Sangu scraper boiler plate

2020-08-15 02:24:57 +05:00
parent 35e96947d9
commit 7d4815fcfa
2 changed files with 80 additions and 0 deletions
--- a/app/Services/Scrapers/SanguScraper.php
+++ b/app/Services/Scrapers/SanguScraper.php
@@ -0,0 +1,57 @@
+<?php
+
+namespace App\Services\Scrapers;
+
+use Goutte\Client;
+use Illuminate\Support\Carbon;
+
+class SanguScraper
+{
+    protected $client;
+
+
+    protected $content;
+    protected $topics = [];
+
+    public function __construct()
+    {
+        $this->client = new Client;
+    }
+
+    public function extract($url)
+    {
+
+        $crawler = $this->client->request('GET', $url);
+
+        $title = $crawler->filter('.brk-title2')->first()->text();
+        $image = $crawler->filter('.main-img img')->first()->attr('src');
+        $author = $crawler->filter('.author span')->first()->text();
+
+        $crawler->filter('article p')->each(function ($node) {
+            $this->content[] = preg_replace("/[a-zA-Z]/","",$node->text());
+        });
+
+        $crawler->filter('.article-tags')->each(function ($node) {
+           
+            $this->topics[] = [
+                "name" => $node->text(),
+                "slug" => str_replace("https://mihaaru.com/", "", $node->attr('href'))
+            ];
+        });
+
+        //Remove all the alphabets from string
+        //preg_replace("/[a-zA-Z]/", "",$string);
+       return [
+            'source'    => 'Sangu',
+            'title'      => $title,
+            'og_title'   => $crawler->filter('meta[property*="og:title"]')->first()->attr('content'),
+            'image'      => $image,
+            'content'    => $this->content,
+            'url'        => $url,
+            'date'       => "",
+            'guid'       => str_replace("https://mihaaru.com/news/","",$url),
+            'author'     => $author,
+            'topics'       => $this->topics
+        ];
+    }
+}