|
1
|
|
|
<?php |
|
2
|
|
|
/* |
|
3
|
|
|
* This file is part of the Slince/China package. |
|
4
|
|
|
* |
|
5
|
|
|
* (c) Slince <[email protected]> |
|
6
|
|
|
* |
|
7
|
|
|
* For the full copyright and license information, please view the LICENSE |
|
8
|
|
|
* file that was distributed with this source code. |
|
9
|
|
|
*/ |
|
10
|
|
|
|
|
11
|
|
|
namespace China\Command; |
|
12
|
|
|
|
|
13
|
|
|
use China\Nationality\Nationality; |
|
14
|
|
|
use Symfony\Component\Console\Input\InputInterface; |
|
15
|
|
|
use Symfony\Component\Console\Output\OutputInterface; |
|
16
|
|
|
use Symfony\Component\Console\Style\SymfonyStyle; |
|
17
|
|
|
use Symfony\Component\DomCrawler\Crawler; |
|
18
|
|
|
|
|
19
|
|
|
class GetNationalityCommand extends CrawlCommand |
|
20
|
|
|
{ |
|
21
|
|
|
/** |
|
22
|
|
|
* @var string |
|
23
|
|
|
*/ |
|
24
|
|
|
const URL = 'https://baike.baidu.com/item/56个民族/383735'; |
|
25
|
|
|
|
|
26
|
|
|
/** |
|
27
|
|
|
* {@inheritdoc} |
|
28
|
|
|
*/ |
|
29
|
|
|
public function configure() |
|
30
|
|
|
{ |
|
31
|
|
|
$this->setName('crawl:nationality'); |
|
32
|
|
|
$this->setDescription('从百度百科采集民族信息'); |
|
33
|
|
|
} |
|
34
|
|
|
|
|
35
|
|
|
/** |
|
36
|
|
|
* {@inheritdoc} |
|
37
|
|
|
*/ |
|
38
|
|
|
public function execute(InputInterface $input, OutputInterface $output) |
|
39
|
|
|
{ |
|
40
|
|
|
$style = new SymfonyStyle($input, $output); |
|
41
|
|
|
|
|
42
|
|
|
$outputFile = static::RESOURCE_DIR.'/nationalities.json'; |
|
43
|
|
|
|
|
44
|
|
|
$crawler = $this->getClient()->request('GET', static::URL); |
|
45
|
|
|
|
|
46
|
|
|
$tables = $crawler->filter('table[log-set-param="table_view"]'); |
|
47
|
|
|
$nationalities = $this->extractPinyinData($tables->eq(1)); |
|
48
|
|
|
$populations = $this->extractPopulationData($tables->eq(2)); |
|
49
|
|
|
|
|
50
|
|
|
$nationalities = $this->mergeData($nationalities,$populations); |
|
51
|
|
|
$this->filesystem->dumpFile($outputFile, \GuzzleHttp\json_encode($nationalities, JSON_UNESCAPED_UNICODE)); |
|
52
|
|
|
|
|
53
|
|
|
$style->writeln(sprintf('<info>Crawl completed, please check the file at "%s"</info>', realpath($outputFile))); |
|
54
|
|
|
} |
|
55
|
|
|
|
|
56
|
|
|
protected function extractPinyinData(Crawler $crawler) |
|
57
|
|
|
{ |
|
58
|
|
|
$nationalities = $crawler->filter('tr')->each(function(Crawler $itemNode){ |
|
59
|
|
|
$data = []; |
|
60
|
|
|
$itemNode->filter('td')->each(function(Crawler $tdNode, $index) use (&$data){ |
|
61
|
|
|
if ($index % 2 === 0) { |
|
62
|
|
|
$data[$index] = [ |
|
63
|
|
|
'name' => trim($tdNode->text()), |
|
64
|
|
|
'pinyin' => false, |
|
65
|
|
|
]; |
|
66
|
|
|
} else { |
|
67
|
|
|
$data[$index - 1]['pinyin'] = trim($tdNode->text()); |
|
68
|
|
|
} |
|
69
|
|
|
}); |
|
70
|
|
|
|
|
71
|
|
|
return $data; |
|
72
|
|
|
}); |
|
73
|
|
|
|
|
74
|
|
|
return call_user_func_array('array_merge', $nationalities); |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
protected function extractPopulationData(Crawler $crawler) |
|
78
|
|
|
{ |
|
79
|
|
|
$data = []; |
|
80
|
|
|
$crawler->filter('tr')->each(function(Crawler $itemNode) use (&$data){ |
|
81
|
|
|
$tds = $itemNode->filter('td'); |
|
82
|
|
|
if (count($tds) > 0) { |
|
83
|
|
|
$name = trim($tds->first()->text()); |
|
84
|
|
|
$data[$name] = trim($tds->eq(1)->text()); |
|
85
|
|
|
} |
|
86
|
|
|
}); |
|
87
|
|
|
|
|
88
|
|
|
return $data; |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
|
|
protected function mergeData($nationalityInfos, $populations) |
|
92
|
|
|
{ |
|
93
|
|
|
$nationalities = []; |
|
94
|
|
|
foreach ($nationalityInfos as $nationalityInfo) { |
|
95
|
|
|
$nationalities[] = new Nationality($nationalityInfo['name'], |
|
96
|
|
|
$nationalityInfo['pinyin'], |
|
97
|
|
|
isset($populations[$nationalityInfo['name']]) ? $populations[$nationalityInfo['name']] : 0 |
|
98
|
|
|
); |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
|
|
return $nationalities; |
|
102
|
|
|
} |
|
103
|
|
|
} |
slince /
China
