|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use Spatie\Robots\RobotsTxt; |
|
6
|
|
|
use League\Csv\Reader; |
|
7
|
|
|
|
|
8
|
|
|
class CrawlerContinue extends Crawler |
|
9
|
|
|
{ |
|
10
|
3 |
|
public function __construct(string $id, string $dataDirectoryBasePath = null) |
|
11
|
|
|
{ |
|
12
|
3 |
|
$this->id = trim($id); |
|
13
|
3 |
|
$this->initDataDirectory($dataDirectoryBasePath); |
|
14
|
|
|
|
|
15
|
3 |
|
$configFilePath = $this->getDataFolder().'/config.json'; |
|
16
|
3 |
|
if (!file_exists($configFilePath)) { |
|
17
|
|
|
throw new \Exception('previous crawl not found (config.json)'); |
|
18
|
|
|
} |
|
19
|
3 |
|
$config = json_decode(file_get_contents($configFilePath), true); |
|
20
|
|
|
|
|
21
|
3 |
|
$this->ignore = new RobotsTxt($config['ignore']); |
|
22
|
3 |
|
$this->limit = $config['limit']; |
|
23
|
3 |
|
$this->userAgent = $config['userAgent']; |
|
24
|
3 |
|
$this->wait = $config['wait']; |
|
25
|
3 |
|
$this->base = $config['base']; |
|
26
|
|
|
|
|
27
|
3 |
|
$this->recorder = new Recorder($this->getDataFolder(), (int) $config['cacheMethod']); |
|
28
|
3 |
|
$this->loadFromPreviousCrawl($config['startUrl']); |
|
29
|
3 |
|
} |
|
30
|
|
|
|
|
31
|
3 |
|
protected function loadFromPreviousCrawl(string $startUrl) |
|
32
|
|
|
{ |
|
33
|
3 |
|
$resultFilePath = $this->getDataFolder().'/index.csv'; |
|
34
|
3 |
|
if (!file_exists($resultFilePath)) { |
|
35
|
|
|
throw new \Exception('previous crawl not found (index.csv)'); |
|
36
|
|
|
} |
|
37
|
|
|
|
|
38
|
3 |
|
$csv = Reader::createFromPath($resultFilePath, 'r'); |
|
39
|
3 |
|
$csv->setHeaderOffset(0); |
|
40
|
|
|
|
|
41
|
3 |
|
$records = $csv->getRecords(); |
|
42
|
3 |
|
foreach ($records as $r) { |
|
43
|
3 |
|
$this->urls[$r['uri']] = new Url($this->base.$r['uri'], 0); |
|
44
|
3 |
|
foreach ($r as $k => $v) { |
|
45
|
3 |
|
if ('can_be_crawled' == $k && !empty($v)) { |
|
46
|
3 |
|
$v = (bool) $v; |
|
47
|
|
|
} |
|
48
|
3 |
|
$this->urls[$r['uri']]->$k = $v; |
|
49
|
|
|
} |
|
50
|
3 |
|
if (!empty($r['can_be_crawled'])) { |
|
51
|
3 |
|
++$this->counter; |
|
52
|
|
|
} |
|
53
|
|
|
} |
|
54
|
|
|
|
|
55
|
3 |
|
$this->currentClick = $r['click'] ?? 0; |
|
56
|
|
|
|
|
57
|
3 |
|
return $startUrl; |
|
58
|
|
|
} |
|
59
|
|
|
} |
|
60
|
|
|
|
PiedWeb /
SeoPocketCrawler
