Alvis::Document::Encoding.",
|
114
|
|
|
|
|
|
|
$ERR_UTF8_CONV=>"Trying to convert to UTF-8 failed.",
|
115
|
|
|
|
|
|
|
$ERR_ENCODING_CONV=>"Converting from the supposed source " .
|
116
|
|
|
|
|
|
|
"encoding to UTF-8 failed.",
|
117
|
|
|
|
|
|
|
$ERR_TYPE_SUFFIX=>"No suffix given for a type.",
|
118
|
|
|
|
|
|
|
$ERR_READ_HTML=>"Reading the HTML failed.",
|
119
|
|
|
|
|
|
|
$ERR_READ_NEWS_XML=>"Reading the news XML failed.",
|
120
|
|
|
|
|
|
|
$ERR_ALVIS_CONV=>"Conversion to Alvis format failed.",
|
121
|
|
|
|
|
|
|
$ERR_ALVIS_SUFFIX=>"No Alvis suffix defined.",
|
122
|
|
|
|
|
|
|
$ERR_NO_OUTPUT_ROOT_DIR=>"No output root directory.",
|
123
|
|
|
|
|
|
|
$ERR_WRITING_OUTPUT=>"Writing the output failed.",
|
124
|
|
|
|
|
|
|
$ERR_DIR_CONV=>"Converting a directory failed.",
|
125
|
|
|
|
|
|
|
$ERR_NO_HTML_F=>"No HTML file.",
|
126
|
|
|
|
|
|
|
$ERR_META_F=>"Opening the meta file failed.",
|
127
|
|
|
|
|
|
|
$ERR_HTML_F=>"Opening the HTML file failed.",
|
128
|
|
|
|
|
|
|
$ERR_NEWS_XML_F=>"Opening the news XML file failed.",
|
129
|
|
|
|
|
|
|
$ERR_DOC_ALVIS_CONV=>"Converting a document to Alvis format failed.",
|
130
|
|
|
|
|
|
|
$ERR_NEWS_XML_PARSE=>"Parsing the news XML failed.",
|
131
|
|
|
|
|
|
|
$ERR_MULTIPLE_SUFFIX_MEANING=>
|
132
|
|
|
|
|
|
|
"Multiple meanings for a single suffix.",
|
133
|
|
|
|
|
|
|
$ERR_OUTPUT_ALVIS=>"Outputting the Alvis records failed.",
|
134
|
|
|
|
|
|
|
$ERR_OUTPUT_SET_OF_RECORDS=>"Outputting a set of records to a " .
|
135
|
|
|
|
|
|
|
"file as a documentCollection failed.",
|
136
|
|
|
|
|
|
|
$ERR_AINODUMP=>"Instantiating Alvis::AinoDump failed.",
|
137
|
|
|
|
|
|
|
$ERR_OPEN_AINODUMP=>"Opening an ainodump file failed.",
|
138
|
|
|
|
|
|
|
$ERR_AINODUMP_PROCESS=>"Processing an ainodump file failed.",
|
139
|
|
|
|
|
|
|
$ERR_DOC_TYPE_WIZARD=>"Instantiating Alvis::Document::Type " .
|
140
|
|
|
|
|
|
|
"failed.",
|
141
|
|
|
|
|
|
|
$ERR_TYPE_GUESS=>"Guessing the document's type failed.",
|
142
|
|
|
|
|
|
|
$ERR_UNK_FILE_TYPE=>"Unrecognized file type.",
|
143
|
|
|
|
|
|
|
$ERR_WIKIPEDIA=>"Instantiating Alvis::Wikipedia::XMLDump failed.",
|
144
|
|
|
|
|
|
|
$ERR_OPEN_WIKIPEDIA=>"Opening the Wikipedia XML dump file failed.",
|
145
|
|
|
|
|
|
|
$ERR_WIKIPEDIA_CONV=>"Extracting the articles from the Wikipedia" .
|
146
|
|
|
|
|
|
|
" XML dump failed."
|
147
|
|
|
|
|
|
|
);
|
148
|
|
|
|
|
|
|
|
149
|
|
|
|
|
|
|
sub _set_err_state
|
150
|
|
|
|
|
|
|
{
|
151
|
|
|
|
|
|
|
my $self=shift;
|
152
|
|
|
|
|
|
|
my $errcode=shift;
|
153
|
|
|
|
|
|
|
my $errmsg=shift;
|
154
|
|
|
|
|
|
|
|
155
|
|
|
|
|
|
|
|
156
|
|
|
|
|
|
|
if (!defined($errcode))
|
157
|
|
|
|
|
|
|
{
|
158
|
|
|
|
|
|
|
confess("set_err_state() called with an undefined argument.");
|
159
|
|
|
|
|
|
|
}
|
160
|
|
|
|
|
|
|
|
161
|
|
|
|
|
|
|
if (exists($ErrMsgs{$errcode}))
|
162
|
|
|
|
|
|
|
{
|
163
|
|
|
|
|
|
|
if ($errcode==$ERR_OK)
|
164
|
|
|
|
|
|
|
{
|
165
|
|
|
|
|
|
|
$self->{errstr}="";
|
166
|
|
|
|
|
|
|
}
|
167
|
|
|
|
|
|
|
else
|
168
|
|
|
|
|
|
|
{
|
169
|
|
|
|
|
|
|
$self->{errstr}.=" " . $ErrMsgs{$errcode};
|
170
|
|
|
|
|
|
|
if (defined($errmsg))
|
171
|
|
|
|
|
|
|
{
|
172
|
|
|
|
|
|
|
$self->{errstr}.=" " . $errmsg;
|
173
|
|
|
|
|
|
|
}
|
174
|
|
|
|
|
|
|
|
175
|
|
|
|
|
|
|
}
|
176
|
|
|
|
|
|
|
}
|
177
|
|
|
|
|
|
|
else
|
178
|
|
|
|
|
|
|
{
|
179
|
|
|
|
|
|
|
confess("Internal error: set_err_state() called with an " .
|
180
|
|
|
|
|
|
|
"unrecognized argument ($errcode).")
|
181
|
|
|
|
|
|
|
}
|
182
|
|
|
|
|
|
|
}
|
183
|
|
|
|
|
|
|
|
184
|
|
|
|
|
|
|
sub clearerr
|
185
|
|
|
|
|
|
|
{
|
186
|
|
|
|
|
|
|
my $self=shift;
|
187
|
|
|
|
|
|
|
|
188
|
|
|
|
|
|
|
$self->{errstr}="";
|
189
|
|
|
|
|
|
|
}
|
190
|
|
|
|
|
|
|
|
191
|
|
|
|
|
|
|
sub errmsg
|
192
|
|
|
|
|
|
|
{
|
193
|
|
|
|
|
|
|
my $self=shift;
|
194
|
|
|
|
|
|
|
|
195
|
|
|
|
|
|
|
return $self->{errstr};
|
196
|
|
|
|
|
|
|
}
|
197
|
|
|
|
|
|
|
|
198
|
|
|
|
|
|
|
############################################################################
|
199
|
|
|
|
|
|
|
#
|
200
|
|
|
|
|
|
|
# Public methods
|
201
|
|
|
|
|
|
|
#
|
202
|
|
|
|
|
|
|
############################################################################
|
203
|
|
|
|
|
|
|
|
204
|
|
|
|
|
|
|
sub new
|
205
|
|
|
|
|
|
|
{
|
206
|
|
|
|
|
|
|
my $proto=shift;
|
207
|
|
|
|
|
|
|
|
208
|
|
|
|
|
|
|
my $class=ref($proto)||$proto;
|
209
|
|
|
|
|
|
|
my $parent=ref($proto)&&$proto;
|
210
|
|
|
|
|
|
|
my $self={};
|
211
|
|
|
|
|
|
|
bless($self,$class);
|
212
|
|
|
|
|
|
|
|
213
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OK);
|
214
|
|
|
|
|
|
|
|
215
|
|
|
|
|
|
|
$self->_init(@_);
|
216
|
|
|
|
|
|
|
|
217
|
|
|
|
|
|
|
if (defined($self->{urlBase}))
|
218
|
|
|
|
|
|
|
{
|
219
|
|
|
|
|
|
|
if ($self->{urlBase}!~/\/$/)
|
220
|
|
|
|
|
|
|
{
|
221
|
|
|
|
|
|
|
$self->{urlBase}.='/';
|
222
|
|
|
|
|
|
|
}
|
223
|
|
|
|
|
|
|
}
|
224
|
|
|
|
|
|
|
|
225
|
|
|
|
|
|
|
$self->{canonicalConverter}=Alvis::Canonical->new();
|
226
|
|
|
|
|
|
|
if (!defined($self->{canonicalConverter}))
|
227
|
|
|
|
|
|
|
{
|
228
|
|
|
|
|
|
|
$self->_set_err_state($ERR_CANONICAL);
|
229
|
|
|
|
|
|
|
return undef;
|
230
|
|
|
|
|
|
|
}
|
231
|
|
|
|
|
|
|
|
232
|
|
|
|
|
|
|
$self->{documentAssembler}=
|
233
|
|
|
|
|
|
|
Alvis::Document->new(includeOriginalDocument=>
|
234
|
|
|
|
|
|
|
$self->{includeOriginalDocument});
|
235
|
|
|
|
|
|
|
if (!defined($self->{documentAssembler}))
|
236
|
|
|
|
|
|
|
{
|
237
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ASSEMBLER);
|
238
|
|
|
|
|
|
|
return undef;
|
239
|
|
|
|
|
|
|
}
|
240
|
|
|
|
|
|
|
|
241
|
|
|
|
|
|
|
$self->{XMLParser}=XML::LibXML->new();
|
242
|
|
|
|
|
|
|
if (!defined($self->{XMLParser}))
|
243
|
|
|
|
|
|
|
{
|
244
|
|
|
|
|
|
|
$self->_set_err_state($ERR_XML_PARSER);
|
245
|
|
|
|
|
|
|
return undef;
|
246
|
|
|
|
|
|
|
}
|
247
|
|
|
|
|
|
|
|
248
|
|
|
|
|
|
|
$self->{encodingWizard}=
|
249
|
|
|
|
|
|
|
Alvis::Document::Encoding->new(defaultEncoding=>undef);
|
250
|
|
|
|
|
|
|
if (!defined($self->{encodingWizard}))
|
251
|
|
|
|
|
|
|
{
|
252
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ENCODING_WIZARD);
|
253
|
|
|
|
|
|
|
return undef;
|
254
|
|
|
|
|
|
|
}
|
255
|
|
|
|
|
|
|
|
256
|
|
|
|
|
|
|
$self->{wikipediaConverter}=
|
257
|
|
|
|
|
|
|
Alvis::Wikipedia::XMLDump->new(expandVariables=>1,
|
258
|
|
|
|
|
|
|
skipRedirects=>0,
|
259
|
|
|
|
|
|
|
dumpCategoryData=>1,
|
260
|
|
|
|
|
|
|
dumpTemplateData=>1);
|
261
|
|
|
|
|
|
|
if (!defined($self->{wikipediaConverter}))
|
262
|
|
|
|
|
|
|
{
|
263
|
|
|
|
|
|
|
$self->_set_err_state($ERR_WIKIPEDIA);
|
264
|
|
|
|
|
|
|
return undef;
|
265
|
|
|
|
|
|
|
}
|
266
|
|
|
|
|
|
|
|
267
|
|
|
|
|
|
|
$self->{docTypeWizard}=
|
268
|
|
|
|
|
|
|
Alvis::Document::Type->new(defaultType=>
|
269
|
|
|
|
|
|
|
$self->{defaultDocType},
|
270
|
|
|
|
|
|
|
defaultSubType=>
|
271
|
|
|
|
|
|
|
$self->{defaultDocSubType});
|
272
|
|
|
|
|
|
|
if (!defined($self->{docTypeWizard}))
|
273
|
|
|
|
|
|
|
{
|
274
|
|
|
|
|
|
|
$self->_set_err_state($ERR_DOC_TYPE_WIZARD);
|
275
|
|
|
|
|
|
|
return undef;
|
276
|
|
|
|
|
|
|
}
|
277
|
|
|
|
|
|
|
|
278
|
|
|
|
|
|
|
return $self;
|
279
|
|
|
|
|
|
|
}
|
280
|
|
|
|
|
|
|
|
281
|
|
|
|
|
|
|
sub _init
|
282
|
|
|
|
|
|
|
{
|
283
|
|
|
|
|
|
|
my $self=shift;
|
284
|
|
|
|
|
|
|
|
285
|
|
|
|
|
|
|
$self->{fileType}=undef;
|
286
|
|
|
|
|
|
|
$self->{sourceEncoding}=undef;
|
287
|
|
|
|
|
|
|
$self->{urlFromBasename}=0;
|
288
|
|
|
|
|
|
|
$self->{outputAtSameLocation}=0;
|
289
|
|
|
|
|
|
|
$self->{alvisSuffix}='alvis';
|
290
|
|
|
|
|
|
|
$self->{outputRootDir}='.';
|
291
|
|
|
|
|
|
|
$self->{outputNPerSubdir}=1000;
|
292
|
|
|
|
|
|
|
$self->{defaultDocType}='text';
|
293
|
|
|
|
|
|
|
$self->{defaultDocSubType}='html';
|
294
|
|
|
|
|
|
|
$self->{defaultEncoding}='iso-8859-1';
|
295
|
|
|
|
|
|
|
$self->{includeOriginalDocument}=1;
|
296
|
|
|
|
|
|
|
$self->{ainodumpWarnings}=1;
|
297
|
|
|
|
|
|
|
$self->{sourceEncodingFromMeta}=0;
|
298
|
|
|
|
|
|
|
|
299
|
|
|
|
|
|
|
if (defined(@_))
|
300
|
|
|
|
|
|
|
{
|
301
|
|
|
|
|
|
|
my %args=@_;
|
302
|
|
|
|
|
|
|
@$self{ keys %args }=values(%args);
|
303
|
|
|
|
|
|
|
}
|
304
|
|
|
|
|
|
|
|
305
|
|
|
|
|
|
|
}
|
306
|
|
|
|
|
|
|
|
307
|
|
|
|
|
|
|
#
|
308
|
|
|
|
|
|
|
# in UTF-8
|
309
|
|
|
|
|
|
|
#
|
310
|
|
|
|
|
|
|
sub HTML
|
311
|
|
|
|
|
|
|
{
|
312
|
|
|
|
|
|
|
my $self=shift;
|
313
|
|
|
|
|
|
|
my $html=shift;
|
314
|
|
|
|
|
|
|
my $meta_txt=shift;
|
315
|
|
|
|
|
|
|
my $opts=shift;
|
316
|
|
|
|
|
|
|
|
317
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OK);
|
318
|
|
|
|
|
|
|
|
319
|
|
|
|
|
|
|
my $meta=Alvis::Document::Meta->new(text=>$meta_txt);
|
320
|
|
|
|
|
|
|
if (!defined($meta))
|
321
|
|
|
|
|
|
|
{
|
322
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META,
|
323
|
|
|
|
|
|
|
"Meta text:\"$meta_txt\".");
|
324
|
|
|
|
|
|
|
return undef;
|
325
|
|
|
|
|
|
|
}
|
326
|
|
|
|
|
|
|
|
327
|
|
|
|
|
|
|
my $src_enc;
|
328
|
|
|
|
|
|
|
if ($opts->{sourceEncoding})
|
329
|
|
|
|
|
|
|
{
|
330
|
|
|
|
|
|
|
$src_enc=$opts->{sourceEncoding};
|
331
|
|
|
|
|
|
|
}
|
332
|
|
|
|
|
|
|
elsif (!exists($opts->{sourceEncoding}) && $self->{sourceEncoding})
|
333
|
|
|
|
|
|
|
{
|
334
|
|
|
|
|
|
|
$src_enc=$self->{sourceEncoding};
|
335
|
|
|
|
|
|
|
}
|
336
|
|
|
|
|
|
|
else
|
337
|
|
|
|
|
|
|
{
|
338
|
|
|
|
|
|
|
# warn "NO SOURCE ENCODING GIVEN IN OPTIONS TO HTML() OR IN new()";
|
339
|
|
|
|
|
|
|
}
|
340
|
|
|
|
|
|
|
|
341
|
|
|
|
|
|
|
if ($opts->{sourceEncodingFromMeta} || $self->{sourceEncodingFromMeta})
|
342
|
|
|
|
|
|
|
{
|
343
|
|
|
|
|
|
|
my $detected=$meta->get('detectedCharSet');
|
344
|
|
|
|
|
|
|
if ($detected)
|
345
|
|
|
|
|
|
|
{
|
346
|
|
|
|
|
|
|
$src_enc=$detected;
|
347
|
|
|
|
|
|
|
}
|
348
|
|
|
|
|
|
|
}
|
349
|
|
|
|
|
|
|
|
350
|
|
|
|
|
|
|
my ($can_doc,$header)=
|
351
|
|
|
|
|
|
|
$self->{canonicalConverter}->HTML($html,
|
352
|
|
|
|
|
|
|
{title=>1,
|
353
|
|
|
|
|
|
|
baseURL=>1,
|
354
|
|
|
|
|
|
|
sourceEncoding=>$src_enc});
|
355
|
|
|
|
|
|
|
if (!defined($can_doc))
|
356
|
|
|
|
|
|
|
{
|
357
|
|
|
|
|
|
|
$self->_set_err_state($ERR_CANDOC_CONV,
|
358
|
|
|
|
|
|
|
$self->{canonicalConverter}->errmsg());
|
359
|
|
|
|
|
|
|
return undef;
|
360
|
|
|
|
|
|
|
}
|
361
|
|
|
|
|
|
|
|
362
|
|
|
|
|
|
|
if (!defined($meta->get('title')))
|
363
|
|
|
|
|
|
|
{
|
364
|
|
|
|
|
|
|
$meta->set('title',$header->{title});
|
365
|
|
|
|
|
|
|
}
|
366
|
|
|
|
|
|
|
if (!defined($meta->get('url')))
|
367
|
|
|
|
|
|
|
{
|
368
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NO_URL);
|
369
|
|
|
|
|
|
|
return undef;
|
370
|
|
|
|
|
|
|
}
|
371
|
|
|
|
|
|
|
else
|
372
|
|
|
|
|
|
|
{
|
373
|
|
|
|
|
|
|
if (!defined($meta->get('baseURL')))
|
374
|
|
|
|
|
|
|
{
|
375
|
|
|
|
|
|
|
if (defined($header->{baseURL}))
|
376
|
|
|
|
|
|
|
{
|
377
|
|
|
|
|
|
|
$meta->set('baseURL',$header->{baseURL});
|
378
|
|
|
|
|
|
|
}
|
379
|
|
|
|
|
|
|
else
|
380
|
|
|
|
|
|
|
{
|
381
|
|
|
|
|
|
|
my $base_URL=$meta->get('url');
|
382
|
|
|
|
|
|
|
$base_URL=~s/\/[^\/]+?$/\//isgo;
|
383
|
|
|
|
|
|
|
$meta->set('baseURL',$base_URL);
|
384
|
|
|
|
|
|
|
}
|
385
|
|
|
|
|
|
|
}
|
386
|
|
|
|
|
|
|
}
|
387
|
|
|
|
|
|
|
|
388
|
|
|
|
|
|
|
my $links=Alvis::Document::Links->new();
|
389
|
|
|
|
|
|
|
if (!defined($links))
|
390
|
|
|
|
|
|
|
{
|
391
|
|
|
|
|
|
|
$self->_set_err_state($ERR_LINKS);
|
392
|
|
|
|
|
|
|
return undef;
|
393
|
|
|
|
|
|
|
}
|
394
|
|
|
|
|
|
|
for my $link (@{$header->{links}})
|
395
|
|
|
|
|
|
|
{
|
396
|
|
|
|
|
|
|
my ($url,$text,$type);
|
397
|
|
|
|
|
|
|
if (exists($link->{url}))
|
398
|
|
|
|
|
|
|
{
|
399
|
|
|
|
|
|
|
$url=$link->{url};
|
400
|
|
|
|
|
|
|
}
|
401
|
|
|
|
|
|
|
if (exists($link->{text}))
|
402
|
|
|
|
|
|
|
{
|
403
|
|
|
|
|
|
|
$text=$link->{text};
|
404
|
|
|
|
|
|
|
}
|
405
|
|
|
|
|
|
|
if (exists($link->{type}))
|
406
|
|
|
|
|
|
|
{
|
407
|
|
|
|
|
|
|
if ($link->{type}=~/^\s*a\s*$/isgo)
|
408
|
|
|
|
|
|
|
{
|
409
|
|
|
|
|
|
|
$type='a';
|
410
|
|
|
|
|
|
|
}
|
411
|
|
|
|
|
|
|
elsif ($link->{type}=~/^\s*i?frame\s*$/isgo)
|
412
|
|
|
|
|
|
|
{
|
413
|
|
|
|
|
|
|
$type='frame';
|
414
|
|
|
|
|
|
|
}
|
415
|
|
|
|
|
|
|
elsif ($link->{type}=~/^\s*img\s*$/isgo)
|
416
|
|
|
|
|
|
|
{
|
417
|
|
|
|
|
|
|
$type='img';
|
418
|
|
|
|
|
|
|
}
|
419
|
|
|
|
|
|
|
}
|
420
|
|
|
|
|
|
|
|
421
|
|
|
|
|
|
|
if (!$links->add($url,$text,$type))
|
422
|
|
|
|
|
|
|
{
|
423
|
|
|
|
|
|
|
$self->_set_err_state($ERR_LINK_ADD,
|
424
|
|
|
|
|
|
|
$links->errmsg());
|
425
|
|
|
|
|
|
|
return undef;
|
426
|
|
|
|
|
|
|
}
|
427
|
|
|
|
|
|
|
}
|
428
|
|
|
|
|
|
|
|
429
|
|
|
|
|
|
|
my $alvisXML=
|
430
|
|
|
|
|
|
|
$self->{documentAssembler}->assemble({canDoc=>$can_doc,
|
431
|
|
|
|
|
|
|
links=>$links,
|
432
|
|
|
|
|
|
|
meta=>$meta,
|
433
|
|
|
|
|
|
|
origText=>$html});
|
434
|
|
|
|
|
|
|
if (!defined($alvisXML))
|
435
|
|
|
|
|
|
|
{
|
436
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ASSEMBLE,
|
437
|
|
|
|
|
|
|
$self->{documentAssembler}->errmsg());
|
438
|
|
|
|
|
|
|
return undef;
|
439
|
|
|
|
|
|
|
}
|
440
|
|
|
|
|
|
|
|
441
|
|
|
|
|
|
|
return $alvisXML;
|
442
|
|
|
|
|
|
|
}
|
443
|
|
|
|
|
|
|
|
444
|
|
|
|
|
|
|
sub newsXML
|
445
|
|
|
|
|
|
|
{
|
446
|
|
|
|
|
|
|
my $self=shift;
|
447
|
|
|
|
|
|
|
my $newsXML=shift;
|
448
|
|
|
|
|
|
|
my $meta_txt=shift;
|
449
|
|
|
|
|
|
|
my $orig_txt=shift;
|
450
|
|
|
|
|
|
|
|
451
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OK);
|
452
|
|
|
|
|
|
|
|
453
|
|
|
|
|
|
|
my $meta=Alvis::Document::Meta->new(text=>$meta_txt);
|
454
|
|
|
|
|
|
|
if (!defined($meta))
|
455
|
|
|
|
|
|
|
{
|
456
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META,
|
457
|
|
|
|
|
|
|
"Meta text:\"$meta_txt\".");
|
458
|
|
|
|
|
|
|
return undef;
|
459
|
|
|
|
|
|
|
}
|
460
|
|
|
|
|
|
|
|
461
|
|
|
|
|
|
|
my @alvisXMLs=();
|
462
|
|
|
|
|
|
|
|
463
|
|
|
|
|
|
|
my $articles=$self->_parse_newsXML($newsXML);
|
464
|
|
|
|
|
|
|
if (!defined($articles))
|
465
|
|
|
|
|
|
|
{
|
466
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NEWS_XML_PARSE);
|
467
|
|
|
|
|
|
|
return undef;
|
468
|
|
|
|
|
|
|
}
|
469
|
|
|
|
|
|
|
for my $article (@$articles)
|
470
|
|
|
|
|
|
|
{
|
471
|
|
|
|
|
|
|
my ($text,$iso_date,$title,$links)=@$article;
|
472
|
|
|
|
|
|
|
if (!defined($text))
|
473
|
|
|
|
|
|
|
{
|
474
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NO_NEWS_XML_TEXT,
|
475
|
|
|
|
|
|
|
"News XML text:\"$newsXML\".");
|
476
|
|
|
|
|
|
|
# OK, ignore
|
477
|
|
|
|
|
|
|
next;
|
478
|
|
|
|
|
|
|
# return undef;
|
479
|
|
|
|
|
|
|
}
|
480
|
|
|
|
|
|
|
$text='' . $text . '';
|
481
|
|
|
|
|
|
|
|
482
|
|
|
|
|
|
|
# Check that the ISO date actually is in ISO format...
|
483
|
|
|
|
|
|
|
if (defined($iso_date))
|
484
|
|
|
|
|
|
|
{
|
485
|
|
|
|
|
|
|
$meta->set('dc:date',$iso_date);
|
486
|
|
|
|
|
|
|
}
|
487
|
|
|
|
|
|
|
|
488
|
|
|
|
|
|
|
my ($can_doc,$header)=
|
489
|
|
|
|
|
|
|
$self->{canonicalConverter}->HTML($text,
|
490
|
|
|
|
|
|
|
{sourceEncoding=>'utf8'});
|
491
|
|
|
|
|
|
|
if (!defined($can_doc))
|
492
|
|
|
|
|
|
|
{
|
493
|
|
|
|
|
|
|
$self->_set_err_state($ERR_CANDOC_CONV,
|
494
|
|
|
|
|
|
|
$self->{canonicalConverter}->errmsg());
|
495
|
|
|
|
|
|
|
return undef;
|
496
|
|
|
|
|
|
|
}
|
497
|
|
|
|
|
|
|
|
498
|
|
|
|
|
|
|
if (defined($title))
|
499
|
|
|
|
|
|
|
{
|
500
|
|
|
|
|
|
|
$meta->set('title',$title);
|
501
|
|
|
|
|
|
|
}
|
502
|
|
|
|
|
|
|
if (!defined($meta->get('url')))
|
503
|
|
|
|
|
|
|
{
|
504
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NO_URL);
|
505
|
|
|
|
|
|
|
return undef;
|
506
|
|
|
|
|
|
|
}
|
507
|
|
|
|
|
|
|
else
|
508
|
|
|
|
|
|
|
{
|
509
|
|
|
|
|
|
|
if (!defined($meta->get('baseURL')))
|
510
|
|
|
|
|
|
|
{
|
511
|
|
|
|
|
|
|
my $base_URL=$meta->get('url');
|
512
|
|
|
|
|
|
|
$base_URL=~s/\/[^\/]+?$/\//isgo;
|
513
|
|
|
|
|
|
|
$meta->set('baseURL',$base_URL);
|
514
|
|
|
|
|
|
|
}
|
515
|
|
|
|
|
|
|
}
|
516
|
|
|
|
|
|
|
|
517
|
|
|
|
|
|
|
my $alvisXML=
|
518
|
|
|
|
|
|
|
$self->{documentAssembler}->assemble({canDoc=>$can_doc,
|
519
|
|
|
|
|
|
|
meta=>$meta,
|
520
|
|
|
|
|
|
|
links=>$links,
|
521
|
|
|
|
|
|
|
origText=>$orig_txt});
|
522
|
|
|
|
|
|
|
if (!defined($alvisXML))
|
523
|
|
|
|
|
|
|
{
|
524
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ASSEMBLE,
|
525
|
|
|
|
|
|
|
$self->{documentAssembler}->errmsg());
|
526
|
|
|
|
|
|
|
return undef;
|
527
|
|
|
|
|
|
|
}
|
528
|
|
|
|
|
|
|
push(@alvisXMLs,$alvisXML);
|
529
|
|
|
|
|
|
|
}
|
530
|
|
|
|
|
|
|
|
531
|
|
|
|
|
|
|
return \@alvisXMLs;
|
532
|
|
|
|
|
|
|
}
|
533
|
|
|
|
|
|
|
|
534
|
|
|
|
|
|
|
sub ainodump
|
535
|
|
|
|
|
|
|
{
|
536
|
|
|
|
|
|
|
my $self=shift;
|
537
|
|
|
|
|
|
|
my $f=shift;
|
538
|
|
|
|
|
|
|
|
539
|
|
|
|
|
|
|
# No meta needed -- one per record in the dump
|
540
|
|
|
|
|
|
|
#
|
541
|
|
|
|
|
|
|
if (!defined(open(AINO,"<:raw",$f)))
|
542
|
|
|
|
|
|
|
{
|
543
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OPEN_AINODUMP,
|
544
|
|
|
|
|
|
|
"File: \"$f\"");
|
545
|
|
|
|
|
|
|
return 0;
|
546
|
|
|
|
|
|
|
}
|
547
|
|
|
|
|
|
|
if (!$self->{ainodumpConverter}
|
548
|
|
|
|
|
|
|
->process_dump(*AINO,
|
549
|
|
|
|
|
|
|
[\&_process_ainodump_doc,$self]))
|
550
|
|
|
|
|
|
|
{
|
551
|
|
|
|
|
|
|
$self->_set_err_state($ERR_AINODUMP_PROCESS,
|
552
|
|
|
|
|
|
|
"File: \"$f\"");
|
553
|
|
|
|
|
|
|
return 0;
|
554
|
|
|
|
|
|
|
}
|
555
|
|
|
|
|
|
|
close(AINO);
|
556
|
|
|
|
|
|
|
|
557
|
|
|
|
|
|
|
return 1;
|
558
|
|
|
|
|
|
|
}
|
559
|
|
|
|
|
|
|
|
560
|
|
|
|
|
|
|
#
|
561
|
|
|
|
|
|
|
# output_cb: [\&_output_wikipedia_article,$arg1,$arg2,...]
|
562
|
|
|
|
|
|
|
# will be called like this:
|
563
|
|
|
|
|
|
|
# _output_wikipedia_article($arg1,$arg2,...,
|
564
|
|
|
|
|
|
|
# $title,$output_format,
|
565
|
|
|
|
|
|
|
# $record_txt,$is_redir)
|
566
|
|
|
|
|
|
|
#
|
567
|
|
|
|
|
|
|
# where $output_format is a global defined in Alvis::Wikipedia::XMLDump
|
568
|
|
|
|
|
|
|
# as $OUTPUT_*
|
569
|
|
|
|
|
|
|
#
|
570
|
|
|
|
|
|
|
#
|
571
|
|
|
|
|
|
|
# progress_cb: [\&_wikipedia_progress,$arg1,$arg2,...] OPTIONAL
|
572
|
|
|
|
|
|
|
# will be called like this:
|
573
|
|
|
|
|
|
|
# _wikipedia_progress($arg1,$arg2,...,
|
574
|
|
|
|
|
|
|
# $prog_txt,$N,$n,$mess)
|
575
|
|
|
|
|
|
|
#
|
576
|
|
|
|
|
|
|
# where $N is the total number of records processed and $n the number of hits
|
577
|
|
|
|
|
|
|
#
|
578
|
|
|
|
|
|
|
# opts: a hash of options with these possible fields:
|
579
|
|
|
|
|
|
|
#
|
580
|
|
|
|
|
|
|
# namespaces ref to a list of namespace identifiers whose
|
581
|
|
|
|
|
|
|
# records to extract
|
582
|
|
|
|
|
|
|
# expandTemplates flag for true template expansion
|
583
|
|
|
|
|
|
|
# templateDumpF template dump file
|
584
|
|
|
|
|
|
|
# outputFormat format for result records
|
585
|
|
|
|
|
|
|
# ($Alvis::Wikipedia::XMLDump::OUTPUT_*)
|
586
|
|
|
|
|
|
|
# categoryWord category namespace identifier (changes with
|
587
|
|
|
|
|
|
|
# language)
|
588
|
|
|
|
|
|
|
# templateWord template namespace identifier (changes with
|
589
|
|
|
|
|
|
|
# language)
|
590
|
|
|
|
|
|
|
# rootCategory root category identifier (changes with
|
591
|
|
|
|
|
|
|
# language)
|
592
|
|
|
|
|
|
|
# date the date of the dump
|
593
|
|
|
|
|
|
|
# dumpCatGraph flag for dumping the category graph
|
594
|
|
|
|
|
|
|
# catGraphDumpF category graph dump file
|
595
|
|
|
|
|
|
|
#
|
596
|
|
|
|
|
|
|
sub wikipedia
|
597
|
|
|
|
|
|
|
{
|
598
|
|
|
|
|
|
|
my $self=shift;
|
599
|
|
|
|
|
|
|
my $f=shift;
|
600
|
|
|
|
|
|
|
my $output_cb=shift;
|
601
|
|
|
|
|
|
|
my $opts=shift;
|
602
|
|
|
|
|
|
|
my $progress_cb=shift;
|
603
|
|
|
|
|
|
|
|
604
|
|
|
|
|
|
|
if (!defined(open(WIKIPEDIA,"<:utf8",$f)))
|
605
|
|
|
|
|
|
|
{
|
606
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OPEN_WIKIPEDIA,
|
607
|
|
|
|
|
|
|
"File: \"$f\"");
|
608
|
|
|
|
|
|
|
return 0;
|
609
|
|
|
|
|
|
|
}
|
610
|
|
|
|
|
|
|
if (!$self->{wikipediaConverter}->extract_records(\*WIKIPEDIA,
|
611
|
|
|
|
|
|
|
$output_cb,
|
612
|
|
|
|
|
|
|
$opts,
|
613
|
|
|
|
|
|
|
$progress_cb))
|
614
|
|
|
|
|
|
|
{
|
615
|
|
|
|
|
|
|
$self->_set_err_state($ERR_WIKIPEDIA_CONV,
|
616
|
|
|
|
|
|
|
"File: \"$f\"");
|
617
|
|
|
|
|
|
|
return 0;
|
618
|
|
|
|
|
|
|
}
|
619
|
|
|
|
|
|
|
|
620
|
|
|
|
|
|
|
close(WIKIPEDIA);
|
621
|
|
|
|
|
|
|
|
622
|
|
|
|
|
|
|
return 1;
|
623
|
|
|
|
|
|
|
}
|
624
|
|
|
|
|
|
|
|
625
|
|
|
|
|
|
|
sub set
|
626
|
|
|
|
|
|
|
{
|
627
|
|
|
|
|
|
|
my $self=shift;
|
628
|
|
|
|
|
|
|
my $param=shift;
|
629
|
|
|
|
|
|
|
my $value=shift;
|
630
|
|
|
|
|
|
|
|
631
|
|
|
|
|
|
|
$self->{$param}=$value;
|
632
|
|
|
|
|
|
|
}
|
633
|
|
|
|
|
|
|
|
634
|
|
|
|
|
|
|
sub read_HTML
|
635
|
|
|
|
|
|
|
{
|
636
|
|
|
|
|
|
|
my $self=shift;
|
637
|
|
|
|
|
|
|
my $f=shift;
|
638
|
|
|
|
|
|
|
my $meta_txt=shift;
|
639
|
|
|
|
|
|
|
|
640
|
|
|
|
|
|
|
my $html_txt="";
|
641
|
|
|
|
|
|
|
|
642
|
|
|
|
|
|
|
# Stupid duplicating of "how the f**k do you read UTF8 in Perl?" fix
|
643
|
|
|
|
|
|
|
my $meta=Alvis::Document::Meta->new(text=>$meta_txt);
|
644
|
|
|
|
|
|
|
if (!defined($meta))
|
645
|
|
|
|
|
|
|
{
|
646
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META,
|
647
|
|
|
|
|
|
|
"Meta text:\"$meta_txt\".");
|
648
|
|
|
|
|
|
|
return undef;
|
649
|
|
|
|
|
|
|
}
|
650
|
|
|
|
|
|
|
|
651
|
|
|
|
|
|
|
my $src_enc;
|
652
|
|
|
|
|
|
|
if ($self->{sourceEncoding})
|
653
|
|
|
|
|
|
|
{
|
654
|
|
|
|
|
|
|
$src_enc=$self->{sourceEncoding};
|
655
|
|
|
|
|
|
|
}
|
656
|
|
|
|
|
|
|
if ($self->{sourceEncodingFromMeta})
|
657
|
|
|
|
|
|
|
{
|
658
|
|
|
|
|
|
|
my $detected=$meta->get('detectedCharSet');
|
659
|
|
|
|
|
|
|
if ($detected)
|
660
|
|
|
|
|
|
|
{
|
661
|
|
|
|
|
|
|
$src_enc=$detected;
|
662
|
|
|
|
|
|
|
}
|
663
|
|
|
|
|
|
|
}
|
664
|
|
|
|
|
|
|
|
665
|
|
|
|
|
|
|
if (defined($src_enc) && $src_enc=~/^\s*utf\s*\-?\s*8\s*$/i)
|
666
|
|
|
|
|
|
|
{
|
667
|
|
|
|
|
|
|
if (!defined(open(H,"<:utf8",$f)))
|
668
|
|
|
|
|
|
|
{
|
669
|
|
|
|
|
|
|
$self->_set_err_state($ERR_HTML_F,
|
670
|
|
|
|
|
|
|
"File: \"$f\".");
|
671
|
|
|
|
|
|
|
return undef;
|
672
|
|
|
|
|
|
|
}
|
673
|
|
|
|
|
|
|
while (my $l=)
|
674
|
|
|
|
|
|
|
{
|
675
|
|
|
|
|
|
|
$html_txt.=$l;
|
676
|
|
|
|
|
|
|
}
|
677
|
|
|
|
|
|
|
close(H);
|
678
|
|
|
|
|
|
|
}
|
679
|
|
|
|
|
|
|
else
|
680
|
|
|
|
|
|
|
{
|
681
|
|
|
|
|
|
|
if (!defined(open(H,"<$f")))
|
682
|
|
|
|
|
|
|
{
|
683
|
|
|
|
|
|
|
$self->_set_err_state($ERR_HTML_F,
|
684
|
|
|
|
|
|
|
"File: \"$f\".");
|
685
|
|
|
|
|
|
|
return undef;
|
686
|
|
|
|
|
|
|
}
|
687
|
|
|
|
|
|
|
while (my $l=)
|
688
|
|
|
|
|
|
|
{
|
689
|
|
|
|
|
|
|
$html_txt.=$l;
|
690
|
|
|
|
|
|
|
}
|
691
|
|
|
|
|
|
|
close(H);
|
692
|
|
|
|
|
|
|
}
|
693
|
|
|
|
|
|
|
|
694
|
|
|
|
|
|
|
return $html_txt;
|
695
|
|
|
|
|
|
|
}
|
696
|
|
|
|
|
|
|
|
697
|
|
|
|
|
|
|
sub read_meta
|
698
|
|
|
|
|
|
|
{
|
699
|
|
|
|
|
|
|
my $self=shift;
|
700
|
|
|
|
|
|
|
my $f=shift;
|
701
|
|
|
|
|
|
|
|
702
|
|
|
|
|
|
|
my $meta_txt="";
|
703
|
|
|
|
|
|
|
|
704
|
|
|
|
|
|
|
if (defined($self->{metaEncoding}))
|
705
|
|
|
|
|
|
|
{
|
706
|
|
|
|
|
|
|
if ($self->{metaEncoding}=~/^\s*utf\s*\-?\s*8\s*$/i)
|
707
|
|
|
|
|
|
|
{
|
708
|
|
|
|
|
|
|
if (!defined(open(M,"<:utf8",$f)))
|
709
|
|
|
|
|
|
|
{
|
710
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META_F,
|
711
|
|
|
|
|
|
|
"File: \"$f\".");
|
712
|
|
|
|
|
|
|
return undef;
|
713
|
|
|
|
|
|
|
}
|
714
|
|
|
|
|
|
|
while (my $l=)
|
715
|
|
|
|
|
|
|
{
|
716
|
|
|
|
|
|
|
$meta_txt.=$l;
|
717
|
|
|
|
|
|
|
}
|
718
|
|
|
|
|
|
|
close(M);
|
719
|
|
|
|
|
|
|
}
|
720
|
|
|
|
|
|
|
else # non-UTF8
|
721
|
|
|
|
|
|
|
{
|
722
|
|
|
|
|
|
|
if (!defined(open(M,"<$f")))
|
723
|
|
|
|
|
|
|
{
|
724
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META_F,
|
725
|
|
|
|
|
|
|
"File: \"$f\".");
|
726
|
|
|
|
|
|
|
return undef;
|
727
|
|
|
|
|
|
|
}
|
728
|
|
|
|
|
|
|
while (my $l=)
|
729
|
|
|
|
|
|
|
{
|
730
|
|
|
|
|
|
|
$meta_txt.=$l;
|
731
|
|
|
|
|
|
|
}
|
732
|
|
|
|
|
|
|
close(M);
|
733
|
|
|
|
|
|
|
|
734
|
|
|
|
|
|
|
eval
|
735
|
|
|
|
|
|
|
{
|
736
|
|
|
|
|
|
|
Encode::from_to($meta_txt,
|
737
|
|
|
|
|
|
|
$self->{metaEncoding},'utf-8',Encode::FB_WARN);
|
738
|
|
|
|
|
|
|
};
|
739
|
|
|
|
|
|
|
if ($@)
|
740
|
|
|
|
|
|
|
{
|
741
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ENCODING_CONV,
|
742
|
|
|
|
|
|
|
"$@. Supposed source encoding of \"$f\":" .
|
743
|
|
|
|
|
|
|
"\"$self->{metaEncoding}\".");
|
744
|
|
|
|
|
|
|
return undef;
|
745
|
|
|
|
|
|
|
}
|
746
|
|
|
|
|
|
|
}
|
747
|
|
|
|
|
|
|
}
|
748
|
|
|
|
|
|
|
else # encoding unknown
|
749
|
|
|
|
|
|
|
{
|
750
|
|
|
|
|
|
|
if (!defined(open(M,"<$f")))
|
751
|
|
|
|
|
|
|
{
|
752
|
|
|
|
|
|
|
$self->_set_err_state($ERR_META_F,
|
753
|
|
|
|
|
|
|
"File: \"$f\".");
|
754
|
|
|
|
|
|
|
return undef;
|
755
|
|
|
|
|
|
|
}
|
756
|
|
|
|
|
|
|
my $meta_txt="";
|
757
|
|
|
|
|
|
|
while (my $l=)
|
758
|
|
|
|
|
|
|
{
|
759
|
|
|
|
|
|
|
$meta_txt.=$l;
|
760
|
|
|
|
|
|
|
}
|
761
|
|
|
|
|
|
|
close(M);
|
762
|
|
|
|
|
|
|
|
763
|
|
|
|
|
|
|
$meta_txt=$self->{encodingWizard}->try_to_convert_to_utf8($meta_txt,
|
764
|
|
|
|
|
|
|
'text',
|
765
|
|
|
|
|
|
|
'plain');
|
766
|
|
|
|
|
|
|
if (!defined($meta_txt))
|
767
|
|
|
|
|
|
|
{
|
768
|
|
|
|
|
|
|
$self->_set_err_state($ERR_UTF8_CONV,
|
769
|
|
|
|
|
|
|
$self->{encodingWizard}->errmsg());
|
770
|
|
|
|
|
|
|
return undef;
|
771
|
|
|
|
|
|
|
}
|
772
|
|
|
|
|
|
|
}
|
773
|
|
|
|
|
|
|
|
774
|
|
|
|
|
|
|
return $meta_txt;
|
775
|
|
|
|
|
|
|
}
|
776
|
|
|
|
|
|
|
|
777
|
|
|
|
|
|
|
sub read_news_XML
|
778
|
|
|
|
|
|
|
{
|
779
|
|
|
|
|
|
|
my $self=shift;
|
780
|
|
|
|
|
|
|
my $f=shift;
|
781
|
|
|
|
|
|
|
|
782
|
|
|
|
|
|
|
if (!defined(open(X,"<:utf8",$f)))
|
783
|
|
|
|
|
|
|
{
|
784
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NEWS_XML_F,
|
785
|
|
|
|
|
|
|
"File: \"$f\".");
|
786
|
|
|
|
|
|
|
return undef;
|
787
|
|
|
|
|
|
|
}
|
788
|
|
|
|
|
|
|
my $txt="";
|
789
|
|
|
|
|
|
|
while (my $l=)
|
790
|
|
|
|
|
|
|
{
|
791
|
|
|
|
|
|
|
$txt.=$l;
|
792
|
|
|
|
|
|
|
}
|
793
|
|
|
|
|
|
|
close(X);
|
794
|
|
|
|
|
|
|
|
795
|
|
|
|
|
|
|
return $txt;
|
796
|
|
|
|
|
|
|
}
|
797
|
|
|
|
|
|
|
|
798
|
|
|
|
|
|
|
sub init_output
|
799
|
|
|
|
|
|
|
{
|
800
|
|
|
|
|
|
|
my $self=shift;
|
801
|
|
|
|
|
|
|
|
802
|
|
|
|
|
|
|
$self->{outputN}=0;
|
803
|
|
|
|
|
|
|
}
|
804
|
|
|
|
|
|
|
|
805
|
|
|
|
|
|
|
sub output_Alvis
|
806
|
|
|
|
|
|
|
{
|
807
|
|
|
|
|
|
|
my $self=shift;
|
808
|
|
|
|
|
|
|
my $alvis_records=shift;
|
809
|
|
|
|
|
|
|
my $base_name=shift;
|
810
|
|
|
|
|
|
|
|
811
|
|
|
|
|
|
|
$self->{recordN}=0;
|
812
|
|
|
|
|
|
|
for my $alvis_record (@$alvis_records)
|
813
|
|
|
|
|
|
|
{
|
814
|
|
|
|
|
|
|
if (!defined($alvis_record))
|
815
|
|
|
|
|
|
|
{
|
816
|
|
|
|
|
|
|
$self->_set_err_state($ERR_DOC_ALVIS_CONV,
|
817
|
|
|
|
|
|
|
"Base name:\"$base_name\"," .
|
818
|
|
|
|
|
|
|
"# of record: $self->{recordN}");
|
819
|
|
|
|
|
|
|
return 0;
|
820
|
|
|
|
|
|
|
}
|
821
|
|
|
|
|
|
|
|
822
|
|
|
|
|
|
|
my $out_f;
|
823
|
|
|
|
|
|
|
if (!defined($self->{alvisSuffix}))
|
824
|
|
|
|
|
|
|
{
|
825
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ALVIS_SUFFIX);
|
826
|
|
|
|
|
|
|
return 0;
|
827
|
|
|
|
|
|
|
}
|
828
|
|
|
|
|
|
|
if ($self->{outputAtSameLocation})
|
829
|
|
|
|
|
|
|
{
|
830
|
|
|
|
|
|
|
$out_f=$base_name . "." . $self->{articleN} . '.' .
|
831
|
|
|
|
|
|
|
$self->{alvisSuffix};
|
832
|
|
|
|
|
|
|
$self->{articleN}++;
|
833
|
|
|
|
|
|
|
if (!$self->_output_set_of_records($alvis_record,$out_f))
|
834
|
|
|
|
|
|
|
{
|
835
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OUTPUT_SET_OF_RECORDS);
|
836
|
|
|
|
|
|
|
return 0;
|
837
|
|
|
|
|
|
|
}
|
838
|
|
|
|
|
|
|
$self->{outputN}++;
|
839
|
|
|
|
|
|
|
print "$self->{outputN}\r";
|
840
|
|
|
|
|
|
|
}
|
841
|
|
|
|
|
|
|
else
|
842
|
|
|
|
|
|
|
{
|
843
|
|
|
|
|
|
|
if (!defined($self->{outputRootDir}))
|
844
|
|
|
|
|
|
|
{
|
845
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NO_OUTPUT_ROOT_DIR);
|
846
|
|
|
|
|
|
|
return 0;
|
847
|
|
|
|
|
|
|
}
|
848
|
|
|
|
|
|
|
my $dir=$self->{outputRootDir} . '/' .
|
849
|
|
|
|
|
|
|
int($self->{outputN} / $self->{outputNPerSubdir});
|
850
|
|
|
|
|
|
|
if ($self->{outputN} % $self->{outputNPerSubdir}==0)
|
851
|
|
|
|
|
|
|
{
|
852
|
|
|
|
|
|
|
mkdir($dir);
|
853
|
|
|
|
|
|
|
}
|
854
|
|
|
|
|
|
|
$out_f=$dir . '/' . $self->{outputN} . '.' .
|
855
|
|
|
|
|
|
|
$self->{alvisSuffix};
|
856
|
|
|
|
|
|
|
|
857
|
|
|
|
|
|
|
if (!$self->_output_set_of_records($alvis_record,$out_f))
|
858
|
|
|
|
|
|
|
{
|
859
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OUTPUT_SET_OF_RECORDS);
|
860
|
|
|
|
|
|
|
return 0;
|
861
|
|
|
|
|
|
|
}
|
862
|
|
|
|
|
|
|
|
863
|
|
|
|
|
|
|
$self->{outputN}++;
|
864
|
|
|
|
|
|
|
print "$self->{outputN}\r";
|
865
|
|
|
|
|
|
|
}
|
866
|
|
|
|
|
|
|
}
|
867
|
|
|
|
|
|
|
|
868
|
|
|
|
|
|
|
return 1;
|
869
|
|
|
|
|
|
|
}
|
870
|
|
|
|
|
|
|
|
871
|
|
|
|
|
|
|
############################################################################
|
872
|
|
|
|
|
|
|
#
|
873
|
|
|
|
|
|
|
# Private methods
|
874
|
|
|
|
|
|
|
#
|
875
|
|
|
|
|
|
|
############################################################################
|
876
|
|
|
|
|
|
|
|
877
|
|
|
|
|
|
|
sub _process_ainodump_doc
|
878
|
|
|
|
|
|
|
{
|
879
|
|
|
|
|
|
|
my $self=shift;
|
880
|
|
|
|
|
|
|
my $text=shift;
|
881
|
|
|
|
|
|
|
my $header=shift;
|
882
|
|
|
|
|
|
|
|
883
|
|
|
|
|
|
|
# print Dumper($header);
|
884
|
|
|
|
|
|
|
# print "\n";
|
885
|
|
|
|
|
|
|
|
886
|
|
|
|
|
|
|
my ($type,$sub_type)=$self->{docTypeWizard}->guess($text);
|
887
|
|
|
|
|
|
|
if (!(defined($type) && defined($sub_type)))
|
888
|
|
|
|
|
|
|
{
|
889
|
|
|
|
|
|
|
$self->_set_err_state($ERR_TYPE_GUESS,
|
890
|
|
|
|
|
|
|
$self->{docTypeWizard}->errmsg());
|
891
|
|
|
|
|
|
|
return 0;
|
892
|
|
|
|
|
|
|
}
|
893
|
|
|
|
|
|
|
|
894
|
|
|
|
|
|
|
# print "TYPE:$type,SUBTYPE:$sub_type\n";
|
895
|
|
|
|
|
|
|
|
896
|
|
|
|
|
|
|
if ($type eq 'text' && $sub_type eq 'html')
|
897
|
|
|
|
|
|
|
{
|
898
|
|
|
|
|
|
|
my $meta_txt;
|
899
|
|
|
|
|
|
|
if (defined($header->{url}))
|
900
|
|
|
|
|
|
|
{
|
901
|
|
|
|
|
|
|
$meta_txt.="url\t$header->{url}\n";
|
902
|
|
|
|
|
|
|
}
|
903
|
|
|
|
|
|
|
if (defined($header->{time}))
|
904
|
|
|
|
|
|
|
{
|
905
|
|
|
|
|
|
|
$meta_txt.="date\t$header->{time}\n";
|
906
|
|
|
|
|
|
|
}
|
907
|
|
|
|
|
|
|
|
908
|
|
|
|
|
|
|
my $base_name;
|
909
|
|
|
|
|
|
|
if (defined($header->{id}))
|
910
|
|
|
|
|
|
|
{
|
911
|
|
|
|
|
|
|
$base_name=$header->{id};
|
912
|
|
|
|
|
|
|
}
|
913
|
|
|
|
|
|
|
else
|
914
|
|
|
|
|
|
|
{
|
915
|
|
|
|
|
|
|
warn "Ainodump document had no ID. URL,time:" .
|
916
|
|
|
|
|
|
|
"($header->{url},$header->{time})\n" if $self->{ainodumpWarnings};
|
917
|
|
|
|
|
|
|
return 1;
|
918
|
|
|
|
|
|
|
}
|
919
|
|
|
|
|
|
|
|
920
|
|
|
|
|
|
|
my $srcenc_setting=$self->{sourceEncoding};
|
921
|
|
|
|
|
|
|
$self->{sourceEncoding}=undef;
|
922
|
|
|
|
|
|
|
my $alvisXML=$self->HTML($text,$meta_txt);
|
923
|
|
|
|
|
|
|
$self->{sourceEncoding}=$srcenc_setting;
|
924
|
|
|
|
|
|
|
if (!defined($alvisXML))
|
925
|
|
|
|
|
|
|
{
|
926
|
|
|
|
|
|
|
$self->_set_err_state($ERR_ALVIS_CONV);
|
927
|
|
|
|
|
|
|
return 0;
|
928
|
|
|
|
|
|
|
}
|
929
|
|
|
|
|
|
|
|
930
|
|
|
|
|
|
|
if (!$self->output_Alvis([$alvisXML],$base_name))
|
931
|
|
|
|
|
|
|
{
|
932
|
|
|
|
|
|
|
$self->_set_err_state($ERR_OUTPUT_ALVIS,
|
933
|
|
|
|
|
|
|
"Base name: \"$base_name\"");
|
934
|
|
|
|
|
|
|
return 0;
|
935
|
|
|
|
|
|
|
}
|
936
|
|
|
|
|
|
|
}
|
937
|
|
|
|
|
|
|
else
|
938
|
|
|
|
|
|
|
{
|
939
|
|
|
|
|
|
|
warn "Ainodump document $header->{id} was not of a convertible " .
|
940
|
|
|
|
|
|
|
"type: $type/$sub_type.\n" if $self->{ainodumpWarnings};
|
941
|
|
|
|
|
|
|
}
|
942
|
|
|
|
|
|
|
|
943
|
|
|
|
|
|
|
return 1;
|
944
|
|
|
|
|
|
|
}
|
945
|
|
|
|
|
|
|
|
946
|
|
|
|
|
|
|
sub _output_set_of_records
|
947
|
|
|
|
|
|
|
{
|
948
|
|
|
|
|
|
|
my $self=shift;
|
949
|
|
|
|
|
|
|
my $set_of_records_txt=shift;
|
950
|
|
|
|
|
|
|
my $path=shift;
|
951
|
|
|
|
|
|
|
|
952
|
|
|
|
|
|
|
if (!defined(open(OUT,">:utf8",$path)))
|
953
|
|
|
|
|
|
|
{
|
954
|
|
|
|
|
|
|
$self->_set_err_state($ERR_WRITING_OUTPUT,"Output file: " .
|
955
|
|
|
|
|
|
|
"\"$path\"");
|
956
|
|
|
|
|
|
|
return 0;
|
957
|
|
|
|
|
|
|
}
|
958
|
|
|
|
|
|
|
print OUT "\n";
|
959
|
|
|
|
|
|
|
print OUT "\n";
|
960
|
|
|
|
|
|
|
print OUT $set_of_records_txt;
|
961
|
|
|
|
|
|
|
print OUT "\n";
|
962
|
|
|
|
|
|
|
close(OUT);
|
963
|
|
|
|
|
|
|
|
964
|
|
|
|
|
|
|
return 1;
|
965
|
|
|
|
|
|
|
}
|
966
|
|
|
|
|
|
|
|
967
|
|
|
|
|
|
|
sub _get_HTML_txt
|
968
|
|
|
|
|
|
|
{
|
969
|
|
|
|
|
|
|
my $self=shift;
|
970
|
|
|
|
|
|
|
my $file_versions=shift;
|
971
|
|
|
|
|
|
|
my $base_name=shift;
|
972
|
|
|
|
|
|
|
my $html_suffix=shift;
|
973
|
|
|
|
|
|
|
|
974
|
|
|
|
|
|
|
my ($html_txt);
|
975
|
|
|
|
|
|
|
if (defined($html_suffix) &&
|
976
|
|
|
|
|
|
|
exists($file_versions->{$base_name}{$html_suffix}))
|
977
|
|
|
|
|
|
|
{
|
978
|
|
|
|
|
|
|
my $html_f=$base_name . "." . $html_suffix;
|
979
|
|
|
|
|
|
|
$html_txt=$self->_read_HTML($html_f);
|
980
|
|
|
|
|
|
|
if (!defined($html_txt))
|
981
|
|
|
|
|
|
|
{
|
982
|
|
|
|
|
|
|
$self->_set_err_state($ERR_READ_HTML,"File:\"$html_f\"");
|
983
|
|
|
|
|
|
|
return undef;
|
984
|
|
|
|
|
|
|
}
|
985
|
|
|
|
|
|
|
}
|
986
|
|
|
|
|
|
|
else # no HTML file
|
987
|
|
|
|
|
|
|
{
|
988
|
|
|
|
|
|
|
$self->_set_err_state($ERR_NO_HTML_F,"Base name:\"$base_name\"");
|
989
|
|
|
|
|
|
|
return undef;
|
990
|
|
|
|
|
|
|
}
|
991
|
|
|
|
|
|
|
|
992
|
|
|
|
|
|
|
return $html_txt;
|
993
|
|
|
|
|
|
|
}
|
994
|
|
|
|
|
|
|
|
995
|
|
|
|
|
|
|
|
996
|
|
|
|
|
|
|
sub _read_HTML
|
997
|
|
|
|
|
|
|
{
|
998
|
|
|
|
|
|
|
my $self=shift;
|
999
|
|
|
|
|
|
|
my $f=shift;
|
1000
|
|
|
|
|
|
|
|
1001
|
|
|
|
|
|
|
if (!defined(open(H,"<$f")))
|
1002
|
|
|
|
|
|
|
{
|
1003
|
|
|
|
|
|
|
$self->_set_err_state($ERR_HTML_F,
|
1004
|
|
|
|
|
|
|
"File: \"$f\".");
|
1005
|
|
|
|
|
|
|
return undef;
|
1006
|
|
|
|
|
|
|
}
|
1007
|
|
|
|
|
|
|
my $txt="";
|
1008
|
|
|
|
|
|
|
while (my $l=)
|
1009
|
|
|
|
|
|
|
{
|
1010
|
|
|
|
|
|
|
$txt.=$l;
|
1011
|
|
|
|
|
|
|
}
|
1012
|
|
|
|
|
|
|
close(H);
|
1013
|
|
|
|
|
|
|
|
1014
|
|
|
|
|
|
|
return $txt;
|
1015
|
|
|
|
|
|
|
}
|
1016
|
|
|
|
|
|
|
|
1017
|
|
|
|
|
|
|
sub _parse_newsXML
|
1018
|
|
|
|
|
|
|
{
|
1019
|
|
|
|
|
|
|
my $self=shift;
|
1020
|
|
|
|
|
|
|
my $newsXML=shift;
|
1021
|
|
|
|
|
|
|
|
1022
|
|
|
|
|
|
|
if ($newsXML=~/^\s*$/isgo)
|
1023
|
|
|
|
|
|
|
{
|
1024
|
|
|
|
|
|
|
return [];
|
1025
|
|
|
|
|
|
|
}
|
1026
|
|
|
|
|
|
|
|
1027
|
|
|
|
|
|
|
my @articles=();
|
1028
|
|
|
|
|
|
|
my ($text,$iso_date,$title,$links);
|
1029
|
|
|
|
|
|
|
|
1030
|
|
|
|
|
|
|
my $doc;
|
1031
|
|
|
|
|
|
|
eval
|
1032
|
|
|
|
|
|
|
{
|
1033
|
|
|
|
|
|
|
$doc=$self->{XMLParser}->parse_string($newsXML);
|
1034
|
|
|
|
|
|
|
};
|
1035
|
|
|
|
|
|
|
if ($@)
|
1036
|
|
|
|
|
|
|
{
|
1037
|
|
|
|
|
|
|
$self->_set_err_state($ERR_XML_PARSE,"$@");
|
1038
|
|
|
|
|
|
|
return undef;
|
1039
|
|
|
|
|
|
|
}
|
1040
|
|
|
|
|
|
|
|
1041
|
|
|
|
|
|
|
my $root=$doc->documentElement();
|
1042
|
|
|
|
|
|
|
|
1043
|
|
|
|
|
|
|
for my $article ($root->getChildrenByTagName('article'))
|
1044
|
|
|
|
|
|
|
{
|
1045
|
|
|
|
|
|
|
$links=Alvis::Document::Links->new();
|
1046
|
|
|
|
|
|
|
if (!defined($links))
|
1047
|
|
|
|
|
|
|
{
|
1048
|
|
|
|
|
|
|
$self->_set_err_state($ERR_LINKS);
|
1049
|
|
|
|
|
|
|
return undef;
|
1050
|
|
|
|
|
|
|
}
|
1051
|
|
|
|
|
|
|
|
1052
|
|
|
|
|
|
|
for my $t ($article->getChildrenByTagName('title'))
|
1053
|
|
|
|
|
|
|
{
|
1054
|
|
|
|
|
|
|
$title=$t->textContent();
|
1055
|
|
|
|
|
|
|
}
|
1056
|
|
|
|
|
|
|
for my $i_d ($article->getChildrenByTagName('iso-date'))
|
1057
|
|
|
|
|
|
|
{
|
1058
|
|
|
|
|
|
|
$iso_date=$i_d->textContent();
|
1059
|
|
|
|
|
|
|
}
|
1060
|
|
|
|
|
|
|
for my $c ($article->getChildrenByTagName('content'))
|
1061
|
|
|
|
|
|
|
{
|
1062
|
|
|
|
|
|
|
$text=$c->textContent();
|
1063
|
|
|
|
|
|
|
}
|
1064
|
|
|
|
|
|
|
for my $ls ($article->getChildrenByTagName('links'))
|
1065
|
|
|
|
|
|
|
{
|
1066
|
|
|
|
|
|
|
for my $l ($ls->getChildrenByTagName('link'))
|
1067
|
|
|
|
|
|
|
{
|
1068
|
|
|
|
|
|
|
my ($l_text,$l_url);
|
1069
|
|
|
|
|
|
|
my $l_type=$l->getAttribute('type');
|
1070
|
|
|
|
|
|
|
for my $l_t ($l->getChildrenByTagName('anchorText'))
|
1071
|
|
|
|
|
|
|
{
|
1072
|
|
|
|
|
|
|
$l_text=$l_t->textContent();
|
1073
|
|
|
|
|
|
|
}
|
1074
|
|
|
|
|
|
|
for my $l_u ($l->getChildrenByTagName('location'))
|
1075
|
|
|
|
|
|
|
{
|
1076
|
|
|
|
|
|
|
$l_url=$l_u->textContent();
|
1077
|
|
|
|
|
|
|
}
|
1078
|
|
|
|
|
|
|
|
1079
|
|
|
|
|
|
|
if (!$links->add($l_url,$l_text,$l_type))
|
1080
|
|
|
|
|
|
|
{
|
1081
|
|
|
|
|
|
|
$self->_set_err_state($ERR_LINK_ADD,
|
1082
|
|
|
|
|
|
|
"Title:\"$title\", " . $links->errmsg());
|
1083
|
|
|
|
|
|
|
next;
|
1084
|
|
|
|
|
|
|
}
|
1085
|
|
|
|
|
|
|
|
1086
|
|
|
|
|
|
|
}
|
1087
|
|
|
|
|
|
|
}
|
1088
|
|
|
|
|
|
|
push(@articles,[$text,$iso_date,$title,$links]);
|
1089
|
|
|
|
|
|
|
}
|
1090
|
|
|
|
|
|
|
|
1091
|
|
|
|
|
|
|
return \@articles;
|
1092
|
|
|
|
|
|
|
}
|
1093
|
|
|
|
|
|
|
|
1094
|
|
|
|
|
|
|
|
1095
|
|
|
|
|
|
|
1;
|
1096
|
|
|
|
|
|
|
|
1097
|
|
|
|
|
|
|
__END__
|