|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Zenstruck\CacheBundle\Url; |
|
4
|
|
|
|
|
5
|
|
|
use Http\Client\HttpClient; |
|
6
|
|
|
use Http\Message\MessageFactory; |
|
7
|
|
|
use Symfony\Component\DomCrawler\Crawler as DomCrawler; |
|
8
|
|
|
|
|
9
|
|
|
/** |
|
10
|
|
|
* @author Kevin Bond <[email protected]> |
|
11
|
|
|
*/ |
|
12
|
|
|
class SitemapUrlProvider implements UrlProvider |
|
13
|
|
|
{ |
|
14
|
|
|
private $sitemaps; |
|
15
|
|
|
private $httpClient; |
|
16
|
|
|
private $messageFactory; |
|
17
|
|
|
private $urls; |
|
18
|
|
|
|
|
19
|
|
|
/** |
|
20
|
|
|
* @param array $sitemaps |
|
21
|
|
|
* @param HttpClient $httpClient |
|
22
|
|
|
* @param MessageFactory $messageFactory |
|
23
|
|
|
*/ |
|
24
|
14 |
|
public function __construct(array $sitemaps, HttpClient $httpClient, MessageFactory $messageFactory) |
|
25
|
|
|
{ |
|
26
|
7 |
|
if (!class_exists('Symfony\\Component\\DomCrawler\\Crawler')) { |
|
27
|
|
|
throw new \RuntimeException('symfony/dom-crawler and symfony/css-selector must be installed to use SitemapUrlProvider.'); |
|
28
|
|
|
} |
|
29
|
14 |
|
$this->sitemaps = $sitemaps; |
|
30
|
14 |
|
$this->httpClient = $httpClient; |
|
31
|
14 |
|
$this->messageFactory = $messageFactory; |
|
32
|
7 |
|
} |
|
33
|
|
|
|
|
34
|
|
|
/** |
|
35
|
|
|
* {@inheritdoc} |
|
36
|
|
|
*/ |
|
37
|
7 |
|
public function count() |
|
38
|
|
|
{ |
|
39
|
7 |
|
return count($this->getUrls()); |
|
40
|
|
|
} |
|
41
|
|
|
|
|
42
|
|
|
/** |
|
43
|
|
|
* {@inheritdoc} |
|
44
|
|
|
*/ |
|
45
|
14 |
|
public function getUrls() |
|
46
|
|
|
{ |
|
47
|
7 |
|
if (null !== $this->urls) { |
|
48
|
7 |
|
return $this->urls; |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
14 |
|
$urls = []; |
|
52
|
|
|
|
|
53
|
14 |
|
foreach ($this->sitemaps as $sitemap) { |
|
54
|
7 |
|
$urls = array_merge($urls, $this->getUrlsForSitemapUrl($sitemap)); |
|
55
|
7 |
|
} |
|
56
|
|
|
|
|
57
|
14 |
|
return $this->urls = $urls; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
|
|
/** |
|
61
|
|
|
* @param string $sitemap |
|
62
|
|
|
* |
|
63
|
|
|
* @return array |
|
64
|
|
|
*/ |
|
65
|
10 |
|
private function getUrlsForSitemapUrl($sitemap) |
|
66
|
|
|
{ |
|
67
|
7 |
|
$path = parse_url($sitemap, PHP_URL_PATH); |
|
68
|
|
|
|
|
69
|
10 |
|
if (null === $path || '/' === trim($path)) { |
|
70
|
5 |
|
return $this->tryDefaultSitemapUrls($sitemap); |
|
71
|
|
|
} |
|
72
|
|
|
|
|
73
|
4 |
|
return $this->parseUrl($sitemap); |
|
74
|
|
|
} |
|
75
|
|
|
|
|
76
|
|
|
/** |
|
77
|
|
|
* @param string $host |
|
78
|
|
|
* |
|
79
|
|
|
* @return array |
|
80
|
|
|
*/ |
|
81
|
10 |
|
private function tryDefaultSitemapUrls($host) |
|
82
|
|
|
{ |
|
83
|
|
|
// try default sitemap_index.xml |
|
84
|
5 |
|
$urls = $this->parseUrl($this->addPathToHost('sitemap_index.xml', $host)); |
|
85
|
|
|
|
|
86
|
10 |
|
if (empty($urls)) { |
|
87
|
|
|
// try default sitemap.xml |
|
88
|
3 |
|
$urls = $this->parseUrl($this->addPathToHost('sitemap.xml', $host)); |
|
89
|
3 |
|
} |
|
90
|
|
|
|
|
91
|
10 |
|
return $urls; |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
/** |
|
95
|
|
|
* @param string $url |
|
96
|
|
|
* |
|
97
|
|
|
* @return array |
|
98
|
|
|
*/ |
|
99
|
10 |
|
private function parseUrl($url) |
|
100
|
|
|
{ |
|
101
|
7 |
|
$response = $this->httpClient->sendRequest($this->messageFactory->createRequest('GET', $url)); |
|
102
|
|
|
|
|
103
|
7 |
|
if (200 !== $response->getStatusCode()) { |
|
104
|
7 |
|
return []; |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
6 |
|
$body = (string) $response->getBody(); |
|
108
|
|
|
|
|
109
|
6 |
|
if (false !== strpos($body, '<sitemapindex')) { |
|
110
|
3 |
|
return $this->parseSitemapIndex($body); |
|
111
|
|
|
} |
|
112
|
|
|
|
|
113
|
6 |
|
return $this->getLocEntries($body); |
|
114
|
|
|
} |
|
115
|
|
|
|
|
116
|
|
|
/** |
|
117
|
|
|
* @param string $body |
|
118
|
|
|
* |
|
119
|
|
|
* @return array |
|
120
|
|
|
*/ |
|
121
|
5 |
|
private function parseSitemapIndex($body) |
|
122
|
|
|
{ |
|
123
|
5 |
|
$urls = []; |
|
124
|
|
|
|
|
125
|
5 |
|
foreach ($this->getLocEntries($body) as $entry) { |
|
126
|
3 |
|
$urls = array_merge($urls, $this->getUrlsForSitemapUrl($entry)); |
|
127
|
3 |
|
} |
|
128
|
|
|
|
|
129
|
5 |
|
return $urls; |
|
130
|
|
|
} |
|
131
|
|
|
|
|
132
|
|
|
/** |
|
133
|
|
|
* @param string $body |
|
134
|
|
|
* |
|
135
|
|
|
* @return array |
|
136
|
|
|
*/ |
|
137
|
9 |
|
private function getLocEntries($body) |
|
138
|
|
|
{ |
|
139
|
6 |
|
$crawler = new DomCrawler($body); |
|
140
|
9 |
|
$entries = []; |
|
141
|
9 |
|
$filter = 'loc'; |
|
142
|
|
|
|
|
143
|
|
|
// check for namespaces |
|
144
|
6 |
|
if (preg_match('/xmlns:/', $body)) { |
|
145
|
2 |
|
$filter = 'default|loc'; |
|
146
|
1 |
|
} |
|
147
|
|
|
|
|
148
|
9 |
|
foreach ($crawler->filter($filter) as $node) { |
|
149
|
6 |
|
$entries[] = $node->nodeValue; |
|
150
|
6 |
|
} |
|
151
|
|
|
|
|
152
|
9 |
|
return $entries; |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
/** |
|
156
|
|
|
* @param string $path |
|
157
|
|
|
* @param string $host |
|
158
|
|
|
* |
|
159
|
|
|
* @return string |
|
160
|
|
|
*/ |
|
161
|
5 |
|
private function addPathToHost($path, $host) |
|
162
|
|
|
{ |
|
163
|
5 |
|
return sprintf('%s/%s', trim($host, '/'), ltrim($path, '/')); |
|
164
|
|
|
} |
|
165
|
|
|
} |
|
166
|
|
|
|
kbond /
ZenstruckCacheBundle
