rov-offline/flowpool_data_update_with_level_v2.py

import time
import traceback
import json
import random

from my_config import set_config
from my_utils import request_post, filter_video_status, send_msg_to_feishu, filter_video_status_app, \
    filter_political_videos
from log import Log
from db_helper import RedisHelper
from odps import ODPS

config_, _ = set_config()
log_ = Log()


def get_videos_from_flow_pool(app_type, size=1000):
    """
    从流量池获取视频，循环获取，直到返回数据为None结束
    :param app_type: 产品标识 type-int
    :param size: 每次获取视频数量，type-int，默认1000
    :return: videos  [{'videoId': 1111, 'flowPool': ''}, ...]
    """
    # 获取批次标识，利用首次获取数据时间戳为标记
    batch_flag = int(time.time()) * 1000 + 111
    log_.info(f"batch_flag: {batch_flag}")
    request_data = {'appType': app_type, 'batchFlag': batch_flag, 'size': size}
    videos = []
    retry = 0
    while True:
        result = request_post(request_url=config_.GET_VIDEOS_FROM_POOL_URL, request_data=request_data)
        if result is None:
            if retry > 2:
                break
            retry += 1
            continue
        if result['code'] != 0:
            log_.info('batch_flag: {}, 获取流量池视频失败'.format(batch_flag))
            if retry > 2:
                break
            retry += 1
            continue
        if not result['data']:
            if retry > 2:
                break
            retry += 1
            continue
        videos.extend(result['data'])
    return videos


def update_remain_view_count(video_info_list):
    """
    获取视频在流量池中的剩余可分发数，并存入对应的redis中
    :param app_type: 产品标识 type-int
    :param video_info_list: 视频信息 (视频id, 流量池标记) type-list，[(video_id, flow_pool), ...]
    :return: data type-list，[(video_id, flow_pool, view_count), ...]
    """
    redis_helper = RedisHelper()
    if not video_info_list:
        return dict()

    remain_videos = dict()
    # 每次请求10个
    for i in range(len(video_info_list)//10 + 1):
        remain_st_time = time.time()
        videos = [{'videoId': info[0], 'flowPool': info[1]} for info in video_info_list[i*10:(i+1)*10]]
        request_data = {'videos': videos}
        result = request_post(request_url=config_.GET_REMAIN_VIEW_COUNT_URL,
                              request_data=request_data, timeout=(0.5, 3))
        log_.info(f"i = {i}, expend time = {(time.time()-remain_st_time)*1000}")
        if result is None:
            continue
        if result['code'] != 0:
            log_.error('获取视频在流量池中的剩余可分发数失败')
            continue
        for item in result['data']:
            if item['distributeCount'] is None:
                continue
            distribute_count = int(item['distributeCount'])
            if distribute_count > 0:
                remain_videos[item['videoId']]=distribute_count
                # 将分发数更新到本地记录
                key_name = f"{config_.LOCAL_DISTRIBUTE_COUNT_PREFIX}{item['videoId']}:{item['flowPool']}"
                redis_helper.set_data_to_redis(key_name=key_name, value=distribute_count, expire_time=25 * 60)
            else:
                # 将本地记录删除
                key_name = f"{config_.LOCAL_DISTRIBUTE_COUNT_PREFIX}{item['videoId']}:{item['flowPool']}"
                redis_helper.del_keys(key_name=key_name)
    log_.info(f"新增加不分发过滤前后整体数量: {len(video_info_list)}:{len(remain_videos)}")
    return remain_videos

def get_flow_pool_recommend_config(flow_pool_id):
    """获取流量池推荐分发配置"""
    result = request_post(request_url=config_.GET_FLOW_POOL_RECOMMEND_CONFIG_URL)
    if result is None:
        return None
    if result['code'] != 0:
        return None
    flow_pool_distribute_config = result['data'].get('flowPoolDistributeConfig')
    if flow_pool_distribute_config:
        if int(eval(flow_pool_distribute_config).get('flowPoolId')) == flow_pool_id:
            return eval(eval(flow_pool_distribute_config).get('distributeRate'))
        else:
            return None
    else:
        return None

def update_flow_pool(flow_pool_id_list):
    """
    获取流量池可分发视频，并将结果上传Redis
    :param app_type: 产品标识 type-int
    :return: None
    """
    # 所有产品都从0取数据
    app_type = 0

    try:
        # 从流量池获取数据
        videos = get_videos_from_flow_pool(app_type=app_type)
        if len(videos) <= 0:
            log_.info('流量池中无需分发的视频')
            return
        # video_id 与 flow_pool, level 进行mapping
        video_ids = set()
        log_.info('流量池中视频数：{}'.format(len(videos)))
        mapping = {}
        for video in videos:
            flow_pool_id = video['flowPoolId'] # 召回使用的切分ID是在这里做的。流量池中的视频是不区分ID的，也不区分层。
            if int(flow_pool_id) not in flow_pool_id_list:
                continue
            # print(f"flow_pool_id: {flow_pool_id}")
            video_id = video['videoId']
            video_ids.add(video_id)
            item_info = {'flowPool': video['flowPool'], 'level': video['level']}
            if video_id in mapping:
                mapping[video_id].append(item_info)
            else:
                mapping[video_id] = [item_info]
        log_.info(f"需更新流量池视频数: {len(video_ids)}")

        # 对视频状态进行过滤
        filtered_videos = filter_video_status(list(video_ids))
        log_.info('filter videos status finished, filtered_videos nums={}'.format(len(filtered_videos)))


        if not filtered_videos:
            log_.info('流量池中视频状态不符合分发')
            return



        # 1. 更新分发数量，过滤分发数量为0的视频
        video_info_list = []
        for video_id in filtered_videos:
            for item in mapping.get(video_id):
                flow_pool = item['flowPool']
                video_info = (video_id, flow_pool)
                if video_info not in video_info_list:
                    video_info_list.append(video_info)

        log_.info(f"video_info_list count = {len(video_info_list)}")
        remain_videos = update_remain_view_count(video_info_list)
        if not remain_videos:
            log_.info('流量池中视频状态不符合分发')
            return


        # 上传数据到redis
        quick_flow_pool_redis_data = set()
        flow_pool_redis_data = dict()
        level_weight = dict()
        for video_id, distribute_count in remain_videos.items():
            for item in mapping.get(video_id):
                flow_pool = item['flowPool']
                level = item['level']
                # 判断是否为快速曝光流量池视频
                value = '{}-{}'.format(video_id, flow_pool)
                flow_pool_id = int(flow_pool.split('#')[0])  # flowPool: 流量池ID#分级ID#级别Level#生命周期ID
                if flow_pool_id == config_.QUICK_FLOW_POOL_ID:
                    quick_flow_pool_redis_data.add(value)
                else:
                    if level not in flow_pool_redis_data:
                        flow_pool_redis_data[level] = set()
                        level_weight[level] = 0
                    flow_pool_redis_data[level].add(value)
                    level_weight[level]=level_weight[level]+distribute_count


        # 2. quick曝光池
        redis_helper = RedisHelper()
        quick_flow_pool_key_name = f"{config_.QUICK_FLOWPOOL_KEY_NAME_PREFIX_SET}{app_type}:{config_.QUICK_FLOW_POOL_ID}"
        # 如果key已存在，删除key
        if redis_helper.key_exists(quick_flow_pool_key_name):
            redis_helper.del_keys(quick_flow_pool_key_name)
        if quick_flow_pool_redis_data:
            log_.info(f"quick_flow_pool_redis_data = {quick_flow_pool_redis_data}")
            redis_helper.add_data_with_set(key_name=quick_flow_pool_key_name, values=quick_flow_pool_redis_data,
                                           expire_time=24 * 3600)
            # 快速流量池分发概率存入redis
            distribute_rate_key_name = f"{config_.QUICK_FLOWPOOL_DISTRIBUTE_RATE_KEY_NAME_PREFIX}{config_.QUICK_FLOW_POOL_ID}"
            distribute_rate = get_flow_pool_recommend_config(flow_pool_id=config_.QUICK_FLOW_POOL_ID)
            if distribute_rate is not None:
                redis_helper.set_data_to_redis(key_name=distribute_rate_key_name, value=distribute_rate,
                                               expire_time=15 * 60)

        # 3. 普通流量池
        for level, videos in flow_pool_redis_data.items():
            log_.info(f"level: {level}, videos_count: {len(videos)}")
            flow_pool_key_name = f"flow:pool:level:item:v2:{app_type}:{level}"
            # 如果key已存在，删除key
            if redis_helper.key_exists(flow_pool_key_name):
                redis_helper.del_keys(flow_pool_key_name)
            # 写入redis
            if videos:
                redis_helper.add_data_with_set(key_name=flow_pool_key_name, values=videos, expire_time=24 * 3600)
                result = redis_helper.get_data_from_set(flow_pool_key_name)
                if not result:
                    result = []
                size = len(result)
                log_.info(f'写入成功key={flow_pool_key_name}:{size}')


        # 4. 写入权重
        weight=json.dumps(level_weight)
        redis_helper.set_data_to_redis(key_name="flow:pool:level:weight:v2", value=weight, expire_time=24*3600)
        log_.info(f'权重 {weight}')


        log_.info('data to redis finished!')


    except Exception as e:
        log_.error('流量池更新失败, appType: {} exception: {}, traceback: {}'.format(
            app_type, e, traceback.format_exc()))
        send_msg_to_feishu(
            webhook=config_.FEISHU_ROBOT['server_robot'].get('webhook'),
            key_word=config_.FEISHU_ROBOT['server_robot'].get('key_word'),
            msg_text='rov-offline{} - 流量池更新失败, appType: {}, exception: {}'.format(config_.ENV_TEXT, app_type, e)
        )

if __name__ == '__main__':
    st_time = time.time()
    # 为避免第一个app_type获取数据不全，等待1min
    log_.info('flow pool predict start...')
    # 获取对应流量池id列表
    redis_helper = RedisHelper()
    flow_pool_abtest_config = redis_helper.get_data_from_redis(key_name=config_.FLOWPOOL_ABTEST_KEY_NAME)
    if flow_pool_abtest_config is not None:
        flow_pool_abtest_config = json.loads(flow_pool_abtest_config)
    else:
        flow_pool_abtest_config = {}
    flow_pool_id_list = flow_pool_abtest_config.get('experimental_flow_set_level', [])
    log_.info('predict start...')
    update_flow_pool(flow_pool_id_list=flow_pool_id_list)
    log_.info('predict end...')
    log_.info(f"expend time = {(time.time() - st_time) * 1000}ms")
    log_.info('flow pool predict end...')

# python flowpool_data_update_with_level.py    测试环境必须手动执行python 才能有数据