|
...
|
...
|
@@ -9,6 +9,7 @@ use App\Models\RouteMap\RouteMap; |
|
|
|
use App\Services\CosService;
|
|
|
|
use App\Services\ProjectServer;
|
|
|
|
use Illuminate\Console\Command;
|
|
|
|
use Illuminate\Support\Facades\Cache;
|
|
|
|
use Illuminate\Support\Facades\DB;
|
|
|
|
use Illuminate\Support\Facades\Redis;
|
|
|
|
|
|
...
|
...
|
@@ -75,29 +76,8 @@ class HtmlLanguageCollect extends Command |
|
|
|
$collect_info->status = CollectTask::STATUS_ING;
|
|
|
|
$collect_info->save();
|
|
|
|
|
|
|
|
//获取英文站域名
|
|
|
|
$domain = $collect_info->domain;
|
|
|
|
if (strpos($domain, '/') !== false) {
|
|
|
|
$domain = substr($domain, 0, strpos($domain, '/'));
|
|
|
|
} else {
|
|
|
|
$domain = str_replace($collect_info->language, 'www', $domain);
|
|
|
|
}
|
|
|
|
|
|
|
|
$web_url_domain = $domain;
|
|
|
|
$home_url = $domain;
|
|
|
|
$url_web_config = 'https://' . $domain . '/wp-content/cache/user_config.text';
|
|
|
|
$data_config = curl_c($url_web_config);
|
|
|
|
if ($data_config) {
|
|
|
|
$web_url_arr = parse_url($data_config['web_url_domain'] ?? '');
|
|
|
|
if (isset($web_url_arr['host'])) {
|
|
|
|
$web_url_domain = $web_url_arr['host'];
|
|
|
|
}
|
|
|
|
|
|
|
|
$home_url_arr = parse_url($data_config['home_url'] ?? '');
|
|
|
|
if (isset($home_url_arr['host'])) {
|
|
|
|
$home_url = $home_url_arr['host'];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
//获取站点正式和测试域名
|
|
|
|
$old_info = $this->getOldDomain($project_id, $collect_info->domain);
|
|
|
|
|
|
|
|
//采集html页面,下载资源到本地并替换
|
|
|
|
try {
|
|
...
|
...
|
@@ -110,7 +90,15 @@ class HtmlLanguageCollect extends Command |
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
$source_list = $this->html_preg($html, $project_id, $collect_info->domain, $web_url_domain, $home_url);
|
|
|
|
//如果有base64图片,先替换掉,再进行资源匹配
|
|
|
|
$new_html = $html;
|
|
|
|
preg_match_all("/data:([^;]*);base64,(.*)?\"/", $new_html, $result_img);
|
|
|
|
$img_base64 = $result_img[2] ?? [];
|
|
|
|
foreach ($img_base64 as $v64) {
|
|
|
|
$new_html = str_replace($v64, '', $new_html);
|
|
|
|
}
|
|
|
|
|
|
|
|
$source_list = $this->html_preg($new_html, $project_id, $collect_info->domain, $old_info['web_url_domain'], $old_info['home_url']);
|
|
|
|
|
|
|
|
if ($source_list) {
|
|
|
|
$html = $this->upload_source($html, $source_list, $project_id);
|
|
...
|
...
|
@@ -162,9 +150,6 @@ class HtmlLanguageCollect extends Command |
|
|
|
case 'blog':
|
|
|
|
$source = RouteMap::SOURCE_BLOG;
|
|
|
|
break;
|
|
|
|
case 'tag':
|
|
|
|
$source = RouteMap::SOURCE_PRODUCT_KEYWORD;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
$source = RouteMap::SOURCE_PRODUCT;
|
|
|
|
break;
|
|
...
|
...
|
@@ -197,6 +182,42 @@ class HtmlLanguageCollect extends Command |
|
|
|
return $task_id;
|
|
|
|
}
|
|
|
|
|
|
|
|
//获取站点老域名
|
|
|
|
protected function getOldDomain($project_id, $domain)
|
|
|
|
{
|
|
|
|
$key = 'project_collect_lan_domain_' . $project_id;
|
|
|
|
|
|
|
|
$data = Cache::get($key);
|
|
|
|
|
|
|
|
if (!$data) {
|
|
|
|
$web_url_domain = $domain;
|
|
|
|
$home_url = $domain;
|
|
|
|
|
|
|
|
$url_web_config = 'https://' . $domain . '/wp-content/cache/user_config.text';
|
|
|
|
$data_config = curl_c($url_web_config);
|
|
|
|
if ($data_config) {
|
|
|
|
$web_url_arr = parse_url($data_config['web_url_domain'] ?? '');
|
|
|
|
if (isset($web_url_arr['host'])) {
|
|
|
|
$web_url_domain = $web_url_arr['host'];
|
|
|
|
}
|
|
|
|
|
|
|
|
$home_url_arr = parse_url($data_config['home_url'] ?? '');
|
|
|
|
if (isset($home_url_arr['host'])) {
|
|
|
|
$home_url = $home_url_arr['host'];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
$data = [
|
|
|
|
'web_url_domain' => $web_url_domain,
|
|
|
|
'home_url' => $home_url,
|
|
|
|
];
|
|
|
|
|
|
|
|
Cache::add($key, $data, 3600);//缓存1小时
|
|
|
|
}
|
|
|
|
|
|
|
|
return $data;
|
|
|
|
}
|
|
|
|
|
|
|
|
//正则匹配html资源
|
|
|
|
protected function html_preg($html, $project_id, $domain, $web_url_domain, $home_url)
|
|
|
|
{
|
|
...
|
...
|
@@ -246,6 +267,13 @@ class HtmlLanguageCollect extends Command |
|
|
|
$check_vc_b && $source[] = $check_vc_b;
|
|
|
|
}
|
|
|
|
|
|
|
|
//a标签下载资源
|
|
|
|
preg_match_all('/<a\s+[^>]*?href\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_a);
|
|
|
|
$down = $result_a[2] ?? [];
|
|
|
|
foreach ($down as $vd) {
|
|
|
|
$check_vd = $this->url_check($vd, $project_id, $domain, $web_url_domain, $home_url);
|
|
|
|
$check_vd && $source[] = $check_vd;
|
|
|
|
}
|
|
|
|
|
|
|
|
return $source;
|
|
|
|
}
|
|
...
|
...
|
@@ -265,7 +293,7 @@ class HtmlLanguageCollect extends Command |
|
|
|
(empty($host) || $host == $web_url_domain || $host == $home_url)
|
|
|
|
&& $path
|
|
|
|
&& (strpos($path, '.') !== false)
|
|
|
|
&& (end($path_arr) != 'html')
|
|
|
|
&& (!in_array(end($path_arr), ['html', 'php', 'com', 'xml']))
|
|
|
|
) {
|
|
|
|
$source = CollectSource::where('project_id', $project_id)->where('origin', $url)->first();
|
|
|
|
if (!$source) {
|
|
...
|
...
|
@@ -329,7 +357,7 @@ class HtmlLanguageCollect extends Command |
|
|
|
continue;
|
|
|
|
}
|
|
|
|
$path_arr = explode('.', $vcs);
|
|
|
|
if(end($path_arr) == 'html'){
|
|
|
|
if(in_array(end($path_arr), ['html', 'php', 'com', 'xml'])){
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
...
|
...
|
|