95 lines
2.6 KiB
PHP
95 lines
2.6 KiB
PHP
<?php
|
|
|
|
namespace App\Services\Scrapers;
|
|
|
|
use Goutte\Client;
|
|
use Illuminate\Support\Carbon;
|
|
|
|
class ThiladhunScraper
|
|
{
|
|
protected $client;
|
|
|
|
protected $title;
|
|
protected $content;
|
|
protected $image;
|
|
protected $author = "unknown";
|
|
protected $topics = [];
|
|
|
|
/**
|
|
* __construct.
|
|
*
|
|
* @return void
|
|
*/
|
|
public function __construct()
|
|
{
|
|
$this->client = new Client();
|
|
}
|
|
|
|
/**
|
|
* extract.
|
|
*
|
|
* @param mixed $url
|
|
* @param mixed $date
|
|
* @param mixed $guid
|
|
*
|
|
* @return array
|
|
*/
|
|
public function extract($url)
|
|
{
|
|
|
|
if (strpos($url, 'gallery') !== false) {
|
|
return null;
|
|
}
|
|
|
|
$crawler = $this->client->request('GET', $url);
|
|
|
|
$crawler->filter('meta[property="og:title"]')->each(function ($node) {
|
|
$this->title = $node->attr('content');
|
|
});
|
|
|
|
|
|
$crawler->filter('div.elementor-widget-container > p')->each(function ($node) {
|
|
$this->content[] = preg_replace("/[a-zA-Z]/", "", $node->text());;
|
|
});
|
|
|
|
$this->image = $crawler->filter('div[data-widget_type*="theme-post-featured-image.default"] figure img')->first()->attr('src');
|
|
|
|
$crawler->filter('.entry-tags a')->each(function ($node) {
|
|
if (!preg_match('/[^A-Za-z0-9-]/', basename($node->attr('href')))) {
|
|
$this->topics[] = [
|
|
"name" => $node->text(),
|
|
"slug" => basename($node->attr('href'))
|
|
];
|
|
}
|
|
});
|
|
|
|
$crawler->filter('a > .elementor-post-info__item--type-author')->each(function ($node) {
|
|
$this->author = $node->text();
|
|
});
|
|
|
|
$urlComponents = parse_url($url);
|
|
$cleanUrl = $urlComponents['scheme'] . '://' . $urlComponents['host'] . $urlComponents['path'];
|
|
|
|
// Extract and modify the guid
|
|
$guid = basename($urlComponents['path']);
|
|
|
|
return [
|
|
'source' => 'Thiladhun News',
|
|
'title' => $this->title,
|
|
'og_title' => str_replace(" | Thiladhun", "", $crawler->filter('title')->first()->text('content')),
|
|
'image' => $this->image,
|
|
'content' => $this->content,
|
|
'date' => Carbon::now(),
|
|
'url' => $cleanUrl,
|
|
'author' => $this->author,
|
|
'guid' => $guid,
|
|
'topics' => $this->topics ?: [
|
|
[
|
|
"name" => "ވަކި މަޢުލޫއެއް ނޭންގެ",
|
|
"slug" => "no-specific-topic"
|
|
]
|
|
]
|
|
];
|
|
}
|
|
}
|