作者 刘锟

update

  1 +<?php
  2 +
  3 +namespace App\Console\Commands\Update;
  4 +
  5 +use App\Models\Collect\CollectSource;
  6 +use App\Models\Collect\CollectTask;
  7 +use App\Models\Com\UpdateLog;
  8 +use App\Models\Com\UpdateOldInfo;
  9 +use App\Models\RouteMap\RouteMap;
  10 +use App\Services\CosService;
  11 +use App\Services\ProjectServer;
  12 +use Illuminate\Console\Command;
  13 +use Illuminate\Support\Facades\DB;
  14 +use Illuminate\Support\Facades\Redis;
  15 +
  16 +/**
  17 + * 4.0,5.0升级到6.0,主站页面采集
  18 + * Class ProjectImport
  19 + * @package App\Console\Commands
  20 + * @author Akun
  21 + * @date 2023/11/10 16:04
  22 + */
  23 +class HtmlSpecialCollect extends Command
  24 +{
  25 + /**
  26 + * The name and signature of the console command.
  27 + *
  28 + * @var string
  29 + */
  30 + protected $signature = 'project_html_special_collect';
  31 +
  32 + /**
  33 + * The console command description.
  34 + *
  35 + * @var string
  36 + */
  37 + protected $description = '执行特殊项目html页面采集';
  38 +
  39 +
  40 + public function handle()
  41 + {
  42 + ini_set('memory_limit', '512M');
  43 + while (true) {
  44 + $this->start_collect();
  45 + }
  46 + }
  47 +
  48 + protected function start_collect()
  49 + {
  50 + $task_id = $this->get_task();
  51 + if ($task_id === false) {
  52 + //所有项目采集完成
  53 + sleep(60);
  54 + return true;
  55 + } elseif ($task_id === 0) {
  56 + //当前项目采集完成
  57 + sleep(2);
  58 + return true;
  59 + }
  60 +
  61 + $task_arr = explode('_', $task_id);
  62 + $project_id = $task_arr[0];
  63 + $collect_id = $task_arr[1];
  64 +
  65 + //设置数据库
  66 + $project = ProjectServer::useProject($project_id);
  67 + if ($project) {
  68 + $collect_info = CollectTask::select(['id', 'domain', 'route'])->where('id', $collect_id)->where('status', CollectTask::STATUS_UN)->where('language', '')->first();
  69 +
  70 + if (!$collect_info) {
  71 + sleep(2);
  72 + return true;
  73 + }
  74 +
  75 + echo 'date:' . date('Y-m-d H:i:s') . ', project_id: ' . $project_id . ', collect_id: ' . $collect_id . ', collect start' . PHP_EOL;
  76 +
  77 + $collect_info->status = CollectTask::STATUS_ING;
  78 + $collect_info->save();
  79 +
  80 + //获取站点原始域名信息
  81 + $old_info = UpdateOldInfo::getOldDomain($project_id, $collect_info->domain);
  82 +
  83 + //采集html页面,下载资源到本地并替换
  84 + try {
  85 + $html = curl_c('https://' . $collect_info->domain . $collect_info->route, false);
  86 + if ($html == '0') {
  87 + $collect_info->status = CollectTask::STATUS_FAIL;
  88 + $collect_info->save();
  89 + echo 'date:' . date('Y-m-d H:i:s') . ', project_id: ' . $project_id . ', collect_id: ' . $collect_id . ', error: no html' . PHP_EOL;
  90 + sleep(2);
  91 + return true;
  92 + }
  93 +
  94 + //如果有base64图片,先替换掉,再进行资源匹配
  95 + $new_html = $html;
  96 + preg_match_all("/data:([^;]*);base64,(.*)?\"/", $new_html, $result_img);
  97 + $img_base64 = $result_img[2] ?? [];
  98 + foreach ($img_base64 as $v64) {
  99 + $new_html = str_replace($v64, '', $new_html);
  100 + }
  101 +
  102 + $source_list = $this->html_preg($new_html, $project_id, $collect_info->domain, $old_info['web_url_domain'], $old_info['home_url']);
  103 +
  104 + if ($source_list) {
  105 + $html = $this->upload_source($html, $source_list, $project_id, $collect_info->domain, $old_info['web_url_domain'], $old_info['home_url']);
  106 + }
  107 + } catch (\Exception $e) {
  108 + $collect_info->status = CollectTask::STATUS_FAIL;
  109 + $collect_info->save();
  110 +
  111 + echo 'date:' . date('Y-m-d H:i:s') . ', project_id: ' . $project_id . ', collect_id: ' . $collect_id . ', error: ' . $e->getMessage() . PHP_EOL;
  112 + sleep(2);
  113 + return true;
  114 + }
  115 +
  116 + $collect_info->html = $html;
  117 + $collect_info->status = CollectTask::STATUS_COM;
  118 + $collect_info->save();
  119 +
  120 + echo 'date:' . date('Y-m-d H:i:s') . ', project_id: ' . $project_id . ', collect_id: ' . $collect_id . ', collect end' . PHP_EOL;
  121 + }
  122 + //关闭数据库
  123 + DB::disconnect('custom_mysql');
  124 +
  125 + sleep(2);
  126 + return true;
  127 + }
  128 +
  129 + //获取任务
  130 + protected function get_task()
  131 + {
  132 + $key = 'console_html_special_collect_task';
  133 + $task_id = Redis::rpop($key);
  134 + if ($task_id) {
  135 + return $task_id;
  136 + }
  137 +
  138 +
  139 + $update_log = UpdateLog::whereIn('project_id', [626])->where('status', UpdateLog::STATUS_COM)->where('collect_status', UpdateLog::COLLECT_STATUS_UN)->orderBy('project_id', 'asc')->first();
  140 + if (!$update_log) {
  141 + return false;
  142 + }
  143 +
  144 + switch ($update_log->api_type) {
  145 + case 'page':
  146 + $source = RouteMap::SOURCE_PAGE;
  147 + break;
  148 + case 'news':
  149 + $source = RouteMap::SOURCE_NEWS;
  150 + break;
  151 + case 'blog':
  152 + $source = RouteMap::SOURCE_BLOG;
  153 + break;
  154 + default:
  155 + $source = RouteMap::SOURCE_PRODUCT;
  156 + break;
  157 + }
  158 +
  159 + $complete = false;
  160 + //设置数据库
  161 + $project = ProjectServer::useProject($update_log->project_id);
  162 + if ($project) {
  163 + $collect_list = CollectTask::select(['id', 'project_id'])->where('project_id', $update_log['project_id'])->where('source', $source)->where('language', '')->where('status', CollectTask::STATUS_UN)->orderBy('id', 'asc')->limit(50)->get();
  164 +
  165 + if ($collect_list->count() == 0) {
  166 + $complete = true;
  167 + } else {
  168 + foreach ($collect_list as $collect) {
  169 + Redis::lpush($key, $collect['project_id'] . '_' . $collect['id']);
  170 + }
  171 + }
  172 + }
  173 + //关闭数据库
  174 + DB::disconnect('custom_mysql');
  175 +
  176 + if ($complete) {
  177 + $update_log->collect_status = UpdateLog::COLLECT_STATUS_MAIN;
  178 + $update_log->save();
  179 + return 0;
  180 + }
  181 +
  182 + $task_id = Redis::rpop($key);
  183 + return $task_id;
  184 + }
  185 +
  186 + //正则匹配html资源
  187 + protected function html_preg($html, $project_id, $domain, $web_url_domain, $home_url)
  188 + {
  189 + $source = [];
  190 +
  191 + if (!$html) {
  192 + return $source;
  193 + }
  194 +
  195 + //image
  196 + preg_match_all('/<img\s+[^>]*?src\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_img);
  197 + $img = $result_img[2] ?? [];
  198 + foreach ($img as $vi) {
  199 + $check_vi = $this->url_check($vi, $project_id, $domain, $web_url_domain, $home_url);
  200 + $check_vi && $source[] = $check_vi;
  201 + }
  202 +
  203 + //js
  204 + preg_match_all('/<script\s+[^>]*?src\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_js);
  205 + $js = $result_js[2] ?? [];
  206 + foreach ($js as $vj) {
  207 + $check_vj = $this->url_check($vj, $project_id, $domain, $web_url_domain, $home_url);
  208 + $check_vj && $source[] = $check_vj;
  209 + }
  210 +
  211 + //video
  212 + preg_match_all('/<source\s+[^>]*?src\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_video);
  213 + $video = $result_video[2] ?? [];
  214 + foreach ($video as $vv) {
  215 + $check_vv = $this->url_check($vv, $project_id, $domain, $web_url_domain, $home_url);
  216 + $check_vv && $source[] = $check_vv;
  217 + }
  218 +
  219 + //css
  220 + preg_match_all('/<link\s+[^>]*?href\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_css);
  221 + $css = $result_css[2] ?? [];
  222 + foreach ($css as $vc) {
  223 + $check_vc = $this->url_check($vc, $project_id, $domain, $web_url_domain, $home_url);
  224 + $check_vc && $source[] = $check_vc;
  225 + }
  226 +
  227 + //css background
  228 + preg_match_all("/url\(['\"]?(\s*[^>]+?)['\"]?\)/i", $html, $result_css_b);
  229 + $css_b = $result_css_b[1] ?? [];
  230 + foreach ($css_b as $vc_b) {
  231 + $check_vc_b = $this->url_check($vc_b, $project_id, $domain, $web_url_domain, $home_url);
  232 + $check_vc_b && $source[] = $check_vc_b;
  233 + }
  234 +
  235 + //a标签下载资源
  236 + preg_match_all('/<a\s+[^>]*?href\s*=\s*(\'|\")(.*?)\\1[^>]*?\/?\s*>/i', $html, $result_a);
  237 + $down = $result_a[2] ?? [];
  238 + foreach ($down as $vd) {
  239 + $check_vd = $this->url_check($vd, $project_id, $domain, $web_url_domain, $home_url);
  240 + $check_vd && $source[] = $check_vd;
  241 + }
  242 +
  243 + return $source;
  244 + }
  245 +
  246 + //判断资源是否需要下载
  247 + protected function url_check($url, $project_id, $domain, $web_url_domain, $home_url)
  248 + {
  249 + if ($url) {
  250 + $url = str_replace('&quot;', '', $url);
  251 + $arr = parse_url($url);
  252 + $scheme = $arr['scheme'] ?? '';
  253 + $host = $arr['host'] ?? '';
  254 + $path = $arr['path'] ?? '';
  255 + $query = $arr['query'] ?? '';
  256 +
  257 + $path_arr = explode('.', $path);
  258 + $path_end = end($path_arr);
  259 + if (
  260 + (empty($scheme) || $scheme == 'https' || $scheme == 'http')
  261 + && (empty($host) || (strpos($web_url_domain, $host) !== false) || (strpos($home_url, $host) !== false))
  262 + && $path
  263 + && (substr($path, 0, 1) == '/')
  264 + && (strpos($path, '.') !== false)
  265 + && (strpos($path_end, 'html') === false)
  266 + && (strpos($path_end, 'php') === false)
  267 + && (strpos($path_end, 'com') === false)
  268 + && (strpos($path_end, 'xml') === false)
  269 + ) {
  270 + $new_url = str_replace($home_url, $web_url_domain, $url);
  271 + $source = CollectSource::where('project_id', $project_id)->where('origin', $new_url)->first();
  272 + if (!$source) {
  273 + return [
  274 + 'download' => true,
  275 + 'url' => $url,
  276 + 'url_complete' => ($scheme ?: 'https') . '://' . ($host ?: $domain) . $path . ($query ? '?' . $query : '')
  277 + ];
  278 + } else {
  279 + return [
  280 + 'download' => false,
  281 + 'url' => $url,
  282 + 'url_complete' => $source['target']
  283 + ];
  284 + }
  285 + } else {
  286 + return false;
  287 + }
  288 + } else {
  289 + return false;
  290 + }
  291 + }
  292 +
  293 + //下载并替换资源
  294 + protected function upload_source($html, $source, $project_id, $domain, $web_url_domain, $home_url)
  295 + {
  296 + foreach ($source as $vs) {
  297 +
  298 + if ($vs['download']) {
  299 + $new_source = CosService::uploadRemote($project_id, 'source', $vs['url_complete']);
  300 + if ($new_source) {
  301 + CollectSource::insert([
  302 + 'project_id' => $project_id,
  303 + 'origin' => $vs['url'],
  304 + 'target' => $new_source,
  305 + 'created_at' => date('Y-m-d H:i:s'),
  306 + 'updated_at' => date('Y-m-d H:i:s'),
  307 + ]);
  308 + $html = str_replace($vs['url'], getImageUrl($new_source), $html);
  309 +
  310 + if (substr($new_source, -3, 3) == 'css' || substr($new_source, -2, 2) == 'js') {
  311 +
  312 + $source_html = curl_c(getImageUrl($new_source), false);
  313 +
  314 + if (substr($new_source, -3, 3) == 'css') {
  315 + preg_match_all("/url\(['\"]?(\s*[^>]+?)['\"]?\)/i", $source_html, $result_source);
  316 + } else {
  317 + preg_match_all("/[large|thumb]+URL:['\"]+(\s*[^>]+?)['\"]+,/i", $source_html, $result_source);
  318 + }
  319 +
  320 + $js_css_source = $result_source[1] ?? [];
  321 + if ($js_css_source) {
  322 + foreach ($js_css_source as $vjs) {
  323 + if (strpos($vjs, 'URL:"') !== false) {
  324 + $vjs = substr($vjs, strpos($vjs, 'URL:"') + 5);
  325 + }
  326 +
  327 + $vjs_down = str_replace('&quot;', '', $vjs);
  328 + if (strpos($vjs_down, 'data:') !== false) {
  329 + //过滤二进制文件
  330 + continue;
  331 + }
  332 + if (strlen($vjs_down) > 255) {
  333 + //过滤太长文件
  334 + continue;
  335 + }
  336 +
  337 + $vjs_down_arr = parse_url($vjs_down);
  338 + $vjs_down_host = $vjs_down_arr['host'] ?? '';
  339 +
  340 + $cos = config('filesystems.disks.cos');
  341 + $cosCdn = $cos['cdn'];
  342 +
  343 + if ($vjs_down_host && $vjs_down_host == $cosCdn) {
  344 + //过滤已经下载的
  345 + continue;
  346 + }
  347 +
  348 + if (empty($vjs_down_host) && substr($vjs_down, 0, 1) != '/') {
  349 + //相对路径
  350 + $url_arr = explode('/', $vs['url']);
  351 + $url_arr[count($url_arr) - 1] = $vjs_down;
  352 + $vjs_down = implode('/', $url_arr);
  353 + }
  354 +
  355 + $vjs_result = $this->url_check($vjs_down, $project_id, $domain, $web_url_domain, $home_url);
  356 + if (!$vjs_result) {
  357 + continue;
  358 + }
  359 +
  360 + if ($vjs_result['download']) {
  361 + $new_vjs = CosService::uploadRemote($project_id, 'source', $vjs_result['url_complete']);
  362 + if ($new_vjs) {
  363 + CollectSource::insert([
  364 + 'project_id' => $project_id,
  365 + 'origin' => $vjs_result['url'],
  366 + 'target' => $new_vjs,
  367 + 'created_at' => date('Y-m-d H:i:s'),
  368 + 'updated_at' => date('Y-m-d H:i:s'),
  369 + ]);
  370 + $source_html = str_replace($vjs, getImageUrl($new_vjs), $source_html);
  371 + }
  372 + } else {
  373 + $source_html = str_replace($vjs, getImageUrl($vjs_result['url_complete']), $source_html);
  374 + }
  375 + }
  376 +
  377 + CosService::uploadRemote($project_id, 'source', $new_source, $new_source, $source_html);
  378 + }
  379 + }
  380 + }
  381 + } else {
  382 + $html = str_replace($vs['url'], getImageUrl($vs['url_complete']), $html);
  383 + }
  384 + }
  385 +
  386 + return $html;
  387 + }
  388 +}