|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Spatie\Crawler\CrawlQueue; |
|
4
|
|
|
|
|
5
|
|
|
use Psr\Http\Message\UriInterface; |
|
6
|
|
|
use Spatie\Crawler\CrawlUrl; |
|
7
|
|
|
use Spatie\Crawler\Exception\InvalidUrl; |
|
8
|
|
|
use Spatie\Crawler\Exception\UrlNotFoundByIndex; |
|
9
|
|
|
|
|
10
|
|
|
class ArrayCrawlQueue implements CrawlQueue |
|
11
|
|
|
{ |
|
12
|
|
|
/** |
|
13
|
|
|
* All known URLs, indexed by URL string. |
|
14
|
|
|
* |
|
15
|
|
|
* @var CrawlUrl[] |
|
16
|
|
|
*/ |
|
17
|
|
|
protected $urls = []; |
|
18
|
|
|
|
|
19
|
|
|
/** |
|
20
|
|
|
* Pending URLs, indexed by URL string. |
|
21
|
|
|
* |
|
22
|
|
|
* @var CrawlUrl[] |
|
23
|
|
|
*/ |
|
24
|
|
|
protected $pendingUrls = []; |
|
25
|
|
|
|
|
26
|
|
|
public function add(CrawlUrl $url) : CrawlQueue |
|
27
|
|
|
{ |
|
28
|
|
|
$urlString = (string) $url->url; |
|
29
|
|
|
|
|
30
|
|
|
if (! isset($this->urls[$urlString])) { |
|
31
|
|
|
$url->setId($urlString); |
|
32
|
|
|
|
|
33
|
|
|
$this->urls[$urlString] = $url; |
|
34
|
|
|
$this->pendingUrls[$urlString] = $url; |
|
35
|
|
|
} |
|
36
|
|
|
|
|
37
|
|
|
return $this; |
|
38
|
|
|
} |
|
39
|
|
|
|
|
40
|
|
|
public function hasPendingUrls() : bool |
|
41
|
|
|
{ |
|
42
|
|
|
return (bool) $this->pendingUrls; |
|
43
|
|
|
} |
|
44
|
|
|
|
|
45
|
|
|
public function getUrlById($id) : CrawlUrl |
|
46
|
|
|
{ |
|
47
|
|
|
if (! isset($this->urls[$id])) { |
|
48
|
|
|
throw new UrlNotFoundByIndex("Crawl url {$id} not found in collection."); |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
return $this->urls[$id]; |
|
52
|
|
|
} |
|
53
|
|
|
|
|
54
|
|
|
public function hasAlreadyBeenProcessed(CrawlUrl $url) : bool |
|
55
|
|
|
{ |
|
56
|
|
|
$url = (string) $url->url; |
|
57
|
|
|
|
|
58
|
|
|
if (isset($this->pendingUrls[$url])) { |
|
59
|
|
|
return false; |
|
60
|
|
|
} |
|
61
|
|
|
|
|
62
|
|
|
if (isset($this->urls[$url])) { |
|
63
|
|
|
return true; |
|
64
|
|
|
} |
|
65
|
|
|
|
|
66
|
|
|
return false; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
public function markAsProcessed(CrawlUrl $crawlUrl) |
|
70
|
|
|
{ |
|
71
|
|
|
$url = (string) $crawlUrl->url; |
|
72
|
|
|
|
|
73
|
|
|
unset($this->pendingUrls[$url]); |
|
74
|
|
|
} |
|
75
|
|
|
|
|
76
|
|
|
/** |
|
77
|
|
|
* @param CrawlUrl|UriInterface $crawlUrl |
|
78
|
|
|
* |
|
79
|
|
|
* @return bool |
|
80
|
|
|
*/ |
|
81
|
|
|
public function has($crawlUrl) : bool |
|
82
|
|
|
{ |
|
83
|
|
|
if ($crawlUrl instanceof CrawlUrl) { |
|
84
|
|
|
$url = (string) $crawlUrl->url; |
|
85
|
|
|
} elseif ($crawlUrl instanceof UriInterface) { |
|
86
|
|
|
$url = (string) $crawlUrl; |
|
87
|
|
|
} else { |
|
88
|
|
|
throw InvalidUrl::unexpectedType($crawlUrl); |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
|
|
return isset($this->urls[$url]); |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
public function getFirstPendingUrl() : ?CrawlUrl |
|
95
|
|
|
{ |
|
96
|
|
|
foreach ($this->pendingUrls as $pendingUrl) { |
|
97
|
|
|
return $pendingUrl; |
|
98
|
|
|
} |
|
99
|
|
|
|
|
100
|
|
|
return null; |
|
101
|
|
|
} |
|
102
|
|
|
} |
|
103
|
|
|
|