|
1
|
|
|
<?php |
|
2
|
|
|
declare(strict_types=1); |
|
3
|
|
|
namespace AOE\Crawler\CrawlStrategy; |
|
4
|
|
|
|
|
5
|
|
|
/* |
|
6
|
|
|
* This file is part of the TYPO3 Crawler Extension. |
|
7
|
|
|
* |
|
8
|
|
|
* It is free software; you can redistribute it and/or modify it under |
|
9
|
|
|
* the terms of the GNU General Public License, either version 2 |
|
10
|
|
|
* of the License, or any later version. |
|
11
|
|
|
* |
|
12
|
|
|
* For the full copyright and license information, please read the |
|
13
|
|
|
* LICENSE.txt file that was distributed with this source code. |
|
14
|
|
|
* |
|
15
|
|
|
* The TYPO3 project - inspiring people to share! |
|
16
|
|
|
*/ |
|
17
|
|
|
|
|
18
|
|
|
use AOE\Crawler\Configuration\ExtensionConfigurationProvider; |
|
19
|
|
|
use Psr\Http\Message\UriInterface; |
|
20
|
|
|
use Psr\Log\LoggerAwareInterface; |
|
21
|
|
|
use Psr\Log\LoggerAwareTrait; |
|
22
|
|
|
use TYPO3\CMS\Core\Core\Environment; |
|
23
|
|
|
use TYPO3\CMS\Core\Utility\ExtensionManagementUtility; |
|
24
|
|
|
use TYPO3\CMS\Core\Utility\GeneralUtility; |
|
25
|
|
|
|
|
26
|
|
|
/** |
|
27
|
|
|
* Executes another process via shell_exec() to include cli/bootstrap.php which in turn |
|
28
|
|
|
* includes the index.php for frontend. |
|
29
|
|
|
*/ |
|
30
|
|
|
class SubProcessExecutionStrategy implements LoggerAwareInterface |
|
31
|
|
|
{ |
|
32
|
|
|
use LoggerAwareTrait; |
|
33
|
|
|
|
|
34
|
|
|
/** |
|
35
|
|
|
* @var array |
|
36
|
|
|
*/ |
|
37
|
|
|
protected $extensionSettings; |
|
38
|
|
|
|
|
39
|
1 |
|
public function __construct(ExtensionConfigurationProvider $configurationProvider = null) |
|
40
|
|
|
{ |
|
41
|
1 |
|
$configurationProvider = $configurationProvider ?? GeneralUtility::makeInstance(ExtensionConfigurationProvider::class); |
|
42
|
1 |
|
$settings = $configurationProvider->getExtensionConfiguration(); |
|
43
|
1 |
|
$this->extensionSettings = is_array($settings) ? $settings : []; |
|
44
|
1 |
|
} |
|
45
|
|
|
|
|
46
|
|
|
/** |
|
47
|
|
|
* Fetches a URL by calling a shell script. |
|
48
|
|
|
* |
|
49
|
|
|
* @param UriInterface $url |
|
50
|
|
|
* @param string $crawlerId |
|
51
|
|
|
* @return array|bool|mixed |
|
52
|
|
|
*/ |
|
53
|
|
|
public function fetchUrlContents(UriInterface $url, string $crawlerId) |
|
54
|
|
|
{ |
|
55
|
|
|
$url = (string)$url; |
|
56
|
|
|
$parsedUrl = parse_url($url); |
|
57
|
|
|
|
|
58
|
|
|
if ($parsedUrl === false) { |
|
59
|
|
|
$this->logger->debug( |
|
60
|
|
|
sprintf('Could not parse_url() for string "%s"', $url), |
|
61
|
|
|
['crawlerId' => $crawlerId] |
|
62
|
|
|
); |
|
63
|
|
|
return false; |
|
64
|
|
|
} |
|
65
|
|
|
|
|
66
|
|
|
if (!in_array($parsedUrl['scheme'], ['','http','https'])) { |
|
67
|
|
|
$this->logger->debug( |
|
68
|
|
|
sprintf('Scheme does not match for url "%s"', $url), |
|
69
|
|
|
['crawlerId' => $crawlerId] |
|
70
|
|
|
); |
|
71
|
|
|
return false; |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
|
|
if (!is_array($parsedUrl)) { |
|
75
|
|
|
return []; |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
$requestHeaders = $this->buildRequestHeaders($parsedUrl, $crawlerId); |
|
79
|
|
|
|
|
80
|
|
|
$cmd = escapeshellcmd($this->extensionSettings['phpPath']); |
|
81
|
|
|
$cmd .= ' '; |
|
82
|
|
|
$cmd .= escapeshellarg(ExtensionManagementUtility::extPath('crawler') . 'cli/bootstrap.php'); |
|
83
|
|
|
$cmd .= ' '; |
|
84
|
|
|
$cmd .= escapeshellarg($this->getFrontendBasePath()); |
|
85
|
|
|
$cmd .= ' '; |
|
86
|
|
|
$cmd .= escapeshellarg($url); |
|
87
|
|
|
$cmd .= ' '; |
|
88
|
|
|
$cmd .= escapeshellarg(base64_encode(serialize($requestHeaders))); |
|
89
|
|
|
|
|
90
|
|
|
$startTime = microtime(true); |
|
91
|
|
|
$content = $this->executeShellCommand($cmd); |
|
92
|
|
|
$this->logger->info($url . ' ' . (microtime(true) - $startTime)); |
|
93
|
|
|
|
|
94
|
|
|
return unserialize($content); |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
protected function buildRequestHeaders(array $url, string $crawlerId) |
|
98
|
|
|
{ |
|
99
|
|
|
$reqHeaders = []; |
|
100
|
|
|
$reqHeaders[] = 'GET ' . $url['path'] . ($url['query'] ? '?' . $url['query'] : '') . ' HTTP/1.0'; |
|
101
|
|
|
$reqHeaders[] = 'Host: ' . $url['host']; |
|
102
|
|
|
if (stristr($url['query'], 'ADMCMD_previewWS')) { |
|
103
|
|
|
$reqHeaders[] = 'Cookie: $Version="1"; be_typo_user="1"; $Path=/'; |
|
104
|
|
|
} |
|
105
|
|
|
$reqHeaders[] = 'Connection: close'; |
|
106
|
|
|
if ($url['user'] != '') { |
|
107
|
|
|
$reqHeaders[] = 'Authorization: Basic ' . base64_encode($url['user'] . ':' . $url['pass']); |
|
108
|
|
|
} |
|
109
|
|
|
$reqHeaders[] = 'X-T3crawler: ' . $crawlerId; |
|
110
|
|
|
$reqHeaders[] = 'User-Agent: TYPO3 crawler'; |
|
111
|
|
|
return $reqHeaders; |
|
112
|
|
|
} |
|
113
|
|
|
|
|
114
|
|
|
|
|
115
|
|
|
/** |
|
116
|
|
|
* Executes a shell command and returns the outputted result. |
|
117
|
|
|
* |
|
118
|
|
|
* @param string $command Shell command to be executed |
|
119
|
|
|
* @return string Outputted result of the command execution |
|
120
|
|
|
*/ |
|
121
|
|
|
protected function executeShellCommand($command) |
|
122
|
|
|
{ |
|
123
|
|
|
return shell_exec($command); |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
|
|
127
|
|
|
/** |
|
128
|
|
|
* Gets the base path of the website frontend. |
|
129
|
|
|
* (e.g. if you call http://mydomain.com/cms/index.php in |
|
130
|
|
|
* the browser the base path is "/cms/") |
|
131
|
|
|
* |
|
132
|
|
|
* @return string Base path of the website frontend |
|
133
|
|
|
*/ |
|
134
|
|
|
protected function getFrontendBasePath() |
|
135
|
|
|
{ |
|
136
|
|
|
$frontendBasePath = '/'; |
|
137
|
|
|
|
|
138
|
|
|
// Get the path from the extension settings: |
|
139
|
|
|
if (isset($this->extensionSettings['frontendBasePath']) && $this->extensionSettings['frontendBasePath']) { |
|
140
|
|
|
$frontendBasePath = $this->extensionSettings['frontendBasePath']; |
|
141
|
|
|
// If empty, try to use config.absRefPrefix: |
|
142
|
|
|
} elseif (isset($GLOBALS['TSFE']->absRefPrefix) && !empty($GLOBALS['TSFE']->absRefPrefix)) { |
|
143
|
|
|
$frontendBasePath = $GLOBALS['TSFE']->absRefPrefix; |
|
144
|
|
|
// If not in CLI mode the base path can be determined from $_SERVER environment: |
|
145
|
|
|
} elseif (!Environment::isCli()) { |
|
146
|
|
|
$frontendBasePath = GeneralUtility::getIndpEnv('TYPO3_SITE_PATH'); |
|
147
|
|
|
} |
|
148
|
|
|
|
|
149
|
|
|
// Base path must be '/<pathSegements>/': |
|
150
|
|
|
if ($frontendBasePath !== '/') { |
|
151
|
|
|
$frontendBasePath = '/' . ltrim($frontendBasePath, '/'); |
|
152
|
|
|
$frontendBasePath = rtrim($frontendBasePath, '/') . '/'; |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
return $frontendBasePath; |
|
156
|
|
|
} |
|
157
|
|
|
} |
|
158
|
|
|
|
AOEpeople /
crawler
