add recall-update-day

config.py

@@ -94,6 +94,19 @@ class BaseConfig(object):
         '19': 'predict_data_19.pickle'
     }
 
+    # 天级规则更新使用数据
+    PROJECT_DAY = 'loghubods'
+    TABLE_DAY = 'video_data_each_day_dataset'
+
+    # 小程序天级规则参数
+    RULE_PARAMS_DAY = {
+        'rule1': {'return_count': 200},
+        # 'rule2': {'return_count': 20, 'score_rule': 0.001},
+        # 'rule3': {'view_type': 'pre-view', 'return_count': 20, 'score_rule': 0.005},
+        # 'rule4': {'cal_score_func': 2, 'return_count': 20, 'score_rule': 0},
+        # 'rule5': {'cal_score_func': 3, 'return_count': 20, 'score_rule': 0},
+    }
+
     # 小程序离线ROV模型结果存放 redis key前缀，完整格式：com.weiqu.video.recall.hot.item.score.{date}
     RECALL_KEY_NAME_PREFIX = 'com.weiqu.video.recall.hot.item.score.'
     # 小程序小时级更新结果存放 redis key前缀，完整格式：com.weiqu.video.recall.item.score.h.{rule_key}.{date}.{h}
@@ -104,6 +117,16 @@ class BaseConfig(object):
     # 小时级视频状态不符合推荐要求的列表 redis key，完整格式：com.weiqu.video.filter.h.item.{rule_key}
     H_VIDEO_FILER = 'com.weiqu.video.filter.h.item.'
 
+    # 小程序天级更新结果存放 redis key前缀，完整格式：com.weiqu.video.recall.item.score.day.{rule_key}.{date}
+    RECALL_KEY_NAME_PREFIX_BY_DAY = 'com.weiqu.video.recall.item.score.day.'
+    # 小程序离线ROV模型结果与小程序天级更新结果去重后 存放 redis key前缀，
+    # 完整格式：com.weiqu.video.recall.hot.item.score.dup.day.now.{rule_key}.{date}
+    RECALL_KEY_NAME_PREFIX_DUP_DAY_NOW = 'com.weiqu.video.recall.hot.item.score.dup.day.now.'
+    # 使用前一天小程序离线ROV模型结果与小程序天级更新结果去重后 存放 redis key前缀，
+    # 完整格式：com.weiqu.video.recall.hot.item.score.dup.day.pre.{rule_key}.{date}
+    RECALL_KEY_NAME_PREFIX_DUP_DAY_PRE = 'com.weiqu.video.recall.hot.item.score.dup.day.pre.'
+
+
     # app应用 小程序离线ROV模型结果存放 redis key前缀，完整格式：com.weiqu.video.recall.hot.item.score.app.{date}
     RECALL_KEY_NAME_PREFIX_APP = 'com.weiqu.video.recall.hot.item.score.app.'
     # app应用 运营提供的小时级数据存放 redis key前缀，完整格式：com.weiqu.video.app.op.item.score.{date}.{h}

rule_rank_day.py

@@ -0,0 +1,201 @@
+import pandas as pd
+from odps import ODPS
+from datetime import datetime, timedelta
+from threading import Timer
+from utils import get_data_from_odps
+from db_helper import RedisHelper
+from config import set_config
+from log import Log
+
+config_, _ = set_config()
+log_ = Log()
+
+features = [
+    'videoid',
+    'preview人数',  # 过去1天预曝光人数
+    'view人数',  # 过去1天曝光人数
+    'play人数',  # 过去1天播放人数
+    'share人数',  # 过去1天分享人数
+    '回流人数',  # 过去1天分享，过去1天回流人数
+    'preview次数',  # 过去1天预曝光次数
+    'view次数',  # 过去1天曝光次数
+    'play次数',  # 过去1天播放次数
+    'share次数',  # 过去1天分享次数
+]
+
+
+def get_rov_redis_key(now_date):
+    # 获取rov模型结果存放key
+    redis_helper = RedisHelper()
+    now_dt = datetime.strftime(now_date, '%Y%m%d')
+    key_name = f'{config_.RECALL_KEY_NAME_PREFIX}{now_dt}'
+    key_dt = now_dt
+    if not redis_helper.key_exists(key_name=key_name):
+        pre_dt = datetime.strftime(now_date - timedelta(days=1), '%Y%m%d')
+        key_name = f'{config_.RECALL_KEY_NAME_PREFIX}{pre_dt}'
+        key_dt = pre_dt
+    return key_name, key_dt
+
+
+def day_data_check(project, table, now_date):
+    """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:
+        dt = datetime.strftime((now_date - timedelta(days=1)), '%Y%m%d')
+        sql = f'select * from {project}.{table} where dt = {dt}'
+        with odps.execute_sql(sql=sql).open_reader() as reader:
+            data_count = reader.count
+    except Exception as e:
+        data_count = 0
+    return data_count
+
+
+def get_feature_data(now_date, project, table):
+    """获取特征数据"""
+    dt = datetime.strftime((now_date - timedelta(days=1)), '%Y%m%d')
+    # dt = '20220425'
+    records = get_data_from_odps(date=dt, project=project, table=table)
+    feature_data = []
+    for record in records:
+        item = {}
+        for feature_name in features:
+            item[feature_name] = record[feature_name]
+        feature_data.append(item)
+    feature_df = pd.DataFrame(feature_data)
+    return feature_df
+
+
+def cal_score1(df):
+    # score1计算公式: score = 回流人数/(view人数+10000)
+    df = df.fillna(0)
+    df['score'] = df['回流人数'] / (df['view人数'] + 1000)
+    df = df.sort_values(by=['score'], ascending=False)
+    return df
+
+
+def cal_score2(df):
+    pass
+
+
+def video_rank_day(df, now_date, rule_key, param):
+    """
+    获取符合进入召回源条件的视频，与每日更新的rov模型结果视频列表进行合并
+    :param df:
+    :param now_date:
+    :param rule_key: 天级规则数据进入条件
+    :param param: 天级规则数据进入条件参数
+    :return:
+    """
+    # 获取rov模型结果
+    redis_helper = RedisHelper()
+    key_name, key_dt = get_rov_redis_key(now_date=now_date)
+    initial_data = redis_helper.get_data_zset_with_index(key_name=key_name, start=0, end=-1, with_scores=True)
+    log_.info(f'initial data count = {len(initial_data)}, key_dt = {key_dt}')
+
+    # 获取符合进入召回源条件的视频
+    return_count = param.get('return_count')
+    h_recall_df = df[df['回流人数'] > return_count]
+    h_recall_videos = h_recall_df['videoid'].to_list()
+    log_.info(f'h_recall videos count = {len(h_recall_videos)}')
+    # 写入对应的redis
+    h_video_ids =[]
+    h_recall_result = {}
+    for video_id in h_recall_videos:
+        score = h_recall_df[h_recall_df['videoid'] == video_id]['score']
+        h_recall_result[int(video_id)] = float(score)
+        h_video_ids.append(int(video_id))
+    h_recall_key_name = \
+        f"{config_.RECALL_KEY_NAME_PREFIX_BY_DAY}{rule_key}.{datetime.strftime(now_date, '%Y%m%d')}"
+    if len(h_recall_result) > 0:
+        redis_helper.add_data_with_zset(key_name=h_recall_key_name, data=h_recall_result, expire_time=7 * 24 * 3600)
+
+    # 去重更新rov模型结果，并另存为redis中
+    initial_data_dup = {}
+    for video_id, score in initial_data:
+        if int(video_id) not in h_video_ids:
+            initial_data_dup[int(video_id)] = score
+    log_.info(f"initial data dup count = {len(initial_data_dup)}")
+
+    now_dt = datetime.strftime(now_date, '%Y%m%d')
+    if key_dt == now_dt:
+        initial_key_name_prefix = config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_NOW
+    else:
+        initial_key_name_prefix = config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_PRE
+    initial_key_name = f"{initial_key_name_prefix}{rule_key}.{now_dt}"
+    if len(initial_data_dup) > 0:
+        redis_helper.add_data_with_zset(key_name=initial_key_name, data=initial_data_dup, expire_time=7 * 24 * 3600)
+
+
+def rank_by_day(now_date, rule_params, project, table):
+    # 获取特征数据
+    feature_df = get_feature_data(now_date=now_date, project=project, table=table)
+    # rank
+    for key, value in rule_params.items():
+        log_.info(f"rule = {key}, param = {value}")
+        # 计算score
+        cal_score_func = value.get('cal_score_func', 1)
+        if cal_score_func == 2:
+            score_df = cal_score2(df=feature_df)
+        else:
+            score_df = cal_score1(df=feature_df)
+        video_rank_day(df=score_df, now_date=now_date, rule_key=key, param=value)
+        # to-csv
+        score_filename = f"score_{key}_{datetime.strftime(now_date, '%Y%m%d')}.csv"
+        score_df.to_csv(f'./data/{score_filename}')
+
+
+def day_rank_bottom(now_date, rule_key):
+    """未按时更新数据，用模型召回数据作为当前的数据"""
+    log_.info(f"rule_key = {rule_key}")
+    now_dt = datetime.strftime(now_date, '%Y%m%d')
+    redis_helper = RedisHelper()
+    key_name, key_dt = get_rov_redis_key(now_date=now_date)
+    initial_data = redis_helper.get_data_zset_with_index(key_name=key_name, start=0, end=-1, with_scores=True)
+    final_data = dict()
+    for video_id, score in initial_data:
+        final_data[video_id] = score
+    if key_dt == now_dt:
+        key_name_prefix = config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_NOW
+    else:
+        key_name_prefix = config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_PRE
+    key_name = f"{key_name_prefix}{rule_key}.{now_dt}"
+    if len(final_data) > 0:
+        redis_helper.add_data_with_zset(key_name=key_name, data=final_data, expire_time=7 * 24 * 3600)
+
+
+def day_timer_check():
+    project = config_.PROJECT_DAY
+    table = config_.TABLE_DAY
+    rule_params = config_.RULE_PARAMS_DAY
+    # return_count_list = [20, 10]
+    now_date = datetime.today()
+    log_.info(f"now_date: {datetime.strftime(now_date, '%Y%m%d')}")
+    now_h = datetime.now().hour
+    now_min = datetime.now().minute
+    # 查看当前天级更新的数据是否已准备好
+    h_data_count = day_data_check(project=project, table=table, now_date=now_date)
+    if h_data_count > 0:
+        log_.info(f'h_data_count = {h_data_count}')
+        # 数据准备好，进行更新
+        rank_by_day(now_date=now_date, rule_params=rule_params, project=project, table=table)
+    elif now_min > 50:
+        log_.info('day_recall data is None!')
+        for key, _ in rule_params.items():
+            day_rank_bottom(now_date=now_date, rule_key=key)
+    else:
+        # 数据没准备好，1分钟后重新检查
+        Timer(60, day_timer_check).start()
+
+
+if __name__ == '__main__':
+    day_timer_check()

rule_rank_day_task.sh

@@ -0,0 +1,7 @@
+source /etc/profile
+echo $ROV_OFFLINE_ENV
+if [[ $ROV_OFFLINE_ENV == 'test' ]]; then
+    cd /data2/rov-offline && /root/anaconda3/bin/python /data2/rov-offline/rule_rank_day.py
+elif [[ $ROV_OFFLINE_ENV == 'pro' ]]; then
+    cd /data/rov-offline && /root/anaconda3/bin/python /data/rov-offline/rule_rank_day.py
+fi

videos_filter.py

@@ -466,6 +466,48 @@ def filter_rov_h():
     log_.info("rov_h pool filter end!")
 
 
+def filter_rov_day():
+    """过滤小程序天级数据"""
+    rule_params = config_.RULE_PARAMS_DAY
+    log_.info("rov_day pool filter start ...")
+    redis_helper = RedisHelper()
+    # 获取当前日期
+    now_date = date.today().strftime('%Y%m%d')
+    log_.info(f'now_date = {now_date}.')
+    for key, value in rule_params.items():
+        log_.info(f"rule = {key}, param = {value}")
+        # 需过滤三个视频列表
+        key_prefix_list = [
+            config_.RECALL_KEY_NAME_PREFIX_BY_DAY,
+            config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_PRE,
+            config_.RECALL_KEY_NAME_PREFIX_DUP_DAY_NOW
+        ]
+        for i, key_prefix in enumerate(key_prefix_list):
+            # 拼接key
+            key_name = f"{key_prefix}{key}.{now_date}"
+            log_.info(f"key_name: {key_name}")
+            # 获取视频
+            data = redis_helper.get_data_zset_with_index(key_name=key_name, start=0, end=-1)
+            if data is None:
+                log_.info("data is None")
+                log_.info("filter end!")
+                continue
+            # 过滤
+            video_ids = [int(video_id) for video_id in data]
+            filtered_result = filter_video_status(video_ids=video_ids)
+            # 求差集，获取需要过滤掉的视频，并从redis中移除
+            filter_videos = set(video_ids) - set(filtered_result)
+            log_.info("video_ids size = {}, filtered size = {}, filter sizer = {}".format(len(video_ids),
+                                                                                          len(filtered_result),
+                                                                                          len(filter_videos)))
+            if len(filter_videos) == 0:
+                log_.info("filter end!")
+                continue
+            redis_helper.remove_value_from_zset(key_name=key_name, value=list(filter_videos))
+
+    log_.info("rov_day pool filter end!")
+
+
 def main():
     try:
         # ROV召回池视频过滤
@@ -493,6 +535,8 @@ def main():
         filter_app_pool()
         # 过滤小程序小时级数据
         filter_rov_h()
+        # 过滤小程序天级数据
+        filter_rov_day()
     except Exception as e:
         log_.error(traceback.format_exc())
         send_msg_to_feishu(