评论数据入库

config/define.php

@@ -53,6 +53,7 @@
     define('WEIBO_LOGIN_COOKIE', '_2A25FMHeEDeRhGeFH41MV8ibNyjSIHXVmTPVMrDV8PUNbmtAYLWLlkW9NekLV22uXkAnuAlCngnjVjozSXR5mujv7');
 
     define('_RQ_SPIDER_WEIBO_BEHAVIOR', 'rq_spider_weibo_behavior');
+    define('_RQ_SPIDER_WEIBO_COMMENT', 'rq_spider_weibo_comment');
 
     define('WEIBO_DETAIL_URL', 'https://weibo.yizherenxin.cn/detail/%s?v=1');
 
@@ -82,4 +83,14 @@
     define('ZHISHIKU_SPIDER_COMMENT', 'zhishiku_spider_comment.log');
 
     define('ZHISHIKU_SPIDER_TEMP_PATH', DATACENTER_ROOT.'/zhishiku_spider_temp/');
-    define('ZHISHIKU_SPIDER_COMMENT_PATH', ZHISHIKU_SPIDER_TEMP_PATH.'/spider_comment/%s/%s.log');
+    define('ZHISHIKU_SPIDER_COMMENT_PATH', ZHISHIKU_SPIDER_TEMP_PATH.'/spider_comment/%s/%s.log');
+
+    //	采集评论状态 0=未采集 1=采集时出现错误中断 2=采集成功
+    define('SPIDER_COMMENT_STATUS_NO', 0);
+    define('SPIDER_COMMENT_STATUS_ERROR', 1);
+    define('SPIDER_COMMENT_STATUS_SUCCESS', 2);
+
+    define('_RS_SPIDER_WEIBO_COMMENT', 'rs_spider_weibo_comment');
+    define('_RC_WEIBO_LOGIN_COOKIE2', 'rc_weibo_login_cookie2');
+
+    define('WEIBO_LOGIN_COOKIE2', '_2A25FdJmRDeRhGeFG71sY9CbOyDuIHXVmC5NZrDV8PUNbmtAYLUj1kW9NeUknrDj_DFv7tUuAQqx1kWSg_y3B4Qe3');

data/dWeibo.php

@@ -24,6 +24,7 @@ class dWeibo extends dBase {
             'comments_count',
             'attitudes_count',
             'status',
+            'spider_comment_status'
         ),
     );
 

model/mSpider.php

@@ -7,6 +7,8 @@ include_once(SERVER_ROOT . "/model/mBase.php");
 
 class mSpider extends mBase {
 
+    public $is_change_ip = false;
+
     public function __construct() {
         $this->obj = new dWeibo();
         $this->tbl = 'spider_weibo';
@@ -15,20 +17,22 @@ class mSpider extends mBase {
     public function getZmhttpIp($task_key) {
         //一分钟内不允许再次获取ip
         $rdobj = $this->initRedis();
-        $ip_request = $rdobj->get('ip_request');
+        $ip_request_key = 'ip_request_'. $task_key;
+        $ip_request = $rdobj->get($ip_request_key);
         if ($ip_request) return false;
 
-        $rdobj->setex('ip_request', 60, 1);
+        $rdobj->setex($ip_request_key, 60, 1);
 
         $url = "http://proxy.siyetian.com/apis_get.html?token=AesJWLNp2a65kaJdXTqFFeNpWT35ERNpnTn1STqFUeORUR31kaNh3TUl0dPRUQy4ERJdXT6lVN.QMxkTO0MjM0cTM&limit=1&type=1&time=&data_format=json&showTimeEnd=true";
         $jsoninfo = $this->getCUrl($url);
 
-        $this->writeLog('spider', 'getZmhttpIp.log', $task_key . '|' . $jsoninfo);
+        $this->writeLog('spider', 'get_zmhttp_ip.log', $task_key . '|' . $jsoninfo);
 
         $data = json_decode($jsoninfo, true);
         //套餐用完 需要换套餐了
         if ($data['code'] == 10019 || $data['code'] == 10005) {
             $this->sendMail(array('1026652509@qq.com'), date('Y年m月d日', time()) . '-微博数据抓取', $data['info']);
+            if($task_key == _RC_SPIDER_WEIBO_COMMENT_USE_IP) $this->sendMail(array('1432334894@qq.com'), date('Y年m月d日', time()) . '-微博评论数据抓取', $data['info']);
             return false;
         }
 
@@ -44,8 +48,8 @@ class mSpider extends mBase {
         return array('ip' => $ip, 'port' => $port);
     }
 
-    public function getRequest($url, $headers = array(), $timeout = 30) {
-        if (empty($this->proxy_ip) || empty($this->proxy_port)) {
+    public function getRequest($url, $headers = array(), $timeout = 30, $use_proxy = true) {
+        if ((empty($this->proxy_ip) || empty($this->proxy_port)) && $use_proxy) {
             $this->setError('配置代理后再来请求吧');
             return false;
         }
@@ -55,7 +59,6 @@ class mSpider extends mBase {
         curl_setopt($ch, CURLOPT_URL, $url);
         curl_setopt($ch, CURLOPT_POST, false);
 
-
         curl_setopt($ch, CURLOPT_RETURNTRANSFER, 1);
         curl_setopt($ch, CURLOPT_REFERER, $url);
 
@@ -64,7 +67,7 @@ class mSpider extends mBase {
         curl_setopt($ch, CURLOPT_HTTP_VERSION, CURL_HTTP_VERSION_1_1);
         curl_setopt($ch, CURLOPT_TIMEOUT, $timeout);
 
-        if ($this->proxy_port) {
+        if ($this->proxy_port && $use_proxy) {
             curl_setopt($ch, CURLOPT_PROXY, 'socks5h://' . $this->proxy_ip . ":" . $this->proxy_port);
         }
 
@@ -82,7 +85,6 @@ class mSpider extends mBase {
 
         $result = curl_exec($ch);
         $this->curl_err = curl_error($ch);
-
         curl_close($ch);
 
         if ($result !== false) return $result;
@@ -328,10 +330,63 @@ class mSpider extends mBase {
             return true;
         }
 
+        preg_match('/Connection timed out after/', $this->curl_err, $match);
+        if ($match[0]) {
+            $this->is_change_ip = true;
+            return true;
+        }
+
         return true;
     }
 
     private function remove4ByteChars($str) {
         return preg_replace('/[\xF0-\xF7][\x80-\xBF]{3}/', '', $str);
     }
+
+    public function spiderComments($cookie, $uid, $wid, $mblogid, $max_id = 0) {
+        $url = "https://weibo.com/ajax/statuses/buildComments?flow=1&is_reload=1&id={$wid}&is_show_bulletin=2&is_mix=0&count=10&uid={$uid}&fetch_level=0&locale=zh-CN";
+        if($max_id > 0) $url = "https://weibo.com/ajax/statuses/buildComments?flow=1&is_reload=1&id={$wid}&is_show_bulletin=2&is_mix=0&max_id={$max_id}&count=10&uid={$uid}&fetch_level=0&locale=zh-CN";
+
+
+        $headers = array(
+            "Cookie: SUB={$cookie}",
+            "x-requested-with: XMLHttpRequest",
+            "User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/136.0.0.0 Safari/537.36",
+            "Referer: https://weibo.com/{$uid}/{$mblogid}"
+        );
+
+        $headers = array(
+            "Cookie: SUB={$cookie}",
+            "x-requested-with: XMLHttpRequest",
+            "User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/138.0.0.0 Safari/537.36",
+            "Referer: https://weibo.com/{$uid}/{$mblogid}"
+        );
+
+        return $this->requestComment($uid, $url, $headers);
+    }
+
+    public function requestComment($uid, $url, $headers) {
+        $res = $this->getRequest($url, $headers, 30, true);
+        if (strpos($res, '登录 - 微博')) {
+            $this->setError('cookie失效');
+            return false;
+        }
+
+        if (empty($res)) {
+            if ($this->curl_err) {
+                $this->getSpiderErrors();
+                $this->setError($this->curl_err);
+            }
+            return false;
+        }
+
+        $data = json_decode($res, true);
+        if (!$data) {
+            $this->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '请求失败:'.$res.'|'.$uid.'|'.$url.'|'.$data);
+            $this->setError('请求为空');
+            return false;
+        }
+
+        return $res;
+    }
 }

model/mWeibo.php

@@ -14,7 +14,7 @@ class mWeibo extends mBase {
         $this->tbl = 'spider_weibo';
     }
 
-    public function getWeiboList($condition, $page_num, $page_size) {
+    public function getWeiboList($condition, $page_num, $page_size, $is_all_data = true) {
         $offset = ($page_num - 1) * $page_size;
 
         $where = "1=1 ";
@@ -32,6 +32,8 @@ class mWeibo extends mBase {
         $list = $this->obj->selectAll($this->tbl, array('sql' => $where, 'vals' => array()), 'created_at desc', array($offset, $page_size));
         if (empty($list)) return array();
 
+        if(!$is_all_data) return $list;
+
         foreach ($list as &$v) {
             $v['name'] = '微博粉丝';
             if ($v['uid'] == 2282201403) $v['name'] = '张宝旬';

model/mWeiboComments.php

@@ -24,14 +24,9 @@ class mWeiboComments extends mBase {
         return $this->obj->select($this->tbl, array('sql' => '`weibo_data_id`=?', 'vals' => array($weibo_data_id)));
     }
 
-    public function isNeedInsertData($weibo_id, $max_weibo_data_id) {
-        $max_weibo_data_id_info = $this->getCommentByWeiboId($max_weibo_data_id, 1, 1);
-        if(empty($max_weibo_data_id_info)) return true;
-
-        $weibo_info = $this->getCommentByWeiboId($weibo_id, 1, 1);
-        if(empty($weibo_info)) return true;
-        if($max_weibo_data_id <= $weibo_info[0]['weibo_data_id']) return false;
-
+    public function isNeedInsertData($weibo_id, $max_weibo_data_id, $comment_count) {
+        $max_weibo_data_id_info = $this->getCommentByWeiboId($max_weibo_data_id, 1, 1, 'id desc');
+        if(empty($max_weibo_data_id_info) || $max_weibo_data_id_info['spider_comment_status']) return true;
         return true;
     }
 
@@ -76,7 +71,7 @@ class mWeiboComments extends mBase {
 
             if(!is_dir($dir)) {
                 $this->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '创建目录失败:'.$source_json_save_path);
-                continue;
+                return false;
             }
 
             if(file_exists($source_json_save_path)) continue;
@@ -84,17 +79,16 @@ class mWeiboComments extends mBase {
             file_put_contents($source_json_save_path, json_encode($comment), LOCK_EX);
             if(!file_exists($source_json_save_path) || filesize($source_json_save_path) < 10) {
                 $this->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '评论保存到文件失败:'.$source_json_save_path);
-                unset($source_data['data'][$key]);
-                continue;
+                return false;
             }
 
             chmod($source_json_save_path, 0755);
             chown($source_json_save_path, 'nobody');
             chgrp($source_json_save_path, 'nobody');
         }
-        if(empty($source_data['data'])) return true;
 
         $sqls = array();
+        $max_id = 0;
         foreach($source_data['data'] as $comment) {
             $content = $comment['text'];
             $weibo_data_id = $comment['id']+0;
@@ -104,6 +98,7 @@ class mWeiboComments extends mBase {
                 'sql' => 'insert into '.$this->tbl.' (`weibo_id`, `weibo_data_id`, `content`, `comment_time`) values (?, ?, ?, ?) ON DUPLICATE KEY UPDATE `weibo_data_id`=?',
                 'vals' => array($weibo_id, $weibo_data_id, $content, $comment_time, $weibo_data_id),
             );
+            $max_id = $weibo_data_id - 1;
         }
 
         $res = $this->obj->execTrans2($sqls);
@@ -113,7 +108,12 @@ class mWeiboComments extends mBase {
             return false;
         }
 
-        return true;
+        $rdata['total_number'] = $source_data['total_number'];
+        $rdata['max_id'] = $max_id;
+        $rdata['weibo_id'] = $weibo_id;
+        $rdata['is_load_all'] = count($source_data['data']) < 20 ? true : false;
+
+        return $rdata;
     }
 
 }

queue/config/daemonconf.php

@@ -1,14 +1,18 @@
 <?php
 
 define('_RC_SPIDER_USE_IP', 'rc_spider_use_ip');
+define('_RC_SPIDER_WEIBO_COMMENT_USE_IP', 'rc_spider_weibo_comment_use_ip');
 define('SPIDER_WEIBO_BEHAVIOR', 'spider_weibo_behavior');
+define('SPIDER_WEIBO_COMMENT', 'spider_weibo_comment');
 
 $GLOBALS['DAEMON_MAXTIME'] = array(
     SPIDER_WEIBO_BEHAVIOR => 7200,
+    SPIDER_WEIBO_COMMENT => 7200,
 );
 
 $GLOBALS['DAEMON_LIST']  = array(
     SPIDER_WEIBO_BEHAVIOR => 'get_weibo_behavior.php',
+    SPIDER_WEIBO_COMMENT => 'spider_weibo_comment.php',
 );
 
 $GLOBALS['DAEMON_LIST_DESC']  = array(
@@ -16,4 +20,5 @@ $GLOBALS['DAEMON_LIST_DESC']  = array(
 
 $GLOBALS['DAEMON_NUMLIMIT']  = array(
     SPIDER_WEIBO_BEHAVIOR => 5,
+    SPIDER_WEIBO_COMMENT => 2,
 );

queue/crontab/add_spider_weibo_comment_task.php

@@ -0,0 +1,38 @@
+<?php
+/**
+ * 抓取微博评论任务
+ * @package crontab
+ */
+
+include_once(dirname(dirname(dirname(__FILE__))) . "/library/publicBase.php");
+
+class addSpiderWeiboCommentTask {
+    public function __construct() {
+        $page_size = 500;
+
+        $condition['uid'] = WEIBO_USER_ZHANG;
+        $condition['spider_comment_status'] = SPIDER_COMMENT_STATUS_NO;
+
+        $obj = new mWeibo();
+        $rdobj = $obj->initRedis();
+        if($rdobj->llen(_RQ_SPIDER_WEIBO_COMMENT)+0 > 50) return true;
+
+        for($page_num=1;$page_num<=1;$page_num++) {
+            $weibo_data = $obj->getWeiboList($condition, $page_num, $page_size);
+            if(empty($weibo_data)) break;
+
+            foreach($weibo_data as $data) {
+                $pdata['weibo_id'] = $data['id'];
+                $pdata['max_id'] = 0;
+
+                if ($rdobj->sIsMember(_RS_SPIDER_WEIBO_COMMENT, $data['id'])) continue;
+                $rdobj->lpush(_RQ_SPIDER_WEIBO_COMMENT, json_encode($pdata));
+                $rdobj->sAdd(_RS_SPIDER_WEIBO_COMMENT, $data['id']);
+            }
+        }
+
+        return true;
+    }
+}
+
+new addSpiderWeiboCommentTask();

queue/deal/spider_weibo_comment.php

@@ -0,0 +1,103 @@
+<?php
+include_once dirname(dirname(__FILE__)).'/base/dealBase.php';
+
+define('PROC_CODE', $argv[2]);
+
+class spiderWeiboComment extends dealBase {
+
+    private $task_key = _RQ_SPIDER_WEIBO_COMMENT;
+
+    public function setPara() {
+        $this->processnum = $GLOBALS['DAEMON_NUMLIMIT'][SPIDER_WEIBO_COMMENT];
+        $this->is_while = true;
+    }
+
+    public function deal() {
+        $obj = new mSpider();
+        $rdobj = $obj->initRedis();
+
+        $ipinfo = $rdobj->get(_RC_SPIDER_WEIBO_COMMENT_USE_IP);
+        $ipdata = json_decode($ipinfo, true);
+        if (!$ipinfo && PROC_CODE == 0 && ($rdobj->llen($this->task_key) + 0 > 0 || !$this->is_while)) $ipdata = $obj->getZmhttpIp(_RC_SPIDER_WEIBO_COMMENT_USE_IP);
+        if (empty($ipdata)) {
+            sleep(2);
+            return false;
+        }
+
+        $obj->proxy_ip = $ipdata['ip'];
+        $obj->proxy_port = $ipdata['port'];
+
+        $rq_data = $rdobj->rpop($this->task_key);
+        if(empty($rq_data)) exit();
+
+        $rq_data = json_decode($rq_data, true);
+
+        $weibo_id = $rq_data['weibo_id']+0;
+        $max_id = $rq_data['max_id']+0;
+        if($weibo_id+0 <= 0 || $max_id < 0) {
+            $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, 'queue:数据格式不正确:'.json_decode( $rq_data));
+            exit;
+        }
+
+        $baselog = $weibo_id . "|" .$max_id . "|" . $ipdata['ip'] . "|" . $ipdata['port'] . "|" . PROC_CODE . "|";
+
+        $wobj = new mWeibo();
+        $weibo_info = $wobj->getWeiboById($weibo_id);
+        if(empty($weibo_info)) {
+            $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '微博信息不存在:'.$baselog);
+            exit;
+        }
+
+        $cookie = $rdobj->get(_RC_WEIBO_LOGIN_COOKIE2);
+        if(empty($cookie)) $cookie = WEIBO_LOGIN_COOKIE2;
+
+        $res = $obj->spiderComments($cookie, $weibo_info['uid'], $weibo_info['wid'], $weibo_info['mblogid'], $max_id);
+        if (!$res) {
+            $error = $obj->getError();
+            $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '队列评论抓取失败，重试中:'.$error.'|'.$baselog);
+
+            if($error == 'cookie失效') {
+                $obj->sendMail(array('1432334894@qq.com'), date('Y年m月d日', time()) . '-微博评论数据抓取', $error);
+                $rdobj->del($this->task_key);
+                $rdobj->del(_RS_SPIDER_WEIBO_COMMENT);
+                exit;
+            }
+
+            $rdobj->lpush($this->task_key, json_encode($rq_data));
+            if ($obj->is_change_ip && PROC_CODE == 0) $rdobj->del(_RC_SPIDER_WEIBO_COMMENT_USE_IP);
+            sleep(30);
+
+            return false;
+        }
+
+        $cobj = new mWeiboComments();
+        $res = $cobj->addComment($weibo_id, $res);
+        if(!$res) {
+            $error = $cobj->getError().'|评论json数据保存到数据库失败~';
+            $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, $baselog.'|'.$error);
+            $wobj->updateWeibo($weibo_id, array('spider_comment_status' => SPIDER_COMMENT_STATUS_ERROR));
+            return false;
+        }
+
+        if(!$res['is_load_all']) {
+            $rdobj->lpush($this->task_key, json_encode($res));
+            return true;
+        }
+
+        $res = $wobj->updateWeibo($weibo_id, array('spider_comment_status' => SPIDER_COMMENT_STATUS_SUCCESS));
+        if(!$res) {
+            $error = $wobj->getError();
+            $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '更新微博状态失败:'.$error.'|'.$baselog);
+            sleep(2);
+            return false;
+        }
+
+        $obj->writeLog(ZHISHIKU_SPIDER_LOG, ZHISHIKU_SPIDER_COMMENT, '抓取成功:'.$baselog);
+        $rdobj->sRem(_RS_SPIDER_WEIBO_COMMENT, $weibo_id);
+        sleep(3);
+
+        return true;
+    }
+}
+
+new spiderWeiboComment();