|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Comrade42\PhpBBParser\Command; |
|
4
|
|
|
|
|
5
|
|
|
use Comrade42\PhpBBParser\Helper\BBCodes\PCREParser; |
|
6
|
|
|
use Goutte\Client; |
|
7
|
|
|
use Symfony\Component\BrowserKit\Cookie; |
|
8
|
|
|
use Symfony\Component\Console\Input\InputInterface; |
|
9
|
|
|
use Symfony\Component\Console\Output\OutputInterface; |
|
10
|
|
|
use Symfony\Component\DomCrawler\Crawler; |
|
11
|
|
|
|
|
12
|
|
|
/** |
|
13
|
|
|
* Class ParseTopicsCommand |
|
14
|
|
|
* @package Comrade42\PhpBBParser\Command |
|
15
|
|
|
*/ |
|
16
|
|
|
class ParseTopicsCommand extends ContainerAwareCommand |
|
17
|
|
|
{ |
|
18
|
|
|
const TOPICS_PER_PAGE = 50; |
|
19
|
|
|
const POSTS_PER_PAGE = 15; |
|
20
|
|
|
|
|
21
|
|
|
protected function configure() |
|
22
|
|
|
{ |
|
23
|
|
|
$this->setName('parse:topics')->setDescription('Parse topics'); |
|
24
|
|
|
} |
|
25
|
|
|
|
|
26
|
|
|
protected function execute(InputInterface $input, OutputInterface $output) |
|
27
|
|
|
{ |
|
28
|
|
|
/** @var \Symfony\Component\Console\Helper\DialogHelper $dialog */ |
|
29
|
|
|
$dialog = $this->getHelperSet()->get('dialog'); |
|
30
|
|
|
/** @var \Doctrine\ORM\EntityManager $entityManager */ |
|
31
|
|
|
$entityManager = $this->container->get('doctrine'); |
|
32
|
|
|
/** @var \Comrade42\PhpBBParser\Bridge\BridgeInterface $entityBridge */ |
|
33
|
|
|
$entityBridge = $this->container->get('bridge'); |
|
34
|
|
|
|
|
35
|
|
|
$choices = array(); |
|
36
|
|
|
foreach ($entityBridge->getForumList($entityManager) as $forum) |
|
37
|
|
|
{ |
|
38
|
|
|
$choices[$forum->getId()] = $forum->getTitle(); |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
|
|
ksort($choices); |
|
42
|
|
|
$forumId = $dialog->select($output, 'Please select forum for parsing:', $choices); |
|
43
|
|
|
|
|
44
|
|
|
$client = new Client(); |
|
45
|
|
|
$client->getCookieJar()->set(new Cookie( |
|
46
|
|
|
$this->container->getParameter('fa_sid_cookie_name'), |
|
47
|
|
|
$this->container->getParameter('fa_sid_cookie_value') |
|
48
|
|
|
)); |
|
49
|
|
|
|
|
50
|
|
|
$baseUrl = rtrim($this->container->getParameter('forum_url'), '/'); |
|
51
|
|
|
|
|
52
|
|
|
for ($offset = 0; ; $offset += static::TOPICS_PER_PAGE) |
|
|
|
|
|
|
53
|
|
|
{ |
|
54
|
|
|
$url = "{$baseUrl}/f{$forumId}p{$offset}-forum"; |
|
55
|
|
|
$output->write(str_pad("<info>⇒ HTTP GET: {$url}</info> ", 80)); |
|
56
|
|
|
|
|
57
|
|
|
$crawler = $client->request('GET', $url); |
|
58
|
|
|
|
|
59
|
|
|
$status = $client->getInternalResponse()->getStatus(); |
|
60
|
|
|
$output->writeln("<info>[{$status}]</info>"); |
|
61
|
|
|
|
|
62
|
|
|
if ($status != 200) |
|
63
|
|
|
{ |
|
64
|
|
|
$offset -= static::TOPICS_PER_PAGE; |
|
65
|
|
|
sleep(1); |
|
66
|
|
|
continue; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
$crawler = $crawler->filter('#main-content ul.topics li.row'); |
|
70
|
|
|
$crawler->each(function (Crawler $node, $index) use ($output, $entityManager, $entityBridge, $client, $baseUrl, $forumId) |
|
71
|
|
|
{ |
|
72
|
|
|
$link = $node->filter('dd.dterm a.topictitle'); |
|
73
|
|
|
$topicId = intval(substr($link->attr('href'), 2, -6)); |
|
74
|
|
|
$title = $link->text(); |
|
75
|
|
|
|
|
76
|
|
|
$link = $node->filter('dd.dterm span.span-tab a'); |
|
77
|
|
|
if ($link->count() == 1) $authorId = intval(substr($link->attr('href'), 2)); |
|
78
|
|
|
else $authorId = 0; |
|
79
|
|
|
|
|
80
|
|
|
preg_match("/background-image:url\('(.+)'\);/iU", $node->filter('dl.icon')->attr('style'), $matches); |
|
81
|
|
|
$matches = explode('/', $matches[1]); |
|
82
|
|
|
$icon = end($matches); |
|
83
|
|
|
|
|
84
|
|
|
switch ($icon) |
|
85
|
|
|
{ |
|
86
|
|
|
case 'topic_unread_locked.gif': |
|
87
|
|
|
case 'topic_read_locked.gif': |
|
88
|
|
|
$isSticky = false; |
|
89
|
|
|
$isLocked = true; |
|
90
|
|
|
break; |
|
91
|
|
|
|
|
92
|
|
|
case 'announce_read.gif': |
|
93
|
|
|
case 'folder_announce.gif': |
|
94
|
|
|
case 'sticky_read.gif': |
|
95
|
|
|
$isSticky = true; |
|
96
|
|
|
$isLocked = false; |
|
97
|
|
|
break; |
|
98
|
|
|
|
|
99
|
|
|
default: |
|
100
|
|
|
$isSticky = false; |
|
101
|
|
|
$isLocked = false; |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
|
|
$links = $node->filter('dd.lastpost a'); |
|
105
|
|
|
if ($links->count() == 2) { |
|
106
|
|
|
$memberUpdatedId = intval(substr($links->first()->attr('href'), 2)); |
|
107
|
|
|
$lastMessageHref = $links->last()->attr('href'); |
|
108
|
|
|
} else { |
|
109
|
|
|
$memberUpdatedId = 0; |
|
110
|
|
|
$lastMessageHref = $links->first()->attr('href'); |
|
111
|
|
|
} |
|
112
|
|
|
|
|
113
|
|
|
$topic = $entityBridge->getTopicEntity($entityManager, $topicId); |
|
114
|
|
|
$topic->setForumId($forumId) |
|
115
|
|
|
->setMemberStartedId($authorId) |
|
116
|
|
|
->setMemberUpdatedId($memberUpdatedId) |
|
117
|
|
|
->setLastMessageId(intval(substr($lastMessageHref, strrpos($lastMessageHref, '#') + 1))) |
|
118
|
|
|
->setTitle($title) |
|
119
|
|
|
->setIsSticky($isSticky) |
|
120
|
|
|
->setIsLocked($isLocked) |
|
121
|
|
|
->setRepliesNumber(intval($node->filter('dd.posts')->text())) |
|
122
|
|
|
->setViewsNumber(intval($node->filter('dd.views')->text())) |
|
123
|
|
|
->setOrder($index + 1); |
|
124
|
|
|
|
|
125
|
|
|
for ($offset = 0; ; $offset += ParseTopicsCommand::POSTS_PER_PAGE) |
|
126
|
|
|
{ |
|
127
|
|
|
$url = "{$baseUrl}/t{$topicId}p{$offset}-topic"; |
|
128
|
|
|
$output->write(str_pad("<info>⇒ HTTP GET: {$url}</info> ", 80)); |
|
129
|
|
|
|
|
130
|
|
|
$crawler = $client->request('GET', $url); |
|
131
|
|
|
|
|
132
|
|
|
$status = $client->getInternalResponse()->getStatus(); |
|
133
|
|
|
$output->writeln("<info>[{$status}]</info>"); |
|
134
|
|
|
|
|
135
|
|
|
if ($status != 200) |
|
136
|
|
|
{ |
|
137
|
|
|
$offset -= ParseTopicsCommand::POSTS_PER_PAGE; |
|
138
|
|
|
sleep(1); |
|
139
|
|
|
continue; |
|
140
|
|
|
} |
|
141
|
|
|
|
|
142
|
|
|
/** @var \Symfony\Component\DomCrawler\Crawler $crawler */ |
|
143
|
|
|
$crawler = $crawler->filter('#main-content div.post')->reduce(function (Crawler $node) |
|
144
|
|
|
{ |
|
145
|
|
|
return is_numeric(substr($node->attr('id'), 1)); |
|
146
|
|
|
}); |
|
147
|
|
|
|
|
148
|
|
|
if ($offset == 0) |
|
149
|
|
|
{ |
|
150
|
|
|
$topic->setFirstMessageId(intval(substr($crawler->first()->attr('id'), 1))); |
|
151
|
|
|
$entityManager->persist($topic); |
|
152
|
|
|
} |
|
153
|
|
|
|
|
154
|
|
|
$crawler->each(function (Crawler $node) use ($output, $entityManager, $entityBridge, $forumId, $topicId) |
|
155
|
|
|
{ |
|
156
|
|
|
$postId = substr($node->attr('id'), 1); |
|
157
|
|
|
|
|
158
|
|
|
$contentRaw = $node->filter('div.postbody div.content div')->html(); |
|
159
|
|
|
$contentParsed = trim(PCREParser::parseAll($contentRaw)); |
|
160
|
|
|
$contentClean = strip_tags($contentParsed); |
|
161
|
|
|
|
|
162
|
|
|
if ($contentClean != $contentParsed) { |
|
163
|
|
|
$output->writeln("\t- parsing problems with post #{$postId}"); |
|
164
|
|
|
} |
|
165
|
|
|
|
|
166
|
|
|
$link = $node->filter('div.postbody p.author a'); |
|
167
|
|
|
$authorId = intval(substr($link->attr('href'), 2)); |
|
168
|
|
|
$authorName = trim($link->text()); |
|
169
|
|
|
|
|
170
|
|
|
$text = trim($node->filter('div.postbody p.author')->text(), " \t\n\r\0\x0B\xC2\xA0"); |
|
171
|
|
|
$text = str_replace($authorName . ' в ', '', $text); |
|
172
|
|
|
$dateTime = new \DateTime($text); |
|
173
|
|
|
|
|
174
|
|
|
$post = $entityBridge->getPostEntity($entityManager, $postId); |
|
175
|
|
|
$post->setForumId($forumId) |
|
176
|
|
|
->setTopicId($topicId) |
|
177
|
|
|
->setAuthorId($authorId) |
|
178
|
|
|
->setAuthorName($authorName) |
|
179
|
|
|
->setCreateDate($dateTime) |
|
180
|
|
|
->setSubject(trim($node->filter('div.postbody h2.topic-title')->text())) |
|
181
|
|
|
->setContent($contentParsed); |
|
182
|
|
|
|
|
183
|
|
|
$entityManager->persist($post); |
|
184
|
|
|
}); |
|
185
|
|
|
|
|
186
|
|
|
if ($crawler->count() < ParseTopicsCommand::POSTS_PER_PAGE) break; |
|
187
|
|
|
} |
|
188
|
|
|
|
|
189
|
|
|
$entityManager->flush(); |
|
190
|
|
|
}); |
|
191
|
|
|
|
|
192
|
|
|
if ($crawler->count() < static::TOPICS_PER_PAGE) break; |
|
193
|
|
|
} |
|
194
|
|
|
} |
|
195
|
|
|
} |
|
196
|
|
|
|
Comrade42 /
phpBB-Parser
Loading history...