1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
import os
import sys
import pandas as pd
os.environ["PYARROW_IGNORE_TIMEZONE"] = "1"
sys.path.append(os.path.dirname(sys.path[0])) # 上级目录
from utils.templates import Templates
# from ..utils.templates import Templates
# 分组排序的udf窗口函数
from pyspark.sql.window import Window
from pyspark.sql import functions as F
class DimAsinFeaturesParquet(Templates):
def __init__(self, site_name='us', block_size=100000):
super(DimAsinFeaturesParquet, self).__init__()
self.site_name = site_name
self.block_size = block_size
self.db_save = f'dim_asin_features_parquet'
self.spark = self.create_spark_object(
app_name=f"{self.db_save}: {self.site_name}, {self.block_size}")
self.df_asin_features = self.spark.sql(f"select 1+1;")
self.df_save = self.spark.sql(f"select 1+1;")
# self.partitions_by = ['site_name', 'block']
self.partitions_by = ['site_name']
self.partitions_num = 600
def read_data(self):
sql = f"select id, asin, img_vector as embedding from ods_asin_extract_features;"
print("sql:", sql)
self.df_save = self.spark.sql(sql).cache()
def handle_data(self):
# self.df_save = self.df_save.withColumn('block', F.floor(self.df_save['id'] / self.block_size))
self.df_save = self.df_save.withColumn('site_name', F.lit(self.site_name))
def handle_data_old(self):
"""
开窗这种方式进行全局索引，会导致所有的数据在1个分区里面，从而只能有1个cpu运行，降低了性能
"""
# 添加索引列 -- index
window = Window.orderBy("id")
self.df_asin_features = self.df_asin_features.withColumn("index", F.row_number().over(window) - 1) # 从0开始
self.df_asin_features.show(20)
# 生成分区列
self.df_asin_features = self.df_asin_features.withColumn('block', F.floor(self.df_asin_features['index'] / self.block_size))
self.df_asin_features = self.df_asin_features.withColumn('site_name', F.lit(self.site_name))
self.df_save = self.df_asin_features
def handle_data_old2(self):
print("分块前分区数量:", self.df_asin_features.rdd.getNumPartitions()) # 642
from pyspark.sql.functions import spark_partition_id
num_partitions = 500 # 你需要根据你的数据和资源来调整这个参数 #
# 第一步: 对数据进行预分区和排序
self.df_asin_features = self.df_asin_features.repartitionByRange(num_partitions, "id").sortWithinPartitions("id")
print("分块后分区数量:", self.df_asin_features.rdd.getNumPartitions())
# 第二步: 在每个分区内部添加索引
# def add_index_in_partition(df):
# # 使用窗口函数在每个分区内部添加索引
# window = Window.orderBy("id")
# df = df.withColumn("index", F.row_number().over(window) - 1) # 从0开始
# return df
# from pyspark.sql.functions import pandas_udf, PandasUDFType
# from pyspark.sql import DataFrame
#
# @pandas_udf("id long, embedding string, index long, block long, site_name string", PandasUDFType.GROUPED_MAP)
# def add_index_in_partition(pdf: pd.DataFrame) -> pd.DataFrame:
# # 使用pandas的cumcount函数在每个分区内部添加索引
# pdf['index'] = pdf.sort_values('id').groupby().cumcount()
# return pdf
from pyspark.sql.functions import pandas_udf, PandasUDFType
@pandas_udf("id long, embedding string, index long", PandasUDFType.GROUPED_MAP)
def add_index_in_partition(df):
df = df.sort_values('id')
df['index'] = range(len(df)) # 或者 df['index'] = df.reset_index().index
return df
self.df_asin_features = self.df_asin_features.groupby(spark_partition_id()).apply(add_index_in_partition)
# 添加全局索引
self.df_asin_features = self.df_asin_features.withColumn("index", F.sum("index").over(Window.orderBy("id")))
# 生成分区列
self.df_asin_features = self.df_asin_features.withColumn('block', F.floor(self.df_asin_features['index'] / self.block_size))
self.df_asin_features = self.df_asin_features.withColumn('site_name', F.lit(self.site_name))
# 存储
self.df_save = self.df_asin_features
# self.df_save.show(10)
if __name__ == '__main__':
handle_obj = DimAsinFeaturesParquet(block_size=200000)
handle_obj.run()
# import os
# import sys
# import pandas as pd
#
# os.environ["PYARROW_IGNORE_TIMEZONE"] = "1"
# sys.path.append(os.path.dirname(sys.path[0])) # 上级目录
# from utils.templates import Templates
# # from ..utils.templates import Templates
# # 分组排序的udf窗口函数
# from pyspark.sql.window import Window
# from pyspark.sql import functions as F
#
#
# class DimAsinFeaturesParquet(Templates):
#
# def __init__(self, site_name='us', block_size=100000):
# super(DimAsinFeaturesParquet, self).__init__()
# self.site_name = site_name
# self.block_size = block_size
# self.db_save = f'dim_asin_features_parquet'
# self.spark = self.create_spark_object(
# app_name=f"{self.db_save}: {self.site_name}, {self.block_size}")
# self.df_asin_features = self.spark.sql(f"select 1+1;")
# self.df_save = self.spark.sql(f"select 1+1;")
# self.partitions_by = ['site_name', 'block']
# self.partitions_num = 50
#
# def read_data(self):
# sql = f"select id, img_vector as embedding from ods_asin_extract_features;"
# print("sql:", sql)
# self.df_asin_features = self.spark.sql(sql).cache()
#
# def handle_data_old(self):
# """
# 开窗这种方式进行全局索引，会导致所有的数据在1个分区里面，从而只能有1个cpu运行，降低了性能
# """
# # 添加索引列 -- index
# window = Window.orderBy("id")
# self.df_asin_features = self.df_asin_features.withColumn("index", F.row_number().over(window) - 1) # 从0开始
# self.df_asin_features.show(20)
# # 生成分区列
# self.df_asin_features = self.df_asin_features.withColumn('block', F.floor(self.df_asin_features['index'] / self.block_size))
# self.df_asin_features = self.df_asin_features.withColumn('site_name', F.lit(self.site_name))
#
# self.df_save = self.df_asin_features
#
# def handle_data(self):
# from pyspark.sql.functions import spark_partition_id
# print("分块前分区数量:", self.df_asin_features.rdd.getNumPartitions())
# num_partitions = 500 # 你需要根据你的数据和资源来调整这个参数
#
# # 第一步: 对数据进行预分区和排序
# self.df_asin_features = self.df_asin_features.repartitionByRange(num_partitions, "id").sortWithinPartitions("id")
#
# # 第二步: 在每个分区内部添加索引
# # def add_index_in_partition(df):
# # # 使用窗口函数在每个分区内部添加索引
# # window = Window.orderBy("id")
# # df = df.withColumn("index", F.row_number().over(window) - 1) # 从0开始
# # return df
#
# # from pyspark.sql.functions import pandas_udf, PandasUDFType
# # from pyspark.sql import DataFrame
# #
# # @pandas_udf("id long, embedding string, index long, block long, site_name string", PandasUDFType.GROUPED_MAP)
# # def add_index_in_partition(pdf: pd.DataFrame) -> pd.DataFrame:
# # # 使用pandas的cumcount函数在每个分区内部添加索引
# # pdf['index'] = pdf.sort_values('id').groupby().cumcount()
# # return pdf
#
# from pyspark.sql.functions import pandas_udf, PandasUDFType
#
# @pandas_udf("id long, embedding string, index long", PandasUDFType.GROUPED_MAP)
# def add_index_in_partition(df):
# df = df.sort_values('id')
# df['index'] = range(len(df)) # 或者 df['index'] = df.reset_index().index
# return df
#
# # print("分块前分区数量:", self.df_asin_features.rdd.getNumPartitions())
# # self.df_asin_features = self.df_asin_features.repartition(1000)
# #
# # print("分块后分区数量:", self.df_asin_features.rdd.getNumPartitions())
#
# self.df_asin_features = self.df_asin_features.groupby(spark_partition_id()).apply(add_index_in_partition)
#
# # 添加全局索引
# self.df_asin_features = self.df_asin_features.withColumn("index", F.sum("index").over(Window.orderBy("id")))
#
# # 生成分区列
# self.df_asin_features = self.df_asin_features.withColumn('block', F.floor(self.df_asin_features['index'] / self.block_size))
# self.df_asin_features = self.df_asin_features.withColumn('site_name', F.lit(self.site_name))
#
# # 存储
# self.df_save = self.df_asin_features
# # self.df_save.show(10)
#
#
# if __name__ == '__main__':
# handle_obj = DimAsinFeaturesParquet(block_size=500000)
# handle_obj.run()