Merge branch 'master' into feature/zhangbo_flow_recall

flowpool_data_update_with_level_supply.py

@@ -0,0 +1,255 @@
+import datetime
+import random
+import time
+import os
+import traceback
+import random
+import json
+
+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
+    :param isSupply: 是否为供给流量池 1 是 0 否
+    :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())
+    request_data = {'appType': app_type, 'batchFlag': batch_flag, 'size': size, 'isSupply': 1}
+    videos = []
+    retry = 0
+    while True:
+        print(config_.GET_VIDEOS_FROM_POOL_URL)
+        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('supply 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 get_videos_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
+
+    # 每次请求10个
+    bu_fen_fa_cnt = 0
+    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"supply i = {i}, expend time = {(time.time()-remain_st_time)*1000}")
+        if result is None:
+            continue
+        if result['code'] != 0:
+            log_.error('supply 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:
+                # 将分发数更新到本地记录
+                key_name = f"{config_.SUPPLY_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_.SUPPLY_LOCAL_DISTRIBUTE_COUNT_PREFIX}{item['videoId']}:{item['flowPool']}"
+                redis_helper.del_keys(key_name=key_name)
+                # 从流量召回池移除
+                value = '{}-{}'.format(item['videoId'], item['flowPool'])
+                for type_name in config_.APP_TYPE:
+                    for level in range(1, 7):
+                        flow_pool_key = \
+                            f"{config_.FLOWPOOL_KEY_NAME_PREFIX_SET_LEVEL_SUPPLY}{config_.APP_TYPE.get(type_name)}:{level}"
+                        redis_helper.remove_value_from_set(key_name=flow_pool_key, values=(value, ))
+                bu_fen_fa_cnt = bu_fen_fa_cnt + 1
+    log_.info(f"supply 新增加不分发过滤前后整体数量: {len(video_info_list)}:{str(bu_fen_fa_cnt)}")
+
+
+def get_flow_pool_data(app_type, video_info_list, 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
+    """
+    try:
+        # 从流量池获取数据
+        videos = get_videos_from_flow_pool(app_type=app_type)
+        log_.info(f"supply app_type: {app_type}")
+        # log_.info(f"videos: {videos}")
+        if len(videos) <= 0:
+            log_.info('supply 流量池中无需分发的视频')
+            return video_info_list
+        # video_id 与 flow_pool, level 进行mapping
+        video_ids = set()
+        log_.info('supply 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"supply 需更新流量池视频数: {len(video_ids)}")
+
+        # 对视频状态进行过滤
+        if app_type == config_.APP_TYPE['APP']:
+            filtered_videos = filter_video_status_app(list(video_ids))
+        else:
+            filtered_videos = filter_video_status(list(video_ids))
+        log_.info('supply filter videos status finished, filtered_videos nums={}'.format(len(filtered_videos)))
+
+        # 涉政视频过滤
+        if app_type not in config_.POLITICAL_RECOMMEND_APP_TYPE_LIST:
+            filtered_videos = filter_political_videos(video_ids=filtered_videos)
+
+        if not filtered_videos:
+            log_.info('supply 流量池中视频状态不符合分发')
+            return video_info_list
+
+        # 上传数据到redis
+        # 普通流量池视频按照层级存储
+        redis_data = dict()
+
+        for video_id in filtered_videos:
+            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)
+                if level not in redis_data:
+                    redis_data[level] = set()
+                redis_data[level].add(value)
+
+                video_info = (video_id, flow_pool)
+                if video_info not in video_info_list:
+                    video_info_list.append(video_info)
+
+        redis_helper = RedisHelper()
+
+        # 普通流量池视频写入redis - 分层存储
+        level_list = []
+        for level, videos in redis_data.items():
+            log_.info(f"supply level: {level}, videos_count: {len(videos)}")
+            level_list.append(level)
+            flow_pool_key_name = f"{config_.FLOWPOOL_KEY_NAME_PREFIX_SET_LEVEL_SUPPLY}{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)
+
+        # 删除此时不存在的level key
+        for i in range(1, 7):
+            if i not in level_list:
+                flow_pool_key_name = f"{config_.FLOWPOOL_KEY_NAME_PREFIX_SET_LEVEL_SUPPLY}{app_type}:{i}"
+                # 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)
+
+        log_.info('supply data to redis finished!')
+
+        return video_info_list
+
+    except Exception as e:
+        log_.error('supply 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{} - supply 流量池更新失败, appType: {}, exception: {}'.format(config_.ENV_TEXT, app_type, e)
+        )
+        return video_info_list
+
+
+def get_data_from_odps(project, sql):
+    """检查数据是否准备好"""
+    odps = ODPS(
+        access_id=config_.ODPS_CONFIG['ACCESSID'],
+        secret_access_key=config_.ODPS_CONFIG['ACCESSKEY'],
+        project=project,
+        endpoint=config_.ODPS_CONFIG['ENDPOINT'],
+        connect_timeout=3000,
+        read_timeout=500000,
+        pool_maxsize=1000,
+        pool_connections=1000
+    )
+
+    try:
+        with odps.execute_sql(sql=sql).open_reader() as reader:
+            data_df = reader.to_pandas()
+    except Exception as e:
+        data_df = None
+    return data_df
+
+
+if __name__ == '__main__':
+    st_time = time.time()
+    # 为避免第一个app_type获取数据不全，等待1min
+    time.sleep(60)
+    log_.info('supply 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('supply_flow_set_level', [])
+    log_.info('supply flow_pool_id_list = {} '.format(flow_pool_id_list))
+    video_info_list = []
+    for app_name, app_type in config_.SUPPLY_APP_TYPE.items():
+        log_.info('{} supply app_type start...'.format(app_name))
+        video_info_list = get_flow_pool_data(app_type=app_type, video_info_list=video_info_list,
+                                                 flow_pool_id_list=flow_pool_id_list)
+        log_.info('{} supply app_type end...'.format(app_name))
+
+    # 更新剩余分发数
+    log_.info(f"supply video_info_list count = {len(video_info_list)}")
+    get_videos_remain_view_count(video_info_list)
+    log_.info('supply flow pool predict end...')
+    log_.info(f"supply expend time = {(time.time() - st_time) * 1000}ms")
+
+# python flowpool_data_update_with_level.py    测试环境必须手动执行python 才能有数据

flowpool_data_update_with_level_supply_task.sh

@@ -0,0 +1,11 @@
+source /etc/profile
+echo $ROV_OFFLINE_ENV
+if [[ $ROV_OFFLINE_ENV == 'test' ]]; then
+#    cd /data2/rov-offline && /root/anaconda3/bin/python /data2/rov-offline/pool_predict.py &
+#    cd /data2/rov-offline && /root/anaconda3/bin/python /data2/rov-offline/flowpool_data_update.py
+    cd /data2/rov-offline && /root/anaconda3/bin/python /data2/rov-offline/flowpool_data_update_with_level_supply.py
+elif [[ $ROV_OFFLINE_ENV == 'pro' ]]; then
+#    cd /data/rov-offline && /root/anaconda3/bin/python /data/rov-offline/pool_predict.py
+#    cd /data/rov-offline && /root/anaconda3/bin/python /data/rov-offline/flowpool_data_update.py
+    cd /data/rov-offline && /root/anaconda3/bin/python /data/rov-offline/flowpool_data_update_with_level_supply.py
+fi

my_config.py

@@ -19,6 +19,11 @@ class BaseConfig(object):
         'JOURNEY': 22,  # 票圈足迹
         'BLESSING_YEAR': 3,  # 票圈福年
         'PIAO_QUAN_BLESSING': 2,  # 票圈 | 祝福
+        'PIAO_QUAN_HAOKAN_VIDEO': 23  # 票圈好看视频
+    }
+    # 供给质量评估 appType
+    SUPPLY_APP_TYPE = {
+        'PIAO_QUAN_HAOKAN_VIDEO': 23
     }
     # 数据存放路径
     DATA_DIR_PATH = './data'
@@ -664,6 +669,8 @@ class BaseConfig(object):
     QUICK_FLOWPOOL_KEY_NAME_PREFIX_SET = 'flow:pool:quick:item:'
     # 流量池视频分层存放 redis key前缀，完整格式 flow:pool:level:item:{appType}:{level}
     FLOWPOOL_KEY_NAME_PREFIX_SET_LEVEL = 'flow:pool:level:item:'
+    # 流量池视频分层存放 redis key前缀，完整格式 flow:pool:level:item:supply:{appType}:{level}
+    FLOWPOOL_KEY_NAME_PREFIX_SET_LEVEL_SUPPLY = 'flow:pool:level:item:supply:'
     # 流量池各层分发概率权重存放 redis key，完整格式 flow:pool:level:recommend:weight
     FLOWPOOL_LEVEL_WEIGHT_KEY_NAME = 'flow:pool:level:recommend:weight'
     # 流量池视频分层存放(按分数排序) redis key前缀，完整格式 flow:pool:level:item:score:{appType}:{level}
@@ -672,6 +679,7 @@ class BaseConfig(object):
     QUICK_FLOWPOOL_DISTRIBUTE_RATE_KEY_NAME_PREFIX = 'flow:pool:quick:distribute:rate:'
     # 本地记录视频的可分发数，控制分发，完整key格式：flow:pool:local:distribute:count:{videoId}:{flowPool}
     LOCAL_DISTRIBUTE_COUNT_PREFIX = 'flow:pool:local:distribute:count:'
+    SUPPLY_LOCAL_DISTRIBUTE_COUNT_PREFIX = 'flow:pool:supply:local:distribute:count:'
 
     # 流量池videoId redis key前缀，完整格式 flow:pool:video:ids:{appType}
     FLOWPOOL_VIDEO_ID_KEY_NAME_PREFIX = 'flow:pool:video:ids:'