|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Spatie\Crawler\CrawlQueue; |
|
4
|
|
|
|
|
5
|
|
|
use Spatie\Crawler\CrawlUrl; |
|
6
|
|
|
use Spatie\Crawler\Exception\UrlNotFoundByIndex; |
|
7
|
|
|
|
|
8
|
|
|
class CollectionCrawlQueue implements RetryableCrawlQueue |
|
9
|
|
|
{ |
|
10
|
|
|
/** @var \Illuminate\Support\Collection|\Tightenco\Collect\Support\Collection */ |
|
11
|
|
|
protected $urls; |
|
12
|
|
|
|
|
13
|
|
|
/** @var \Illuminate\Support\Collection|\Tightenco\Collect\Support\Collection */ |
|
14
|
|
|
protected $pendingUrls; |
|
15
|
|
|
|
|
16
|
|
|
public function __construct() |
|
17
|
|
|
{ |
|
18
|
|
|
$this->urls = collect(); |
|
19
|
|
|
|
|
20
|
|
|
$this->pendingUrls = collect(); |
|
21
|
|
|
} |
|
22
|
|
|
|
|
23
|
|
|
public function getUrls() |
|
24
|
|
|
{ |
|
25
|
|
|
return $this->urls; |
|
26
|
|
|
} |
|
27
|
|
|
|
|
28
|
|
|
public function getPendingUrls() |
|
29
|
|
|
{ |
|
30
|
|
|
return $this->pendingUrls; |
|
31
|
|
|
} |
|
32
|
|
|
|
|
33
|
|
|
public function add(CrawlUrl $url): CrawlQueue |
|
34
|
|
|
{ |
|
35
|
|
|
if ($this->has($url)) { |
|
36
|
|
|
return $this; |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
$this->urls->push($url); |
|
40
|
|
|
|
|
41
|
|
|
$url->setId($this->urls->keys()->last()); |
|
42
|
|
|
$this->pendingUrls->push($url); |
|
43
|
|
|
|
|
44
|
|
|
return $this; |
|
45
|
|
|
} |
|
46
|
|
|
|
|
47
|
|
|
public function hasPendingUrls(): bool |
|
48
|
|
|
{ |
|
49
|
|
|
return (bool) $this->pendingUrls->count(); |
|
50
|
|
|
} |
|
51
|
|
|
|
|
52
|
|
|
/** |
|
53
|
|
|
* @param mixed $id |
|
54
|
|
|
* |
|
55
|
|
|
* @return \Spatie\Crawler\CrawlUrl|null |
|
56
|
|
|
*/ |
|
57
|
|
|
public function getUrlById($id): CrawlUrl |
|
58
|
|
|
{ |
|
59
|
|
|
if (! isset($this->urls->values()[$id])) { |
|
60
|
|
|
throw new UrlNotFoundByIndex("#{$id} crawl url not found in collection"); |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
return $this->urls->values()[$id]; |
|
64
|
|
|
} |
|
65
|
|
|
|
|
66
|
|
|
public function hasAlreadyBeenProcessed(CrawlUrl $url): bool |
|
67
|
|
|
{ |
|
68
|
|
|
return ! $this->contains($this->pendingUrls, $url) && $this->contains($this->urls, $url); |
|
69
|
|
|
} |
|
70
|
|
|
|
|
71
|
|
|
public function markAsProcessed(CrawlUrl $crawlUrl) |
|
72
|
|
|
{ |
|
73
|
|
|
$this->pendingUrls = $this->pendingUrls |
|
74
|
|
|
->reject(function (CrawlUrl $crawlUrlItem) use ($crawlUrl) { |
|
75
|
|
|
return (string) $crawlUrlItem->url === (string) $crawlUrl->url; |
|
76
|
|
|
}); |
|
77
|
|
|
} |
|
78
|
|
|
|
|
79
|
|
|
/** |
|
80
|
|
|
* @param CrawlUrl|\Psr\Http\Message\UriInterface $crawlUrl |
|
81
|
|
|
* |
|
82
|
|
|
* @return bool |
|
83
|
|
|
*/ |
|
84
|
|
|
public function has($crawlUrl): bool |
|
85
|
|
|
{ |
|
86
|
|
|
if (! $crawlUrl instanceof CrawlUrl) { |
|
87
|
|
|
$crawlUrl = CrawlUrl::create($crawlUrl); |
|
88
|
|
|
} |
|
89
|
|
|
|
|
90
|
|
|
if ($this->contains($this->urls, $crawlUrl)) { |
|
91
|
|
|
return true; |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
return false; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
/** @return \Spatie\Crawler\CrawlUrl|null */ |
|
98
|
|
|
public function getFirstPendingUrl() |
|
99
|
|
|
{ |
|
100
|
|
|
return $this->pendingUrls->first(); |
|
101
|
|
|
} |
|
102
|
|
|
|
|
103
|
|
|
/** |
|
104
|
|
|
* @param \Illuminate\Support\Collection|\Tightenco\Collect\Support\Collection $collection |
|
105
|
|
|
* @param \Spatie\Crawler\CrawlUrl $searchCrawlUrl |
|
106
|
|
|
* |
|
107
|
|
|
* @return bool |
|
108
|
|
|
*/ |
|
109
|
|
|
protected function contains($collection, CrawlUrl $searchCrawlUrl): bool |
|
110
|
|
|
{ |
|
111
|
|
|
foreach ($collection as $crawlUrl) { |
|
112
|
|
|
if ((string) $crawlUrl->url === (string) $searchCrawlUrl->url) { |
|
113
|
|
|
return true; |
|
114
|
|
|
} |
|
115
|
|
|
} |
|
116
|
|
|
|
|
117
|
|
|
return false; |
|
118
|
|
|
} |
|
119
|
|
|
|
|
120
|
|
|
/** |
|
121
|
|
|
* @param CrawlUrl $crawlUrl |
|
122
|
|
|
* |
|
123
|
|
|
* @return void |
|
124
|
|
|
*/ |
|
125
|
|
|
public function retry(CrawlUrl $crawlUrl) |
|
126
|
|
|
{ |
|
127
|
|
|
if (! $this->contains($this->urls, $crawlUrl)) { |
|
128
|
|
|
return; |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
if ($this->contains($this->pendingUrls, $crawlUrl)) { |
|
132
|
|
|
return; |
|
133
|
|
|
} |
|
134
|
|
|
|
|
135
|
|
|
$this->pendingUrls->push($crawlUrl); |
|
136
|
|
|
} |
|
137
|
|
|
} |
|
138
|
|
|
|
spatie /
crawler
