|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Spatie\Crawler\CrawlQueue; |
|
4
|
|
|
|
|
5
|
|
|
use Predis\Client; |
|
6
|
|
|
use Psr\Http\Message\UriInterface; |
|
7
|
|
|
use Spatie\Crawler\CrawlUrl; |
|
8
|
|
|
use Spatie\Crawler\Exception\InvalidUrl; |
|
9
|
|
|
use Spatie\Crawler\Exception\UrlNotFoundByIndex; |
|
10
|
|
|
|
|
11
|
|
|
/** |
|
12
|
|
|
* Implementation of CrawlQueue using Redis Hashes |
|
13
|
|
|
*/ |
|
14
|
|
|
class RedisCrawlQueue implements CrawlQueue |
|
15
|
|
|
{ |
|
16
|
|
|
// All known URLs, indexed by URL string. |
|
17
|
|
|
const URLS = 'urls'; |
|
18
|
|
|
// Pending URLs, indexed by URL string. |
|
19
|
|
|
const PENDING_URLS = 'pending'; |
|
20
|
|
|
|
|
21
|
|
|
/** |
|
22
|
|
|
* Redis Instance |
|
23
|
|
|
* @var \Predis\Client |
|
24
|
|
|
*/ |
|
25
|
|
|
private $redis; |
|
26
|
|
|
|
|
27
|
|
|
public function __construct(?Client $redis = null) |
|
28
|
|
|
{ |
|
29
|
|
|
$this->redis = $redis; |
|
30
|
|
|
if (is_null($redis)) { |
|
31
|
|
|
$this->redis = new Client(); |
|
32
|
|
|
} |
|
33
|
|
|
} |
|
34
|
|
|
|
|
35
|
|
|
public function add(CrawlUrl $url) : CrawlQueue |
|
36
|
|
|
{ |
|
37
|
|
|
$urlString = (string) $url->url; |
|
38
|
|
|
|
|
39
|
|
|
if (!$this->has($urlString)) { |
|
40
|
|
|
$url->setId($urlString); |
|
41
|
|
|
|
|
42
|
|
|
$this->redis->hset(self::URLS, $urlString, serialize($url)); |
|
43
|
|
|
$this->redis->hset(self::PENDING_URLS, $urlString, serialize($url)); |
|
44
|
|
|
} |
|
45
|
|
|
|
|
46
|
|
|
return $this; |
|
47
|
|
|
} |
|
48
|
|
|
|
|
49
|
|
|
public function has($crawlUrl) : bool |
|
50
|
|
|
{ |
|
51
|
|
|
if ($crawlUrl instanceof CrawlUrl) { |
|
52
|
|
|
$url = (string) $crawlUrl->url; |
|
53
|
|
|
} elseif ($crawlUrl instanceof UriInterface) { |
|
54
|
|
|
$url = (string) $crawlUrl; |
|
55
|
|
|
} elseif (is_string($crawlUrl)) { |
|
56
|
|
|
$url = $crawlUrl; |
|
57
|
|
|
} else { |
|
58
|
|
|
throw InvalidUrl::unexpectedType($crawlUrl); |
|
59
|
|
|
} |
|
60
|
|
|
|
|
61
|
|
|
return (bool) $this->redis->hexists(self::URLS, $url); |
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
public function hasPendingUrls() : bool |
|
65
|
|
|
{ |
|
66
|
|
|
return (bool) $this->redis->hlen(self::PENDING_URLS); |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
public function getUrlById($id) : CrawlUrl |
|
70
|
|
|
{ |
|
71
|
|
|
if (!$this->has($id)) { |
|
72
|
|
|
throw new UrlNotFoundByIndex("Crawl url {$id} not found in hashes."); |
|
73
|
|
|
} |
|
74
|
|
|
return unserialize($this->redis->hget(self::URLS, $id)); |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
public function getFirstPendingUrl() : ?CrawlUrl |
|
78
|
|
|
{ |
|
79
|
|
|
$keys = $this->redis->hkeys(self::PENDING_URLS); |
|
80
|
|
|
|
|
81
|
|
|
foreach ($keys as $key) { |
|
82
|
|
|
return unserialize($this->redis->hget(self::PENDING_URLS, $key)); |
|
83
|
|
|
} |
|
84
|
|
|
|
|
85
|
|
|
return null; |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
public function hasAlreadyBeenProcessed(CrawlUrl $url) : bool |
|
89
|
|
|
{ |
|
90
|
|
|
$url = (string) $url->url; |
|
91
|
|
|
|
|
92
|
|
|
if ($this->redis->hexists(self::PENDING_URLS, $url)) { |
|
93
|
|
|
return false; |
|
94
|
|
|
} |
|
95
|
|
|
|
|
96
|
|
|
if ($this->redis->hexists(self::URLS, $url)) { |
|
97
|
|
|
return true; |
|
98
|
|
|
} |
|
99
|
|
|
|
|
100
|
|
|
return false; |
|
101
|
|
|
} |
|
102
|
|
|
|
|
103
|
|
|
public function markAsProcessed(CrawlUrl $crawlUrl) |
|
104
|
|
|
{ |
|
105
|
|
|
$this->redis->hdel(self::PENDING_URLS, (string) $crawlUrl->url); |
|
106
|
|
|
} |
|
107
|
|
|
} |
|
108
|
|
|
|
spatie /
crawler
