|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler\Command; |
|
4
|
|
|
|
|
5
|
|
|
use PiedWeb\SeoPocketCrawler\Crawler; |
|
6
|
|
|
use PiedWeb\SeoPocketCrawler\CrawlerContinue; |
|
7
|
|
|
use PiedWeb\SeoPocketCrawler\CrawlerRestart; |
|
8
|
|
|
use Symfony\Component\Console\Command\Command; |
|
9
|
|
|
use Symfony\Component\Console\Input\InputArgument; |
|
10
|
|
|
use Symfony\Component\Console\Input\InputInterface; |
|
11
|
|
|
use Symfony\Component\Console\Input\InputOption; |
|
12
|
|
|
use Symfony\Component\Console\Output\OutputInterface; |
|
13
|
|
|
|
|
14
|
|
|
class CrawlerCommand extends Command |
|
15
|
|
|
{ |
|
16
|
|
|
protected static $defaultName = 'crawler:go'; |
|
17
|
|
|
|
|
18
|
|
|
protected $id; |
|
19
|
|
|
|
|
20
|
|
|
protected function configure() |
|
21
|
|
|
{ |
|
22
|
|
|
$this->setDescription('Crawl a website.'); |
|
23
|
|
|
|
|
24
|
|
|
$this |
|
25
|
|
|
->addArgument( |
|
26
|
|
|
'start', |
|
27
|
|
|
InputArgument::REQUIRED, |
|
28
|
|
|
'Define where the crawl start. Eg: https://piedweb.com' |
|
29
|
|
|
.PHP_EOL.'You can specify an id from a previous crawl. Other options will not be listen.' |
|
30
|
|
|
.PHP_EOL.'You can use `last` to continue the last crawl (just stopped).' |
|
31
|
|
|
) |
|
32
|
|
|
->addOption('limit', 'l', InputOption::VALUE_REQUIRED, 'Define where a depth limit', 5) |
|
33
|
|
|
->addOption( |
|
34
|
|
|
'ignore', |
|
35
|
|
|
'i', |
|
36
|
|
|
InputOption::VALUE_REQUIRED, |
|
37
|
|
|
'Virtual Robots.txt to respect (could be a string or an URL).' |
|
38
|
|
|
) |
|
39
|
|
|
->addOption( |
|
40
|
|
|
'user-agent', |
|
41
|
|
|
'u', |
|
42
|
|
|
InputOption::VALUE_REQUIRED, |
|
43
|
|
|
'Define the user-agent used during the crawl.', |
|
44
|
|
|
'SEO Pocket Crawler - PiedWeb.com/seo/crawler' |
|
45
|
|
|
) |
|
46
|
|
|
->addOption( |
|
47
|
|
|
'wait', |
|
48
|
|
|
'w', |
|
49
|
|
|
InputOption::VALUE_REQUIRED, |
|
50
|
|
|
'In Microseconds, the time to wait between 2 requests. Default 0,1s.', |
|
51
|
|
|
100000 |
|
52
|
|
|
) |
|
53
|
|
|
->addOption( |
|
54
|
|
|
'cache-method', |
|
55
|
|
|
'c', |
|
56
|
|
|
InputOption::VALUE_REQUIRED, |
|
57
|
|
|
'In Microseconds, the time to wait between two request. Default : 100000 (0,1s).', |
|
58
|
|
|
\PiedWeb\SeoPocketCrawler\Recorder::CACHE_ID |
|
59
|
|
|
) |
|
60
|
|
|
->addOption( |
|
61
|
|
|
'restart', |
|
62
|
|
|
'r', |
|
63
|
|
|
InputOption::VALUE_REQUIRED, |
|
64
|
|
|
'Permit to restart a previous crawl. Values 1 = fresh restart, 2 = restart from cache' |
|
65
|
|
|
) |
|
66
|
|
|
; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
protected function execute(InputInterface $input, OutputInterface $output) |
|
70
|
|
|
{ |
|
71
|
|
|
$this->checkArguments($input); |
|
72
|
|
|
|
|
73
|
|
|
$start = microtime(true); |
|
74
|
|
|
|
|
75
|
|
|
$crawler = $this->initCrawler($input); |
|
76
|
|
|
|
|
77
|
|
|
$output->writeln(['', '', 'Crawl starting !', '============', '', 'ID: '.$crawler->getConfig()->getId()]); |
|
78
|
|
|
$output->writeln([ |
|
79
|
|
|
null !== $this->id ? ($input->getOption('restart') ? 'Restart' : 'Continue') : '', |
|
80
|
|
|
'', |
|
81
|
|
|
'Details : ', |
|
82
|
|
|
'- Crawl starting at '.$crawler->getConfig()->getBase().$crawler->getConfig()->getStartUrl(), |
|
83
|
|
|
'- User-Agent used `'.$crawler->getConfig()->getUserAgent(), |
|
84
|
|
|
'- `'.$crawler->getConfig()->getWait().' ms between two requests', |
|
85
|
|
|
]); |
|
86
|
|
|
|
|
87
|
|
|
$crawler->crawl(); |
|
88
|
|
|
|
|
89
|
|
|
$end = microtime(true); |
|
90
|
|
|
|
|
91
|
|
|
$output->writeln(['', '---------------', 'Crawl succeed', 'You can find your data in ']); |
|
92
|
|
|
|
|
93
|
|
|
echo realpath($crawler->getConfig()->getDataFolder()).'/data.csv'.PHP_EOL; |
|
94
|
|
|
|
|
95
|
|
|
$output->writeln(['', '', '----Chrono----', (round(($end - $start), 2)).'s', '', '']); |
|
96
|
|
|
|
|
97
|
|
|
return 0; |
|
98
|
|
|
} |
|
99
|
|
|
|
|
100
|
|
|
public function checkArguments(InputInterface $input) |
|
101
|
|
|
{ |
|
102
|
|
|
if (! filter_var($input->getArgument('start'), FILTER_VALIDATE_URL)) { |
|
103
|
|
|
$this->id = $input->getArgument('start'); |
|
104
|
|
|
} |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
/** |
|
108
|
|
|
* @return Crawler |
|
109
|
|
|
*/ |
|
110
|
|
|
public function initCrawler(InputInterface $input) |
|
111
|
|
|
{ |
|
112
|
|
|
if (null === $this->id) { |
|
113
|
|
|
return new Crawler( |
|
114
|
|
|
(string) $input->getArgument('start'), |
|
115
|
|
|
$this->loadVirtualRobotsTxt($input), |
|
116
|
|
|
intval($input->getOption('limit')), |
|
117
|
|
|
(string) $input->getOption('user-agent'), |
|
118
|
|
|
intval($input->getOption('cache-method')), |
|
119
|
|
|
intval($input->getOption('wait')), |
|
120
|
|
|
! $input->getOption('quiet') |
|
121
|
|
|
); |
|
122
|
|
|
} |
|
123
|
|
|
|
|
124
|
|
|
if ($input->getOption('restart')) { |
|
125
|
|
|
return new CrawlerRestart( |
|
126
|
|
|
$this->id, |
|
127
|
|
|
2 == $input->getOption('restart') ? true : false, // $fromCache |
|
128
|
|
|
! $input->getOption('quiet') |
|
129
|
|
|
); |
|
130
|
|
|
} |
|
131
|
|
|
|
|
132
|
|
|
return new CrawlerContinue($this->id, ! $input->getOption('quiet')); |
|
133
|
|
|
} |
|
134
|
|
|
|
|
135
|
|
|
public function loadVirtualRobotsTxt(InputInterface $input) |
|
136
|
|
|
{ |
|
137
|
|
|
if (null === $input->getOption('ignore')) { |
|
138
|
|
|
return ''; |
|
139
|
|
|
} |
|
140
|
|
|
|
|
141
|
|
|
$ignore = (string) $input->getOption('ignore'); |
|
142
|
|
|
|
|
143
|
|
|
if (filter_var($ignore, FILTER_VALIDATE_URL)) { |
|
144
|
|
|
return \PiedWeb\Curl\Request::get($ignore); |
|
145
|
|
|
} |
|
146
|
|
|
|
|
147
|
|
|
if (file_exists($ignore)) { |
|
148
|
|
|
return file_get_contents($ignore); |
|
149
|
|
|
} |
|
150
|
|
|
|
|
151
|
|
|
throw new \Exception('An error occured with your --ignore option'); |
|
152
|
|
|
} |
|
153
|
|
|
} |
|
154
|
|
|
|
PiedWeb /
SeoPocketCrawler
