LongArticlesJob/coldStartTasks/crawler/weixinCategoryCrawler.py

"""
@author: luojunhui
抓取全局品类文章
"""

import time

from tqdm import tqdm

from applications import WeixinSpider, Functions, longArticlesMySQL

# 常量
ACCOUNT_GOOD_STATUS = 1
DEFAULT_VIEW_COUNT = 0
DEFAULT_LIKE_COUNT = 0
DEFAULT_ARTICLE_STATUS = 1
DEFAULT_TIMESTAMP = 1704038400


class weixinCategory(object):
    """
    微信全局品类账号抓取
    """

    def __init__(self):
        self.db_client_lam = longArticlesMySQL()
        self.spider = WeixinSpider()
        self.function = Functions()

    def get_account_list(self, account_category):
        """
        获取账号
        :param account_category 品类
        :return:
        """
        sql = f"""
            select gh_id, account_source, account_name, account_category, latest_update_time
            from long_articles_accounts 
            where account_category = '{account_category}' and is_using = {ACCOUNT_GOOD_STATUS};
            """
        account_tuple = self.db_client_lam.select(sql)
        result = [
            {
                "gh_id": i[0],
                "platform": i[1],
                "account_name": i[2],
                "category": i[3],
                "latest_timestamp": i[4],
            }
            for i in account_tuple
        ]
        return result

    def insert_data_into_db(self, gh_id, category, article_list):
        """
        将数据更新到数据库
        :return:
        """
        for article_obj in article_list:
            detail_article_list = article_obj["AppMsg"]["DetailInfo"]
            for obj in detail_article_list:
                try:
                    show_stat = self.function.show_desc_to_sta(obj["ShowDesc"])
                    show_view_count = show_stat.get("show_view_count", DEFAULT_VIEW_COUNT)
                    show_like_count = show_stat.get("show_like_count", DEFAULT_LIKE_COUNT)
                    insert_sql = f"""
                        insert into crawler_meta_article
                        (platform, mode, category, out_account_id, article_index, title, link, read_cnt, like_cnt, description, publish_time, crawler_time, status, unique_index)
                        VALUES 
                        (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s);
                    """
                    self.db_client_lam.update(
                        sql=insert_sql,
                        params=(
                            "weixin",
                            "account",
                            category,
                            gh_id,
                            obj['ItemIndex'],
                            obj["Title"],
                            obj["ContentUrl"],
                            show_view_count,
                            show_like_count,
                            obj["Digest"],
                            obj["send_time"],
                            int(time.time()),
                            DEFAULT_ARTICLE_STATUS,
                            self.function.generateGzhId(obj["ContentUrl"]),
                        ),
                    )
                except Exception as e:
                    print(e)

    def update_latest_account_timestamp(self, gh_id):
        """
        更新账号的最新时间戳
        :return:
        """
        select_sql = f"""
            SELECT publish_time 
            From crawler_meta_article 
            WHERE out_account_id = '{gh_id}'
            ORDER BY publish_time DESC LIMIT 1;
        """
        result = self.db_client_lam.select(select_sql)
        time_stamp = result[0][0]
        dt_str = self.function.time_stamp_to_str(time_stamp)
        update_sql = f"""
            update long_articles_accounts
            set latest_update_time = %s
            where gh_id = %s;
        """
        self.db_client_lam.update(sql=update_sql, params=(dt_str, gh_id))

    def update_each_account(self, gh_id, category, latest_time_stamp, index=None):
        """
        更新账号文章
        :return:
        """
        response = self.spider.update_msg_list(ghId=gh_id, index=index)
        msg_list = response.get("data", {}).get("data")
        if msg_list:
            last_article_in_this_msg = msg_list[-1]
            self.insert_data_into_db(
                gh_id=gh_id, category=category, article_list=msg_list
            )
            last_time_stamp_in_this_msg = last_article_in_this_msg["AppMsg"]["BaseInfo"]["UpdateTime"]
            if latest_time_stamp < last_time_stamp_in_this_msg:
                next_cursor = response["data"]["next_cursor"]
                return self.update_each_account(
                    gh_id=gh_id,
                    latest_time_stamp=latest_time_stamp,
                    category=category,
                    index=next_cursor,
                )
            else:
                # 更新最近抓取时间
                self.update_latest_account_timestamp(gh_id=gh_id)
                print("账号时间更新成功")
        else:
            print("No more data")

    def deal(self, category_list):
        """

        :param category_list:
        :return:
        """
        for category in category_list:
            account_list = self.get_account_list(category)
            for account in tqdm(account_list):
                try:
                    gh_id = account['gh_id']
                    category = account['category']
                    try:
                        timestamp = int(account['latest_timestamp'].timestamp())
                    except Exception as e:
                        timestamp = DEFAULT_TIMESTAMP
                    self.update_each_account(
                        gh_id=gh_id,
                        category=category,
                        latest_time_stamp=timestamp
                    )
                    print("success")
                except Exception as e:
                    print("fail because of {}".format(e))

    def deal_accounts(self, account_list):
        """
        input account list
        :param account_list:
        :return:
        """
        account_tuple = tuple(account_list)
        sql = f"""
            SELECT gh_id, account_name, account_category, latest_update_time 
            FROM long_articles_accounts
            WHERE account_name in {account_tuple};
        """
        response = self.db_client_lam.select(sql)
        for account in tqdm(response):
            try:
                gh_id = account[0]
                category = account[2]
                try:
                    latest_timestamp = account[3].timestamp()
                except Exception as e:
                    latest_timestamp = DEFAULT_TIMESTAMP
                self.update_each_account(
                    gh_id=gh_id,
                    category=category,
                    latest_time_stamp=latest_timestamp
                )
            except Exception as e:
                print(e)