82 lines
2.3 KiB
PHP
82 lines
2.3 KiB
PHP
<?php
|
|
|
|
namespace App\Services\Scrapers;
|
|
|
|
use Goutte\Client;
|
|
use Symfony\Component\HttpClient\HttpClient;
|
|
|
|
class MihaaruScraper
|
|
{
|
|
protected $client;
|
|
|
|
protected $title;
|
|
protected $content;
|
|
protected $image;
|
|
protected $topics = [];
|
|
protected $author;
|
|
|
|
public function __construct()
|
|
{
|
|
$this->client = new Client(
|
|
HttpClient::create([
|
|
"proxy" => config('karudhaas.proxy.host')
|
|
])
|
|
);
|
|
}
|
|
|
|
public function extract($url, $date = null)
|
|
{
|
|
|
|
$crawler = $this->client->request('GET', $url);
|
|
|
|
$crawler->filter('h1')->each(function ($node) {
|
|
$this->title = $node->text();
|
|
});
|
|
|
|
$this->image = $crawler->filter('.w-full.flex.flex-col.items-end.max-w-3xl.mb-10.relative img')->attr('src');
|
|
|
|
$crawler->filter('.by-line address')->each(function ($node) {
|
|
$author = $node->text();
|
|
//Trim all the white spaces
|
|
$spacetrim = str_replace(' ', '', $author);
|
|
//Replace multiple spaces and newlines with a single space
|
|
$cleaneddata = trim(preg_replace('/\s\s+/', ' ', $spacetrim));
|
|
$this->author = $cleaneddata;
|
|
});
|
|
|
|
$crawler->filter('.text-faseyha')->each(function ($node) {
|
|
$this->content[] = $node->text();
|
|
});
|
|
|
|
$crawler->filter('.items-end a')->each(function ($node) {
|
|
|
|
try {
|
|
$topicName = $node->filter('span')->text();
|
|
$topicSlug = ltrim($node->attr('href'), '/');
|
|
} catch (\Throwable $th) {
|
|
return;
|
|
}
|
|
|
|
$this->topics[] = [
|
|
"name" => $topicName,
|
|
"slug" => $topicSlug
|
|
];
|
|
});
|
|
|
|
//Remove all the alphabets from string
|
|
//preg_replace("/[a-zA-Z]/", "",$string);
|
|
return [
|
|
'source' => 'Mihaaru',
|
|
'title' => $this->title,
|
|
'og_title' => $crawler->filter('meta[property*="og:title"]')->first()->attr('content'),
|
|
'image' => $this->image,
|
|
'content' => $this->content,
|
|
'url' => $url,
|
|
'date' => $date,
|
|
'guid' => str_replace("https://mihaaru.com/news/", "", $url),
|
|
'author' => $this->author,
|
|
'topics' => $this->topics
|
|
];
|
|
}
|
|
}
|