import sys
import pickle
import numpy as np
import pandas as pd

sys.path.insert(0, '/Users/puhaoran/Desktop/code/algpro/rcpro/eva')

def get_debug_sample(sample_nums=10000):
    """ 训练集中采样 """
    train_df = pd.read_csv('./inputs/train_click_log.csv')
    train_user_ids = train_df.user_id.unique()
    sample_user_ids = np.random.choice(train_user_ids, size=sample_nums, replace=False)
    all_click_df = train_df[train_df.user_id.isin(sample_user_ids)]
    all_click_df = all_click_df.drop_duplicates((['user_id', 'click_article_id', 'click_timestamp']))
    return all_click_df

def get_train_df(strategy='local'):
    """ 线上策略或线下策略 """ 
    if strategy == 'local':
        all_click_df = pd.read_csv('./inputs/train_click_log.csv')
    else:
        train_df = pd.read_csv('./inputs/train_click_log.csv')
        test_df = pd.read_csv('./inputs/testA_click_log.csv')
        all_click_df = train_df.append(test_df)
    all_click_df = all_click_df.drop_duplicates((['user_id', 'click_article_id', 'click_timestamp']))
    return all_click_df

def get_item_info():
    """ 读取物品数据 """
    item_df = pd.read_csv('./inputs/articles.csv')
    item_df = item_df.rename(columns={'article_id': 'click_article_id'})
    return item_df

def get_emb_dict():
    """ 获取embedding数据 """
    item_emb_df = pd.read_csv('./inputs/articles_emb.csv')
    cols = [x for x in item_emb_df.columns if 'emb' in x]
    t = np.ascontiguousarray(item_emb_df[cols])
    # 归一化
    t = t / np.linalg.norm(t, axis=1, keepdims=True)
    item_emb_dict = dict(zip(item_emb_df['article_id'], t))
    pickle.dump(item_emb_dict, open('./inputs/item_emb.pkl', 'wb'))
    return item_emb_dict

data = get_debug_sample()
min_max_scaler = lambda x: (x-np.min(x)) / (np.max(x)- np.min(x))
# 对时间戳归一化，用于关联规则时计算权重
data['click_timestamp'] = data[['click_timestamp']].apply(min_max_scaler)
item_df = get_item_info()
get_emb_dict = get_emb_dict()

##### 定义多路召回字典
multiple_recall_dict = {
    'cold_boot': {},
    'itemcf_recall': {},
    'emb_recall': {},
    'youtubednn_recall': {},
}

阿斯蒂芬