1 | #!/usr/bin/perl -w |
---|
2 | |
---|
3 | # |
---|
4 | # This perl script is intended to perform movie data lookups in russian |
---|
5 | # based on the www.kinox.ru website |
---|
6 | # |
---|
7 | # For more information on MythVideo's external movie lookup mechanism, see |
---|
8 | # the README file in this directory. |
---|
9 | # |
---|
10 | # Author: Denys Dmytriyenko (denis AT denix DOT org) |
---|
11 | # Based on the allocine script by Xavier Hervy |
---|
12 | # |
---|
13 | |
---|
14 | # Note: |
---|
15 | # Encoding on the Web page is cp1251 |
---|
16 | # Internal encoding of this script is koi8-r |
---|
17 | # The output of this script is in utf8 (set by "outcp" below) |
---|
18 | |
---|
19 | use LWP::Simple; # libwww-perl providing simple HTML get actions |
---|
20 | use HTML::Entities; |
---|
21 | use URI::Escape qw(uri_unescape uri_escape uri_escape_utf8); |
---|
22 | |
---|
23 | no encoding; |
---|
24 | |
---|
25 | use Encode; |
---|
26 | use vars qw($opt_h $opt_r $opt_d $opt_i $opt_v $opt_D $opt_M $opt_P $opt_originaltitle $opt_casting $opt_u_dummy); |
---|
27 | use Getopt::Long; |
---|
28 | |
---|
29 | $title = "KinoX Query"; |
---|
30 | $version = "v0.03"; |
---|
31 | $author = "Denys Dmytriyenko"; |
---|
32 | |
---|
33 | # This is the output encoding |
---|
34 | $outcp = "utf8"; |
---|
35 | |
---|
36 | # binmode() does not work for some reason |
---|
37 | # The output ends up being in the wrong encoding |
---|
38 | #binmode(STDOUT, ":utf8"); |
---|
39 | |
---|
40 | # display usage |
---|
41 | sub usage { |
---|
42 | print "usage: $0 -hviocMPD [parameters]\n"; |
---|
43 | print " -h, --help help\n"; |
---|
44 | print " -v, --version display version\n"; |
---|
45 | print " -i, --info display info\n"; |
---|
46 | print " -o, --originaltitle concatenate title and original title\n"; |
---|
47 | print " -c, --casting with -D option, grap the complete actor list (much slower)\n"; |
---|
48 | print "\n"; |
---|
49 | print " -M <query>, --movie query> get movie list\n"; |
---|
50 | print " -D <movieid>, --data <movieid> get movie data\n"; |
---|
51 | # print " -P <movieid>, --poster <movieid> get movie poster\n"; |
---|
52 | exit(-1); |
---|
53 | } |
---|
54 | |
---|
55 | # display 1-line of info that describes the version of the program |
---|
56 | sub version { |
---|
57 | print "$title ($version) by $author\n" |
---|
58 | } |
---|
59 | |
---|
60 | # display 1-line of info that can describe the type of query used |
---|
61 | sub info { |
---|
62 | print "Performs queries using the www.kinox.ru website.\n"; |
---|
63 | } |
---|
64 | |
---|
65 | # display detailed help |
---|
66 | sub help { |
---|
67 | version(); |
---|
68 | info(); |
---|
69 | usage(); |
---|
70 | } |
---|
71 | |
---|
72 | # returns text within 'data' between 'beg' and 'end' matching strings |
---|
73 | sub parseBetween { |
---|
74 | my ($data, $beg, $end)=@_; # grab parameters |
---|
75 | |
---|
76 | my $ldata = lc($data); |
---|
77 | my $start = index($ldata, lc($beg)) + length($beg); |
---|
78 | my $finish = index($ldata, lc($end), $start); |
---|
79 | |
---|
80 | |
---|
81 | if ($start != (length($beg) -1) && $finish != -1) { |
---|
82 | my $result = substr($data, $start, $finish - $start); |
---|
83 | # dont use decode entities &npsp; => spécial characters bug in html::entities ? |
---|
84 | #decode_entities($result); |
---|
85 | return removenbsp($result); |
---|
86 | } |
---|
87 | return ""; |
---|
88 | } |
---|
89 | |
---|
90 | # use to replace by " " (instead of decode_entities) |
---|
91 | sub removenbsp { |
---|
92 | my ($data)=@_; # grab parameters |
---|
93 | |
---|
94 | my $ldata = lc($data); |
---|
95 | my $start = index($ldata, " "); |
---|
96 | while ($start != -1){ |
---|
97 | $data = substr($data, 0, $start). " " .substr($data, $start+6, length($data)); |
---|
98 | $ldata = lc($data); |
---|
99 | $start = index($ldata, " "); |
---|
100 | } |
---|
101 | return $data; |
---|
102 | } |
---|
103 | |
---|
104 | |
---|
105 | # returns text within 'data' without tag |
---|
106 | sub removeTag { |
---|
107 | my ($data)=@_; # grab parameters |
---|
108 | |
---|
109 | my $ldata = lc($data); |
---|
110 | my $start = index($ldata, "<"); |
---|
111 | my $finish = index($ldata, ">", $start)+1; |
---|
112 | while ($start != -1 && $finish != -1){ |
---|
113 | $data = substr($data, 0, $start).substr($data, $finish, length($data)); |
---|
114 | $ldata = lc($data); |
---|
115 | $start = index($ldata, "<"); |
---|
116 | $finish = index($ldata, ">", $start)+1; |
---|
117 | } |
---|
118 | return $data; |
---|
119 | } |
---|
120 | |
---|
121 | # get Movie Data |
---|
122 | sub getMovieData { |
---|
123 | my ($movieid)=@_; # grab movieid parameter |
---|
124 | if (defined $opt_d) { printf("# looking for movie id: '%s'\n", $movieid);} |
---|
125 | |
---|
126 | # get the search results page |
---|
127 | my $request = "http://www.kinox.ru/index.asp?comm=4&num=" . $movieid; |
---|
128 | if (defined $opt_d) { printf("# request: '%s'\n", $request); } |
---|
129 | my $response = get $request; |
---|
130 | |
---|
131 | # parse title and year |
---|
132 | my $sub = parseBetween($response, "<h1>", "</h1>"); |
---|
133 | my ($sub1, $countries) = split("<br>", $sub); |
---|
134 | |
---|
135 | $countries = removeTag($countries); |
---|
136 | $countries =~ s/[\n\r]/ /g; |
---|
137 | Encode::from_to($countries, "windows-1251", $outcp); |
---|
138 | |
---|
139 | my ($title, $original_title) = split("<font size=4 color=#000000> / </font>", $sub1); |
---|
140 | $title = removeTag($title); |
---|
141 | if (!$original_title) { $original_title = "" ;} |
---|
142 | $original_title = removeTag($original_title); |
---|
143 | |
---|
144 | Encode::from_to($title, "windows-1251", $outcp); |
---|
145 | Encode::from_to($original_title, "windows-1251", $outcp); |
---|
146 | |
---|
147 | if (defined $opt_originaltitle){ |
---|
148 | if ($original_title ne ""){ |
---|
149 | $title = $title . " (" . $original_title . ")"; |
---|
150 | } |
---|
151 | } |
---|
152 | |
---|
153 | # parse director |
---|
154 | my $dirq = "<b>òÅÖÉÓÓÅÒ:</b>"; |
---|
155 | Encode::from_to($dirq, "koi8-r", "windows-1251"); |
---|
156 | my $director = parseBetween($response, $dirq, "</a>"); |
---|
157 | $director = removeTag($director); |
---|
158 | $director =~ s/\s{2,}//; |
---|
159 | Encode::from_to($director, "windows-1251", $outcp); |
---|
160 | |
---|
161 | # parse plot |
---|
162 | my $plotq = "<b>ëÒÁÔËÏÅ ÓÏÄÅÒÖÁÎÉÅ:</b>"; |
---|
163 | Encode::from_to($plotq, "koi8-r", "windows-1251"); |
---|
164 | my $plot = parseBetween($response, $plotq, "</p>"); |
---|
165 | $plot = removeTag($plot); |
---|
166 | $plot =~ s/\s{2,}//; |
---|
167 | Encode::from_to($plot, "windows-1251", $outcp); |
---|
168 | |
---|
169 | # parse cast |
---|
170 | my $castq = "<b>÷ ÒÏÌÑÈ:</b>"; |
---|
171 | Encode::from_to($castq, "koi8-r", "windows-1251"); |
---|
172 | my $cast = parseBetween($response, $castq, "<b>"); |
---|
173 | $cast = removeTag($cast); |
---|
174 | $cast =~ s/\s{2,}//; |
---|
175 | $cast =~ s/\s\(.*?\)//g; |
---|
176 | $cast =~ s/\s*,\s*/,/g; |
---|
177 | $cast =~ s/\.$//; |
---|
178 | Encode::from_to($cast, "windows-1251", $outcp); |
---|
179 | |
---|
180 | # studio, year, genres, runtime |
---|
181 | $sub = parseBetween($response, "<td colspan=2 bgcolor=f8f8f8 align=\"center\" valign=\"top\">", "</td></tr></table>"); |
---|
182 | $sub =~ s/ //g; |
---|
183 | $sub =~ s/ //g; |
---|
184 | Encode::from_to($sub, "windows-1251", $outcp); |
---|
185 | |
---|
186 | my $beg = "<font color=\"#008000\">"; |
---|
187 | my $end = "</font>"; |
---|
188 | |
---|
189 | my $start = index($sub, $beg); |
---|
190 | my $finish = index($sub, $end, $start); |
---|
191 | |
---|
192 | $start += length($beg); |
---|
193 | my $studio = substr($sub, $start, $finish - $start); |
---|
194 | $studio = removeTag($studio); |
---|
195 | |
---|
196 | $sub = substr($sub, - (length($sub) - $finish)); |
---|
197 | |
---|
198 | $start = index($sub, $beg); |
---|
199 | $finish = index($sub, $end, $start); |
---|
200 | |
---|
201 | $start += length($beg); |
---|
202 | my $year = substr($sub, $start, $finish - $start); |
---|
203 | $year = removeTag($year); |
---|
204 | |
---|
205 | $sub = substr($sub, - (length($sub) - $finish)); |
---|
206 | |
---|
207 | $start = index($sub, $beg); |
---|
208 | $finish = index($sub, $end, $start); |
---|
209 | |
---|
210 | $start += length($beg); |
---|
211 | my $genres = substr($sub, $start, $finish - $start); |
---|
212 | $genres = removeTag($genres); |
---|
213 | $genres =~ s|\s*/\s*|,|g; |
---|
214 | |
---|
215 | $sub = substr($sub, - (length($sub) - $finish)); |
---|
216 | |
---|
217 | $start = index($sub, $beg); |
---|
218 | $finish = index($sub, $end, $start); |
---|
219 | |
---|
220 | $start += length($beg); |
---|
221 | my $runtime = substr($sub, $start, $finish - $start); |
---|
222 | $runtime = removeTag($runtime); |
---|
223 | |
---|
224 | # output fields (these field names must match what MythVideo is looking for) |
---|
225 | print "Title:$title\n"; |
---|
226 | if (!(defined $opt_originaltitle)){ |
---|
227 | print "OriginalTitle:$original_title\n"; |
---|
228 | } |
---|
229 | print "Year:$year\n"; |
---|
230 | print "Director:$director\n"; |
---|
231 | print "Plot:$plot\n"; |
---|
232 | print "Runtime:$runtime\n"; |
---|
233 | print "Cast:$cast\n"; |
---|
234 | print "Genres:$genres\n"; |
---|
235 | print "Countries:$countries\n"; |
---|
236 | } |
---|
237 | |
---|
238 | # dump Movie list: 1 entry per line, each line as 'movieid:Movie Title' |
---|
239 | sub getMovieList { |
---|
240 | my ($filename, $options)=@_; # grab parameters |
---|
241 | |
---|
242 | # If we wanted to inspect the file for any reason we can do that now |
---|
243 | |
---|
244 | # |
---|
245 | # Convert filename into a query string |
---|
246 | # (use same rules that Metadata::guesTitle does) |
---|
247 | my $query = $filename; |
---|
248 | |
---|
249 | $query = uri_unescape($query); # in case it was escaped |
---|
250 | # Strip off the file extension |
---|
251 | if (rindex($query, '.') != -1) { |
---|
252 | $query = substr($query, 0, rindex($query, '.')); |
---|
253 | } |
---|
254 | # Strip off anything following '(' - people use this for general comments |
---|
255 | if (rindex($query, '(') != -1) { |
---|
256 | $query = substr($query, 0, rindex($query, '(')); |
---|
257 | } |
---|
258 | # Strip off anything following '[' - people use this for general comments |
---|
259 | if (rindex($query, '[') != -1) { |
---|
260 | $query = substr($query, 0, rindex($query, '[')); |
---|
261 | } |
---|
262 | # Strip off anything following '-' - people use this for general comments |
---|
263 | if (index($query, '-') != -1) { |
---|
264 | $query = substr($query, 0, index($query, '-')); |
---|
265 | } |
---|
266 | |
---|
267 | # IMDB searches do better if any trailing ,The is left off |
---|
268 | $query =~ /(.*), The$/i; |
---|
269 | if ($1) { $query = $1; } |
---|
270 | Encode::from_to($query, "koi8-r", "windows-1251"); |
---|
271 | |
---|
272 | # prepare the url |
---|
273 | $query = uri_escape($query); |
---|
274 | if (!$options) { $options = "" ;} |
---|
275 | if (defined $opt_d) { |
---|
276 | printf("# query: '%s', options: '%s'\n", $query, $options); |
---|
277 | } |
---|
278 | |
---|
279 | my $count = 0; |
---|
280 | my $typerecherche = 3; |
---|
281 | |
---|
282 | while (($typerecherche <=5) && ($count ==0)){ |
---|
283 | # get the search results page |
---|
284 | my $request = "http://www.kinox.ru/index.asp?comm=1&fop=false&pack=0&kw=$query"; |
---|
285 | if (defined $opt_d) { printf("# request: '%s'\n", $request); } |
---|
286 | my $response = get $request; |
---|
287 | if (defined $opt_d) { printf("# response: '%s'\n", $response); } |
---|
288 | |
---|
289 | # |
---|
290 | # don't try to invent if it doesn't exist |
---|
291 | # |
---|
292 | my $notfnd = "ÎÉÞÅÇÏ ÎÅÂÙÌÏ ÎÁÊÄÅÎÏ"; |
---|
293 | Encode::from_to($notfnd, "koi8-r", "windows-1251"); |
---|
294 | return if $response =~ /$notfnd/; |
---|
295 | |
---|
296 | # extract possible matches |
---|
297 | # possible matches are grouped in several catagories: |
---|
298 | # exact, partial, and approximate |
---|
299 | my $exact_matches = $response; |
---|
300 | # parse movie list from matches |
---|
301 | my $beg = "<a class=l2 href=\"index.asp?comm=4&num="; |
---|
302 | my $end = "</a>"; |
---|
303 | my $begy = "colspan=2 align=center>"; |
---|
304 | my $endy = "</td>"; |
---|
305 | |
---|
306 | my @movies; |
---|
307 | |
---|
308 | my $data = $exact_matches; |
---|
309 | if ($data eq "") { |
---|
310 | if (defined $opt_d) { printf("# no results\n"); } |
---|
311 | $typerecherche = $typerecherche +2 ; |
---|
312 | }else{ |
---|
313 | my $start = index($data, $beg); |
---|
314 | my $finish = index($data, $end, $start); |
---|
315 | |
---|
316 | my $title; |
---|
317 | while ($start != -1) { |
---|
318 | $start += length($beg); |
---|
319 | my $sub = substr($data, $start, $finish - $start); |
---|
320 | my ($movienum, $moviename) = split("\">", $sub); |
---|
321 | $title = removeTag($moviename); |
---|
322 | $moviename = removeTag($moviename); |
---|
323 | |
---|
324 | $title =~ s/\s{2,}//; |
---|
325 | Encode::from_to($title, "windows-1251", $outcp); |
---|
326 | |
---|
327 | # advance data to next field |
---|
328 | $data = substr($data, - (length($data) - $finish)); |
---|
329 | |
---|
330 | $start = index($data, $begy); |
---|
331 | $finish = index($data, $endy, $start); |
---|
332 | $start += length($begy); |
---|
333 | $sub = substr($data, $start, $finish - $start); |
---|
334 | my $movieyear = removeTag($sub); |
---|
335 | |
---|
336 | if ($movieyear){$title = $title." (".$movieyear.")"; } |
---|
337 | $moviename=$title ; |
---|
338 | |
---|
339 | # advance data to next movie |
---|
340 | $data = substr($data, - (length($data) - $finish)); |
---|
341 | $start = index($data, $beg); |
---|
342 | $finish = index($data, $end, $start); |
---|
343 | |
---|
344 | # add to array |
---|
345 | $movies[$count++] = $movienum . ":" . $moviename; |
---|
346 | } |
---|
347 | |
---|
348 | # display array of values |
---|
349 | for $movie (@movies) { |
---|
350 | print "$movie\n"; |
---|
351 | } |
---|
352 | } |
---|
353 | } |
---|
354 | } |
---|
355 | |
---|
356 | # |
---|
357 | # Main Program |
---|
358 | # |
---|
359 | |
---|
360 | # parse command line arguments |
---|
361 | |
---|
362 | GetOptions( "utf8" => \$opt_u_dummy, |
---|
363 | "version" => \$opt_v, |
---|
364 | "info" => \$opt_i, |
---|
365 | "originaltitle" => \$opt_originaltitle, |
---|
366 | "casting" => \$opt_casting, |
---|
367 | "Data" => \$opt_D, |
---|
368 | "Movie" => \$opt_M, |
---|
369 | "Poster" => \$opt_P |
---|
370 | ); |
---|
371 | |
---|
372 | |
---|
373 | #$opt_d = 1; |
---|
374 | |
---|
375 | # print out info |
---|
376 | if (defined $opt_v) { version(); exit 1; } |
---|
377 | if (defined $opt_i) { info(); exit 1; } |
---|
378 | |
---|
379 | # print out usage if needed |
---|
380 | if (defined $opt_h || $#ARGV<0) { help(); } |
---|
381 | |
---|
382 | if (defined $opt_D) { |
---|
383 | # take movieid from cmdline arg |
---|
384 | $movieid = shift || die "Usage : $0 -D <movieid>\n"; |
---|
385 | getMovieData($movieid); |
---|
386 | } |
---|
387 | |
---|
388 | elsif (defined $opt_M) { |
---|
389 | # take query from cmdline arg |
---|
390 | #$options = shift || die "Usage : $0 -M <query>\n"; |
---|
391 | my $query; |
---|
392 | my $options = ''; |
---|
393 | foreach $key (0 .. $#ARGV) { |
---|
394 | $query .= $ARGV[$key]. ' '; |
---|
395 | } |
---|
396 | getMovieList($query, $options); |
---|
397 | } |
---|