fixing Scrapper

This commit is contained in:
ROMANGOLIENKO 2024-01-31 15:42:53 +03:00
parent 7a585aa3cd
commit 72191cbb16
1 changed files with 39 additions and 24 deletions

View File

@ -93,41 +93,56 @@ class PageScrapper
$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://")) {
$tmp = explode('/', rawurldecode($href));
foreach ( $tmp as $k => $v) {
$tmp[$k] = rawurlencode($v);
}
$href = implode('/', $tmp);
$html0 = str_replace($href, 'https://mkgtu.ru' . $href, $html0);
$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://")) {
$tmp = explode('/', rawurldecode($src));
foreach ( $tmp as $k => $v) {
$tmp[$k] = rawurlencode($v);
}
$src = implode('/', $tmp);
$html0 = str_replace($src, 'https://mkgtu.ru' . $src, $html0);
$unchanged = $srclink->attr('src');
$changed = $srclink->src = 'https://mkgtu.ru' . $src;
$html0 = str_replace(urldecode($unchanged), $changed, $html0);
}
}
str_replace('st yle', 'style', $html0);
// foreach ($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;
}