Roman_applicant-site/app/Services/PageScrapper.php

<?php

namespace App\Services;

use DiDom\Document;

/*
class PageScrapper
{
    private string $url;
    private string $contentMarker;

    public function __construct($url, $contentMarker = '<div class=["\']content_info["\']>')
    {
        $this->url = $url;
        $this->contentMarker = $contentMarker;
    }

    public function getHTML()
    {

        $page = file_get_contents($this->url);
        $strForPregMatch = "/" . "{$this->contentMarker}" . "(.*)<\/div>/is";

        $arr = [];
        $rez = preg_match_all($strForPregMatch, $page, $arr);

        return $content = $arr[0][0];
    }

    public function normalizeURLFile($content)
    {

        $rez = preg_match_all('/<a href="(.*)">/isU', $content, $arr);
        $arr[1] = array_unique($arr[1]);
        foreach ($arr[1] as $el) {
            if (!str_starts_with($el, 'https')) {
                $content = str_replace($el, 'https://mkgtu.ru' . $el, $content);
            }
        }
        $rez = preg_match_all('/src="(.*)">/isU', $content, $arr);
        $arr[1] = array_unique($arr[1]);
        foreach ($arr[1] as $el) {
            if (!str_starts_with($el, 'https') && str_contains($el, 'upload')) {
                $content = str_replace($el, 'https://mkgtu.ru' . $el, $content);
            }
        }

        return $content;
    }
    public function cutHTML($content, $strForScissors)
    {


        $arr = [];
        //<footer(.*)<\/footer>
        //safdsaf sdfdasf<footer>--------------------------------fsdfdasf <\/footer> asdfdasf asdf
        $rez = preg_match_all($strForScissors, $content, $arr);
        //$arr[1][0] = '>--------------------------------fsdfdasf ';
        $content = str_replace($arr[0], '', $content);
        //safdsaf sdfdasf<footer<\/footer> asdfdasf asdf

        return $content;
    }
}
*/

class PageScrapper
{
    private string $url;
    private string $contentMarker;

    public function __construct($url, $contentMarker)
    {
        $this->url = $url;
        $this->contentMarker = $contentMarker;
        $this->document = new Document($this->url, true);
    }
    public function getFullHTML()
    {
        return $this->document;
    }
    public function printHTML()
    {
        $rez = $this->document;
        $content = $rez->first($this->contentMarker)->html();
        return $content;
    }
    public function normalizePath()
    {
        $rez = $this->document;
        $content = $rez->first($this->contentMarker);
        $links = $content->find('a[href]');
        $srclinks = $content->find('img[src]');

        $html0 = $content->html();

        preg_match_all('/<a href="(.*)">/isU', $html0, $arr);
        foreach ($arr[1] as $el) {
            $html0 = str_replace($el, urldecode($el), $html0);
        }
        preg_match_all('/<img src="(.*)">/isU', $html0, $arr2);
        foreach ($arr2[1] as $el) {
            $html0 = str_replace($el, urldecode($el), $html0);
        }

        foreach ($links as $k => $link) {
            $href = $link->attr('href');
            if (!str_contains($link->attr('href'), "https://")) {
                $unchanged = $link->attr('href');
                $changed = $link->href = 'https://mkgtu.ru' . $href;
                $html0 = str_replace(urldecode($unchanged), $changed, $html0);
            }
        }






        foreach ($srclinks as $k => $srclink) {
            $src = $srclink->attr('src');
            if (!str_contains($srclink->attr('src'), "https://")) {
                $unchanged = $srclink->attr('src');
                $changed = $srclink->src = 'https://mkgtu.ru' . $src;
                $html0 = str_replace(urldecode($unchanged), $changed, $html0);
            }
        }
//        foreach ($srclinks as $k => $srclink) {
//            $src = $srclink->attr('src');
//
//            if (!str_contains($srclink->attr('src'), "https://")) {
//
//
//                $tmp = explode('/', rawurldecode($src));
//                foreach ( $tmp as $k => $v) {
//                    $tmp[$k] = rawurlencode($v);
//                }
//                $src = implode('/', $tmp);
//
//
//                $html0 = str_replace($src, 'https://mkgtu.ru' . $src, $html0);
//            }
//        }
//        str_replace('st yle', 'style', $html0);

        return $html0;
    }
}