piaoquan_crawler/common/video_item.py

import time

from common.public import clean_title
from .aliyun_log import AliyunLogger


class VideoItem(object):
    """
    function: 当扫描进一条视频的时候，对该视频的基本信息进行处理，保证发送给 pipeline和 etl 的 video_dict 是正确的
    __init__: 初始化空json 对象，用来存储视频信息
    add_video_info: 把视频信息存储到 item 对象中
    check_item: 检查 item 对象中的各个元素以及处理
    """

    def __init__(self):
        self.item = {}

    def add_video_info(self, key, value):
        self.item[key] = value

    def check_item(self):
        """
        判断item 里面的字段，是否符合要求
        字段分为 3 类：
        1. 必须存在数据的字段： ["video_id", "user_id", "user_name", "out_user_id", "out_video_id", "session", "video_url", "cover_url", "platform", "strategy"]
        2. 不存在默认为 0 的字段 ：["duration", "play_cnt", "like_cnt", "comment_cnt", "share_cnt", "width", "height"]
        3. 需要后出理的字段： video_title, publish_time
        """
        if self.item.get("video_title"):
            self.item["video_title"] = clean_title(self.item["video_title"])
        else:
            return False
        if self.item.get("publish_time_stamp"):
            publish_time_str = time.strftime(
                "%Y-%m-%d %H:%M:%S", time.localtime(self.item["publish_time_stamp"])
            )
            self.add_video_info("publish_time_str", publish_time_str)
        else:
            publish_time_stamp = int(time.time())
            publish_time_str = time.strftime(
                "%Y-%m-%d %H:%M:%S", time.localtime(publish_time_stamp)
            )
            self.add_video_info("publish_time_stamp", publish_time_stamp)
            self.add_video_info("publish_time_str", publish_time_str)
        self.add_video_info("publish_time", publish_time_str)
        if not self.item.get("update_time_stamp"):
            self.add_video_info("update_time_stamp", int(time.time()))

        # 如果不存在，默认值为 0
        config_keys = [
            "duration",
            "play_cnt",
            "like_cnt",
            "comment_cnt",
            "share_cnt",
            "width",
            "height",
        ]
        for config_key in config_keys:
            if self.item.get(config_key):
                continue
            else:
                self.add_video_info(config_key, 0)

        # 必须存在的元素，若不存在则会报错
        must_keys = [
            "video_id",
            "user_id",
            "user_name",
            "out_video_id",
            "session",
            "video_url",
            "cover_url",
            "platform",
            "strategy",
        ]
        """
        video_id, out_video_id 均为站外视频 id
        usr_id: 站内用户 id
        out_user_id: 站外用户 id
        user_name: 站外用户名称
        """
        for m_key in must_keys:
            if self.item.get(m_key):
                continue
            else:
                # print(m_key)
                return False
        return True

    def produce_item(self):
        flag = self.check_item()
        if flag:
            return self.item
        else:
            return False