Context Navigation

abc_website @ 587

Revision 587, 25.3 kB (checked in by lincoln, 6 years ago)
commit --set overrides for abc_website/abc2_website
Property svn:executable set to ``*

Line
1	#!/usr/bin/perl -w
2
3	# ABC/ABC2 au_tv guide grabber - runs from "Shepherd" master grabber
4	# * written by ltd
5	# * uses ABC website for ABC or ABC2 data
6	# * when used in conjunction with Shepherd, shepherd can collect other channels
7	# using other grabbers
8	# * this does NOT use any config file - all settings are passed in from shepherd
9
10	# changelog:
11	# 1.50 22sep06 added support for "shepherd" master grabber script
12	# 1.51 02oct06 --ready option
13	# 1.52 03oct06 split out abc grabber into its own grabber
14	# 1.55 09oct06 formalize --cheap option
15	# 1.56 20oct06 misc cleanups
16	# 1.60 11nov06 fix midday time calculation
17	# 1.70 16nov06 also use "printable" TV guide to determine 'station close'
18	# 2.00 23nov06 simplified
19
20	use strict;
21
22	my $progname = "abc_website";
23	my $chan_id = "ABC";
24	my $version = "2.11";
25
26	use LWP::UserAgent;
27	use XMLTV;
28	use POSIX qw(strftime mktime);
29	use Getopt::Long;
30	use HTML::TreeBuilder;
31	use Data::Dumper;
32	use Storable;
33
34	#
35	# constants
36	#
37	my $urls;
38	$urls->{station_close}->{ABC} = "http://www.abc.net.au/tv/guide/abctvweekguide.htm";
39	$urls->{station_close}->{ABC2} = "http://www.abc.net.au/tv/guide/abc2weekguide.htm";
40	$urls->{guide}->{ABC} = "http://www.abc.net.au/tv/guide/netw";
41	$urls->{guide}->{ABC2} = "http://www.abc.net.au/tv/guide/abc2";
42
43	#
44	# some initial cruft
45	#
46
47	my $script_start_time = time;
48	my $gmt_offset;
49	my %stats;
50	my $channels, my $opt_channels, my $gaps;
51	my $tv_guide;
52	my $data_cache;
53	my $override_settings = { };
54	my @station_close_data;
55	my $writer;
56	my %amp = ( nbsp => ' ', qw{ amp & lt < gt > apos ' quot " } );
57
58	my $ua;
59	$ua = LWP::UserAgent->new('timeout' => 30, 'keep_alive' => 30, 'agent' => "Shepherd / $progname $version");
60	$ua->env_proxy;
61	# $ua->cookie_jar({});
62	$ua->conn_cache(LWP::ConnCache->new());
63	$\| = 1;
64
65	#
66	# parse command line
67	#
68
69	my $opt_days = 7; # default
70	my $opt_offset = 0; # default
71	my $opt_timezone = "1000"; # default
72	my $opt_outputfile = $progname.".xmltv"; # default
73	my $opt_configfile = $progname.".conf"; # ignored
74	my $opt_cache_file = $progname.".storable.cache";
75	my $opt_old_cache_file = $progname.".cache";
76	my $opt_channels_file= "";
77	my $opt_gaps_file= "";
78	my $opt_no_cache = 0;
79	my $opt_cheap = 0;
80	my $opt_fast = 0;
81	my $opt_warper = 0;
82	my $opt_obfuscate = 0;
83	my $opt_do_extra_days = 0;
84	my $opt_set = "";
85	my $opt_help = 0;
86	my $opt_version = 0;
87	my $opt_desc = 0;
88	my $opt_dont_retry = 0;
89	my $debug = 0;
90	my $lang = "en";
91	my $region = 94;
92	my $time_offset = 0;
93
94	GetOptions(
95	'region=i' => \$region,
96	'days=i' => \$opt_days,
97	'offset=i' => \$opt_offset,
98	'timezone=s' => \$opt_timezone,
99	'channels_file=s' => \$opt_channels_file,
100	'gaps_file=s' => \$opt_gaps_file,
101	'output=s' => \$opt_outputfile,
102	'config-file=s' => \$opt_configfile,
103	'cache-file=s' => \$opt_cache_file,
104	'do-extra-days' => \$opt_do_extra_days,
105	'fast' => \$opt_fast,
106	'no-cache' => \$opt_no_cache,
107	'cheap' => \$opt_cheap,
108	'debug+' => \$debug,
109	'warper' => \$opt_warper,
110	'lang=s' => \$lang,
111	'obfuscate' => \$opt_obfuscate,
112	'no-retry' => \$opt_dont_retry,
113	'set=s' => \$opt_set,
114	'help' => \$opt_help,
115	'verbose' => \$opt_help,
116	'version' => \$opt_version,
117	'ready' => \$opt_version,
118	'desc' => \$opt_desc,
119	'v' => \$opt_help);
120
121	&help if ($opt_help);
122
123	if ($opt_version \|\| $opt_desc) {
124	printf "%s %s\n",$progname,$version;
125	printf "%s is a details-aware grabber that collects decent quality data using the ABC website for %s only.",$progname,$chan_id if $opt_desc;
126	exit(0);
127	}
128
129	&set_override if ($opt_set ne "");
130
131	die "no channel file specified, see --help for instructions\n", if ($opt_channels_file eq "");
132
133	#
134	# go go go!
135	#
136
137	my $starttime = time;
138	&read_cache if ($opt_no_cache == 0);
139
140	&log(sprintf "going to %s%s %s%d%s days%s of data into %s (%s%s)",
141	($opt_gaps_file ne "" ? "micro-gap " : ""),
142	($opt_cheap ? "verify (cache-validate)" : "grab"),
143	($opt_do_extra_days ? "somewhere between " : ""),
144	$opt_days,
145	($opt_do_extra_days ? " to 28" : ""),
146	($opt_offset ? " (skipping first $opt_offset days)" : ""),
147	$opt_outputfile,
148	($opt_no_cache ? "without caching" : "with caching"),
149	($opt_warper ? ", anonymously" : ""));
150
151	# read channels file
152	if (-r $opt_channels_file) {
153	local (@ARGV, $/) = ($opt_channels_file);
154	no warnings 'all'; eval <>; die "$@" if $@;
155	} else {
156	die "WARNING: channels file $opt_channels_file could not be read: $!\n";
157	}
158	die "nothing to do; $chan_id not in channels lineup!\n" if (!defined $channels->{$chan_id});
159
160	# if just filling in microgaps, parse gaps
161	if ($opt_gaps_file ne "") {
162	if (-r $opt_gaps_file) {
163	local (@ARGV, $/) = ($opt_gaps_file);
164	no warnings 'all'; eval <>; die "$@" if $@;
165	} else {
166	die "WARNING: gaps_file $opt_gaps_file could not be read: $!\n";
167	}
168	}
169
170	my %writer_args = ( encoding => 'ISO-8859-1' );
171	my $fh = new IO::File(">$opt_outputfile") \|\| die "can't open $opt_outputfile: $!";
172	$writer_args{OUTPUT} = $fh;
173
174	$writer = new XMLTV::Writer(%writer_args);
175	$writer->start( { 'source-info-name' => "$progname $version", 'generator-info-name' => "$progname $version"} );
176	$writer->write_channel( { 'display-name' => [[ $chan_id, $lang ]], 'id' => $channels->{$chan_id} } );
177
178	&get_station_close($channels->{$chan_id}, $urls->{station_close}->{$chan_id});
179	&get_abc_data($channels->{$chan_id}, $urls->{guide}->{$chan_id});
180	&write_cache if ($opt_no_cache == 0);
181
182	$writer->end;
183
184	&print_stats;
185	exit(0);
186
187	######################################################################################################
188	# help
189
190	sub help
191	{
192	print<<EOF
193	$progname $version
194
195	options are as follows:
196	--help show these help options
197	--days=N fetch 'n' days of data (default: $opt_days)
198	--output=file send xml output to file (default: "$opt_outputfile")
199	--config-file=file (ignored - historically used by grabbers not not this one)
200	--no-cache don't use a cache to optimize (reduce) number of web queries
201	--cheap validate contents of cache - fetch summary only, not details
202	--cache-file=file where to store cache (default "$opt_cache_file")
203	--fast don't run slow - get data as quick as you can - not recommended
204	--debug increase debug level
205	--warper fetch data using WebWarper web anonymizer service
206	--obfuscate pretend to be a proxy servicing multiple clients
207	--do-extra-days fetch extra (21 days) from ABC website
208	--no-retry if webserver is rejecting our request, don't retry (default: do retry)
209	--lang=[s] set language of xmltv output data (default $lang)
210
211	--region=N set region for where to collect data from (default: $region)
212	--channels_file=file where to get channel data from (if not set manually)
213	--timezone=HHMM timezone for channel data (default: $opt_timezone)
214
215	--set (option):(1/0) setting override options (1=enable, 0=disable)
216	do_extra_days:1/0 enable/disable fetching up to 24 days
217	fast:1/0 enable/disable extra-fast grab speed (not recommended)
218	debug:1/0 enable/disable debugging
219
220	EOF
221	;
222
223	exit(0);
224	}
225
226	######################################################################################################
227
228	sub set_override
229	{
230	&read_cache;
231	my ($setting, $val) = split(/:/,$opt_set);
232
233	die "--set format is (setting):(value) where value is 0 for disable, 1 for enable.\n"
234	if (($val ne "0") && ($val ne "1"));
235
236	die "unknown '--set' parameter '$setting', see --help for details.\n"
237	if (($setting ne "do_extra_days") &&
238	($setting ne "fast") &&
239	($setting ne "debug"));
240
241	$override_settings->{$setting} = $val;
242	printf "%s: override parameter %s: %s\n", $progname, $setting, ($val eq "0" ? "disabled" : "enabled");
243
244	&write_cache;
245	exit(0);
246	}
247
248	######################################################################################################
249	# populate cache
250
251	sub read_cache
252	{
253	if (-r $opt_cache_file) {
254	my $store = Storable::retrieve($opt_cache_file);
255	$data_cache = $store->{data_cache};
256	$override_settings = $store->{override_settings};
257
258	# apply settings overrides
259	$opt_do_extra_days = 1 if ((defined $override_settings->{do_extra_days}) && ($override_settings->{do_extra_days} == 1));
260	$opt_fast = 1 if ((defined $override_settings->{fast}) && ($override_settings->{fast} == 1));
261	$debug = 1 if ((defined $override_settings->{debug}) && ($override_settings->{debug} > 0));
262	} else {
263	printf "WARNING: no programme cache $opt_cache_file - have to fetch all details\n";
264
265	# try to write to it - if directory doesn't exist this will then cause an error
266	&write_cache;
267	}
268	}
269
270	######################################################################################################
271	# write out updated cache
272
273	sub write_cache
274	{
275	# cleanup old entries from cache
276	for my $cache_key (keys %{$data_cache}) {
277	my ($starttime, @rest) = split(/,/,$cache_key);
278	if ($starttime < (time-86400)) {
279	delete $data_cache->{$cache_key};
280	$stats{removed_items_from_cache}++;
281	}
282	}
283
284	my $store;
285	$store->{data_cache} = $data_cache;
286	$store->{override_settings} = $override_settings;
287	Storable::store($store, $opt_cache_file);
288	}
289
290	######################################################################################################
291
292	sub get_abc_data
293	{
294	my ($xmlid,$urlbase) = @_;
295	my $try_to_add_abc_detail;
296	my $unprocessed_programmes = 0;
297	my $stop_fetching = 0;
298	my @unprocessed_progname, my @unprocessed_starttime, my @unprocessed_url;
299
300	my $to_skip = $opt_offset;
301	my $daynum = 0;
302	my @gap_s, my @gap_e;
303
304	$opt_days = 28 if (($opt_do_extra_days) && ($opt_gaps_file eq "") && ($opt_offset == 0) && ($opt_days == 7));
305	my $days_left = $opt_days;
306
307	DAYS: while ($days_left > 0) {
308	my $currtime = $starttime + ($daynum * 86400);
309	$days_left--;
310	$daynum++;
311
312	if ($to_skip > 0) {
313	$to_skip--;
314	next;
315	}
316
317	if ($opt_gaps_file ne "") { # micro-gap mode!
318	my $found_gap_match = 0;
319
320	if ((defined $gaps) && (defined $gaps->{$chan_id})) {
321	foreach my $g (@{($gaps->{$chan_id})}) {
322	my ($s, $e) = split(/-/,$g);
323	if (($s >= $currtime) && ($s <= ($currtime+86400))) {
324	$found_gap_match++;
325	push(@gap_s,$s);
326	push(@gap_e,$e);
327	printf "including day %d channel '%s' gap start %d, gap end %d\n",
328	$daynum, $chan_id, $s, $e if $debug;
329	}
330	}
331	}
332	next if (!$found_gap_match); # no gaps for this day - skip!
333	}
334
335	my @timeattr = localtime($currtime); # 0=sec,1=min,2=hour,3=day,4=month,5=year,6=wday,7=yday,8=isdst
336	$timeattr[0] = 0; # zero seconds
337
338	my $url = sprintf "%s/%s.htm",$urlbase, POSIX::strftime("%Y%m/%Y%m%d",localtime($currtime));
339
340	my $data;
341	my $tree;
342	my $tries = 0;
343	while (($tries < 5) && (!defined $tree)) {
344	$tries++;
345	&log((sprintf "fetching %s summary data: day %d of %d%s",
346	$xmlid, $daynum, $opt_days, ($tries > 1 ? " (try $tries)" : "")));
347	$data = &get_url($url);
348	$tree = HTML::TreeBuilder->new_from_content($data) if ((defined $data) && ($data ne ""));
349
350	if ((!$tree) && ($tries < 5)) {
351	# if fetching extra days, bail out at first error
352	if ($daynum > 7) {
353	&log("failed to fetch $url, assuming we only have $daynum days..");
354	$days_left = 0;
355	next DAYS;
356	}
357
358	# slow down ..
359	my $waittime = ($tries*10)+int(rand(10));
360	&log("failed to fetch $url, retrying in $waittime secs");
361	sleep($waittime);
362	$stats{slept_for} += $waittime;
363	}
364	}
365	if (!defined $tree) {
366	&log("failed to fetch $url after $tries attempts; skipping");
367
368	die "couldn't fetch first daily page after $tries attempts, network is probably down. aborting!"
369	if ((!defined $stats{abc_daily_pages}) \|\| ($stats{abc_daily_pages} == 0));
370
371	next;
372	}
373
374	my $seen_programmes = 0;
375	my $seen_pm = 0;
376
377	for ($tree->look_down('_tag' => 'div', 'class' => 'scheduleDiv')) {
378	foreach my $tree_tr ($_->look_down('_tag' => 'tr')) {
379	if (my $tree_row = $tree_tr->look_down('_tag' => 'th', 'scope' => 'row')) {
380	if ($tree_row->as_text() =~ /^(\d+):(\d+)(.)m/) {
381	$timeattr[2] = $1; # hour
382	$timeattr[1] = $2; # min
383
384	if ($3 eq "p") {
385	# pm
386	$timeattr[2] += 12 if ($timeattr[2] != 12);
387	$seen_pm = 1;
388	}
389	my $found_time = mktime(@timeattr);
390
391	# handle programmes that are after midnight
392	if (($seen_pm) && ($3 eq "a")) {
393	if ($timeattr[2] == 12) {
394	$found_time += (126060); # 12:xx am
395	} else {
396	$found_time += (246060);
397	}
398	}
399
400	if ($tree_tr->look_down('_tag' => 'td')) {
401	foreach my $prog ($tree_tr->look_down('_tag' => 'a')) {
402	my $programme = $prog->as_text();
403	my $progurl = $prog->attr('href');
404
405	if ($progurl =~ /^\/tv\/guide\//) {
406	printf "day %d time '%s' (%s) prog: %s url: %s\n",
407	$daynum,$tree_row->as_text(),POSIX::strftime("%Y%m%d%H%M", localtime($found_time)),
408	$programme,$progurl if ($debug && $debug > 1);
409
410	$unprocessed_progname[$unprocessed_programmes] = $programme;
411	$unprocessed_starttime[$unprocessed_programmes] = $found_time;
412	$unprocessed_url[$unprocessed_programmes] = "http://www.abc.net.au".$progurl;
413	$unprocessed_programmes++;
414	$seen_programmes++;
415	} else {
416	printf "ignoring prog %s because url %s is not a detail page\n",
417	$programme,$progurl if $debug;
418	}
419	}
420	}
421	}
422	}
423	}
424	}
425
426	if ($seen_programmes > 0) {
427	$stats{abc_daily_pages}++;
428
429	if (defined $station_close_data[$daynum]) {
430	# get station-close time from the previously-fetched "weekly programme guide"
431
432	$unprocessed_progname[$unprocessed_programmes] = "Station Close";
433	$unprocessed_starttime[$unprocessed_programmes] = $station_close_data[$daynum];
434	$unprocessed_url[$unprocessed_programmes] = "";
435	$unprocessed_programmes++;
436	} else {
437	# throw away last programme from each day - we can't use it as
438	# we don't have a 'stop' time for it
439
440	printf "throwing away '%s' (%s) because we won't have a valid stop time\n",
441	$unprocessed_progname[$unprocessed_programmes-1],
442	POSIX::strftime("%Y%m%d%H%M", localtime($unprocessed_starttime[$unprocessed_programmes-1]))
443	if $debug;
444	$unprocessed_progname[$unprocessed_programmes-1] = "";
445	}
446	} else {
447	# if we were trying to fetch more than 7 days, stop on first day with no programmes
448	if ($daynum > 7) {
449	&log("failed to fetch $url, assuming we only have $daynum days..");
450	$days_left = 0;
451	next DAYS;
452	}
453	}
454	}
455
456	# have 'n' days of this channel unprocessed - process it!
457	&log((sprintf "have summary data, now fetching detail pages for up to %d programmes..",$unprocessed_programmes-2));
458
459	for (my $i = 0; $i < ($unprocessed_programmes-1); $i++) {
460	next if ($unprocessed_progname[$i] eq "");
461
462	# if we are micro-gap fetching, only include programmes which match our micro gaps
463	if ($opt_gaps_file ne "") {
464	my $found_gap_match = 0;
465	for (my $g_num = 0; $g_num < $#gap_s; $g_num++) {
466	$found_gap_match++
467	if ((($gap_s[$g_num] >= $unprocessed_starttime[$i]) &&
468	($gap_s[$g_num] <= $unprocessed_starttime[$i+1])) \|\|
469	(($gap_e[$g_num] >= $unprocessed_starttime[$i]) &&
470	($gap_e[$g_num] <= $unprocessed_starttime[$i+1])) \|\|
471	(($gap_s[$g_num] <= $unprocessed_starttime[$i]) &&
472	($gap_e[$g_num] >= $unprocessed_starttime[$i+1])));
473	}
474	next if (!$found_gap_match);
475
476	$stats{programme_gaps_used}++;
477	printf "gap-fetching: including prog '%s', start %d, end %d\n", $unprocessed_progname[$i],
478	$unprocessed_starttime[$i], $unprocessed_starttime[$i+1] if $debug;
479	}
480
481	$stats{programmes}++;
482	my $prog;
483
484	my $cache_key = sprintf "%d,%d,%s,%s", $unprocessed_starttime[$i], $unprocessed_starttime[$i+1], $xmlid, $unprocessed_progname[$i];
485
486	$prog->{'channel'} = $xmlid;
487	$prog->{'start'} = POSIX::strftime("%Y%m%d%H%M", localtime($unprocessed_starttime[$i]));
488	$prog->{'stop'} = POSIX::strftime("%Y%m%d%H%M", localtime($unprocessed_starttime[$i+1]));
489	$prog->{'title'} = [[ $unprocessed_progname[$i], $lang ]];
490
491	if (defined $data_cache->{$cache_key}) {
492	$stats{used_cached_data}++;
493	} else {
494	if ((!$opt_cheap) && ($unprocessed_url[$i] ne "")) {
495	$stats{portal_detail_pages}++;
496	&get_one_abc_event($cache_key, $unprocessed_url[$i]);
497
498	if (($stats{portal_detail_pages} % 25) == 1) {
499	&log((sprintf " .. at %s detail page %d of %d (used %d cached entries)",
500	$xmlid, ($i+1), $unprocessed_programmes-2,
501	(defined $stats{used_cached_data} ? $stats{used_cached_data} : 0)));
502
503	if (!$opt_fast) {
504	# slow down ..
505	my $waittime = 3 + int(rand(10));
506	sleep($waittime);
507	$stats{slept_for} += $waittime;
508	}
509	}
510	}
511	}
512
513	if (defined $data_cache->{$cache_key}) {
514	$prog->{'sub-title'} = [[ $data_cache->{$cache_key}->{subtitle}, $lang ]]
515	if $data_cache->{$cache_key}->{subtitle};
516	$prog->{'desc'} = [[ $data_cache->{$cache_key}->{desc}, $lang ]]
517	if $data_cache->{$cache_key}->{desc};
518	$prog->{'category'} = [[ $data_cache->{$cache_key}->{genre}, $lang ]]
519	if $data_cache->{$cache_key}->{genre};
520	$prog->{'previously-shown'} = { } if (defined $data_cache->{$cache_key}->{repeat});
521	$prog->{'subtitles'} = [ { 'type' => 'teletext' } ] if (defined $data_cache->{$cache_key}->{cc});
522	$prog->{'rating'} = [ [ $data_cache->{$cache_key}->{rating}, 'ABA', undef] ]
523	if (defined $data_cache->{$cache_key}->{rating});
524	}
525
526	&cleanup($prog);
527	$writer->write_programme($prog);
528	}
529	}
530
531	######################################################################################################
532
533	sub get_one_abc_event
534	{
535	my ($cache_key, $url) = @_;
536	my $seen_programme = 0;
537
538	my $tries = 0;
539	my $tree;
540	while (($tries < 5) && (!$tree)) {
541	$tries++;
542	my $data = &get_url($url);
543	$tree = HTML::TreeBuilder->new_from_content($data) if ((defined $data) && ($data ne ""));
544
545	if ((!$tree) && ($tries < 5)) {
546	# slow down ..
547	my $waittime = ($tries*20)+int(rand(20));
548	&log("failed to fetch $url on try $tries, retrying in $waittime secs");
549	sleep($waittime);
550	$stats{slept_for} += $waittime;
551	}
552	}
553	if (!defined $tree) {
554	&log("failed to fetch $url after $tries attempts; skipping");
555	return;
556	}
557
558	if (my $inner_tree = $tree->look_down('_tag' => 'div', 'class' => 'column2')) {
559	my $event_title = undef, my $event_subtitle = undef, my $event_description = undef, my $event_genre = undef;
560
561	if (my $prog_h2 = $inner_tree->look_down('_tag' => 'h2')) {
562	my $full_title = $prog_h2->as_HTML();
563	($event_title,$event_subtitle) = split(/<br>/,$full_title);
564
565	$event_title =~ s/(<[a-zA-Z0-9]+\>)//g; # remove html tags
566	$event_title =~ s/(^\n\|\n$)//g; # strip trailing/leading blank lines
567
568	if ($event_subtitle) {
569	$event_subtitle =~ s/(<[\/a-zA-Z0-9]+\>)//g; # remove html tags
570	$event_subtitle =~ s/(^\n\|\n$)//g; # strip trailing/leading blank lines
571	$data_cache->{$cache_key}->{subtitle} = $event_subtitle;
572	}
573	}
574
575	my $paranum = 0;
576	my $seen_genre = 0;
577	foreach my $para ($inner_tree->look_down('_tag' => 'p')) {
578	$paranum++;
579
580	if (($paranum > 1) && (!($para->as_text() =~ /^Go to website/)) && (!($para->as_text() =~ /^Send to a Friend/))) {
581	if (my $try_genre = $para->look_down('_tag' => 'a')) {
582	$data_cache->{$cache_key}->{genre} = $try_genre->as_text();
583	$seen_genre = 1;
584	}
585
586	if (!$seen_genre) {
587	$data_cache->{$cache_key}->{desc} .= $para->as_text() . "\n";
588	} else {
589	$data_cache->{$cache_key}->{repeat} = 1 if ($para->as_text() =~ /Repeat/);
590	$data_cache->{$cache_key}->{cc} = 1 if ($para->as_text() =~ /CC/);
591	$data_cache->{$cache_key}->{rating} = $1 if ($para->as_text() =~ /(M\|PG\|G)/);
592	}
593	}
594	}
595
596	if (defined $data_cache->{$cache_key}->{desc}) {
597	$data_cache->{$cache_key}->{desc} =~ s/(^\n\|\n$)//g; # strip trailing/leading blank lines
598	$data_cache->{$cache_key}->{desc} =~ s/(^\s+\|\s+$)//g; # strip trailing/leading spaces
599	delete $data_cache->{$cache_key}->{desc} if ($data_cache->{$cache_key}->{desc} eq "");
600	}
601
602	$seen_programme++;
603	$stats{added_cached_data}++;
604
605	&write_cache if (($opt_no_cache == 0) &&
606	(($stats{added_cached_data} % 30) == 0)); # incrementally write
607	}
608
609	if ($seen_programme == 0) {
610	printf "WARNING: failed to parse any programme data from '%s' - blocked/rate-limited/format-changed?\n",$url;
611	$stats{failed_to_parse_portal_detail_page}++;
612	}
613	}
614
615	######################################################################################################
616	# logic to fetch a page via http
617
618	sub get_url
619	{
620	my $url = shift;
621	my $response;
622	my $attempts = 0;
623	my ($raw, $page, $base);
624
625	$url =~ s#^http://#http://webwarper.net/ww/# if $opt_warper;
626	my $request = HTTP::Request->new(GET => $url);
627	$request->header('Accept-Encoding' => 'gzip');
628
629	if ($opt_obfuscate) {
630	$ua->agent('Mozilla/5.0 (Windows; U; Windows NT 5.1; en-us');
631	my $randomaddr = sprintf "203.%d.%d.%d",rand(255),rand(255),(rand(254)+1);
632	$request->header('Via' => '1.0 proxy:81 (Squid/2.3.STABLE3)');
633	$request->header('X-Forwarded-For' => $randomaddr);
634	}
635	$response = $ua->request($request);
636	if (!($response->is_success)) {
637	$stats{http_failed_requests}++;
638	return undef;
639	}
640
641	$stats{bytes_fetched} += do {use bytes; length($response->content)};
642	$stats{http_successful_requests}++;
643
644	if ($response->header('Content-Encoding') &&
645	$response->header('Content-Encoding') eq 'gzip') {
646	$stats{compressed_pages} += do {use bytes; length($response->content)};
647	$response->content(Compress::Zlib::memGunzip($response->content));
648	}
649	return $response->content;
650	}
651
652	######################################################################################################
653
654	sub log
655	{
656	my ($entry) = @_;
657	printf "%s\n", $entry;
658	}
659
660	######################################################################################################
661
662	sub print_stats
663	{
664	printf "STATS: %s v%s completed in %d seconds", $progname, $version, (time-$script_start_time);
665	foreach my $key (sort keys %stats) {
666	printf ", %d %s",$stats{$key},$key;
667	}
668	printf "\n";
669	}
670
671	######################################################################################################
672	# descend a structure and clean up various things, including stripping
673	# leading/trailing spaces in strings, translations of html stuff etc
674	# -- taken & modified from Michael 'Immir' Smith's excellent tv_grab_au
675
676	sub cleanup {
677	my $x = shift;
678	if (ref $x eq "REF") { cleanup($_) }
679	elsif (ref $x eq "HASH") { cleanup(\$_) for values %$x }
680	elsif (ref $x eq "ARRAY") { cleanup(\$_) for @$x }
681	elsif (defined $$x) {
682	$$x =~ s/&(#(\d+)\|(.*?));/ $2 ? chr($2) : $amp{$3}\|\|' ' /eg;
683	$$x =~ s/[^\x20-\x7f]/ /g;
684	$$x =~ s/(^\s+\|\s+$)//g;
685	}
686	}
687
688	######################################################################################################
689
690	# strptime type date parsing - BUT - if no timezone is present, treat time as being in localtime
691	# rather than the various other perl implementation which treat it as being in UTC/GMT
692	sub parse_xmltv_date
693	{
694	my $datestring = shift;
695	my @t; # 0=sec,1=min,2=hour,3=day,4=month,5=year,6=wday,7=yday,8=isdst
696	my $tz_offset = 0;
697
698	# work out GMT offset - we only do this once
699	if (!$gmt_offset) {
700	my $tzstring = strftime("%z", localtime(time));
701
702	$gmt_offset = (6060) int(substr($tzstring,1,2)); # hr
703	$gmt_offset += (60 * int(substr($tzstring,3,2))); # min
704	$gmt_offset *= -1 if (substr($tzstring,0,1) eq "-"); # +/-
705	}
706
707	if ($datestring =~ /^(\d{4})(\d{2})(\d{2})(\d{2})(\d{2})/) {
708	($t[5],$t[4],$t[3],$t[2],$t[1],$t[0]) = (int($1)-1900,int($2)-1,int($3),int($4),int($5),0);
709	($t[6],$t[7],$t[8]) = (-1,-1,-1);
710
711	# if input data has a timezone offset, then offset by that
712	if ($datestring =~ /\+(\d{2})(\d{2})/) {
713	$tz_offset = $gmt_offset - (($1(6060)) + ($2*60));
714	} elsif ($datestring =~ /\-(\d{2})(\d{2})/) {
715	$tz_offset = $gmt_offset + (($1(6060)) + ($2*60));
716	}
717
718	my $e = mktime(@t);
719	return ($e+$tz_offset) if ($e > 1);
720	}
721	return undef;
722	}
723
724	######################################################################################################
725
726	sub get_station_close
727	{
728	my ($xmlid,$url) = @_;
729	my $tries = 0;
730	my $tree;
731	while (($tries < 3) && (!$tree)) {
732	$tries++;
733	&log("fetching (weekly) station close data for $xmlid (attempt $tries)");
734	my $data = &get_url($url);
735	$tree = HTML::TreeBuilder->new_from_content($data) if ((defined $data) && ($data ne ""));
736
737	if ((!$tree) && ($tries < 3)) {
738	# slow down ..
739	my $waittime = ($tries*10)+int(rand(10));
740	&log("failed to fetch $url, retrying in $waittime secs");
741	sleep($waittime);
742	$stats{slept_for} += $waittime;
743	}
744	}
745	if (!defined $tree) {
746	&log("failed to fetch $url after $tries attempts; skipping");
747	return;
748	}
749
750	my $to_skip = $opt_offset;
751	my $daynum = 0;
752	my $last_td_text;
753
754	foreach my $tree_td ($tree->look_down('_tag' => 'td')) {
755	if ($tree_td->as_text() =~ /^\.\.\.programs start at /) {
756	if (defined $last_td_text) {
757	if ($to_skip > 0) {
758	$to_skip--;
759	} else {
760	# 0=sec,1=min,2=hour,3=day,4=month,5=year,6=wday,7=yday,8=isdst
761	my @timeattr = localtime($starttime + ($daynum*86400));
762	$timeattr[0] = 0; # zero seconds
763
764	if ($last_td_text =~ /^(\d+):(\d+)(.)m/) {
765	$timeattr[2] = $1; # hour
766	$timeattr[1] = $2; # min
767
768	if ($3 eq "p") {
769	# pm
770	$timeattr[2] += 12 if ($timeattr[2] != 12);
771	}
772	my $found_time = mktime(@timeattr);
773
774	if ($3 eq "a") {
775	# am - must be tomorrow
776	if ($timeattr[2] == 12) {
777	$found_time += (126060); # 12:xx am
778	} else {
779	$found_time += (246060);
780	}
781	}
782
783	$daynum++;
784	$station_close_data[$daynum] = $found_time;
785
786	printf "station close time for day %d is %s\n",
787	$daynum, POSIX::strftime("%Y%m%d%H%M", localtime($found_time))
788	if $debug;
789	}
790	}
791	}
792	}
793	$last_td_text = $tree_td->as_text();
794	}
795	}
796
797	######################################################################################################

Note: See TracBrowser for help on using the browser.

Context Navigation

root/grabbers/abc_website @ 587

Download in other formats: