Django ORM执行原生SQL
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
|
# extra # 在QuerySet的基础上继续执行子语句 # extra(self, select=None, where=None, params=None, tables=None, order_by=None, select_params=None) # select和select_params是一组,where和params是一组,tables用来设置from哪个表 # Entry.objects.extra(select={'new_id': "select col from sometable where othercol > %s"}, select_params=(1,)) # Entry.objects.extra(where=['headline=%s'], params=['Lennon']) # Entry.objects.extra(where=["foo='a' OR bar = 'a'", "baz = 'a'"]) # Entry.objects.extra(select={'new_id': "select id from tb where id > %s"}, select_params=(1,), order_by=['-nid']) 举个例子: models.UserInfo.objects.extra( select = { 'newid' : 'select count(1) from app01_usertype where id>%s' }, select_params = [ 1 ,], where = [ 'age>%s' ], params = [ 18 ,], order_by = [ '-age' ], tables = [ 'app01_usertype' ] ) """ select app01_userinfo.id, (select count(1) from app01_usertype where id>1) as newid from app01_userinfo,app01_usertype where app01_userinfo.age > 18 order by app01_userinfo.age desc """ # 执行原生SQL # 更高灵活度的方式执行原生SQL语句 # from django.db import connection, connections # cursor = connection.cursor() # cursor = connections['default'].cursor() # cursor.execute("""SELECT * from auth_user where id = %s""", [1]) # row = cursor.fetchone() |
QuerySet方法大全
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
|
################################################################## # PUBLIC METHODS THAT ALTER ATTRIBUTES AND RETURN A NEW QUERYSET # ################################################################## def all ( self ) # 获取所有的数据对象 def filter ( self , * args, * * kwargs) # 条件查询 # 条件可以是:参数,字典,Q def exclude( self , * args, * * kwargs) # 条件查询 # 条件可以是:参数,字典,Q def select_related( self , * fields) 性能相关:表之间进行join连表操作,一次性获取关联的数据。 总结: 1. select_related主要针一对一和多对一关系进行优化。 2. select_related使用SQL的JOIN语句进行优化,通过减少SQL查询的次数来进行优化、提高性能。 def prefetch_related( self , * lookups) 性能相关:多表连表操作时速度会慢,使用其执行多次SQL查询在Python代码中实现连表操作。 总结: 1. 对于多对多字段(ManyToManyField)和一对多字段,可以使用prefetch_related()来进行优化。 2. prefetch_related()的优化方式是分别查询每个表,然后用Python处理他们之间的关系。 def annotate( self , * args, * * kwargs) # 用于实现聚合group by查询 from django.db.models import Count, Avg, Max , Min , Sum v = models.UserInfo.objects.values( 'u_id' ).annotate(uid = Count( 'u_id' )) # SELECT u_id, COUNT(ui) AS `uid` FROM UserInfo GROUP BY u_id v = models.UserInfo.objects.values( 'u_id' ).annotate(uid = Count( 'u_id' )). filter (uid__gt = 1 ) # SELECT u_id, COUNT(ui_id) AS `uid` FROM UserInfo GROUP BY u_id having count(u_id) > 1 v = models.UserInfo.objects.values( 'u_id' ).annotate(uid = Count( 'u_id' ,distinct = True )). filter (uid__gt = 1 ) # SELECT u_id, COUNT( DISTINCT ui_id) AS `uid` FROM UserInfo GROUP BY u_id having count(u_id) > 1 def distinct( self , * field_names) # 用于distinct去重 models.UserInfo.objects.values( 'nid' ).distinct() # select distinct nid from userinfo 注:只有在PostgreSQL中才能使用distinct进行去重 def order_by( self , * field_names) # 用于排序 models.UserInfo.objects. all ().order_by( '-id' , 'age' ) def extra( self , select = None , where = None , params = None , tables = None , order_by = None , select_params = None ) # 构造额外的查询条件或者映射,如:子查询 Entry.objects.extra(select = { 'new_id' : "select col from sometable where othercol > %s" }, select_params = ( 1 ,)) Entry.objects.extra(where = [ 'headline=%s' ], params = [ 'Lennon' ]) Entry.objects.extra(where = [ "foo='a' OR bar = 'a'" , "baz = 'a'" ]) Entry.objects.extra(select = { 'new_id' : "select id from tb where id > %s" }, select_params = ( 1 ,), order_by = [ '-nid' ]) def reverse( self ): # 倒序 models.UserInfo.objects. all ().order_by( '-nid' ).reverse() # 注:如果存在order_by,reverse则是倒序,如果多个排序则一一倒序 def defer( self , * fields): models.UserInfo.objects.defer( 'username' , 'id' ) 或 models.UserInfo.objects. filter (...).defer( 'username' , 'id' ) #映射中排除某列数据 def only( self , * fields): #仅取某个表中的数据 models.UserInfo.objects.only( 'username' , 'id' ) 或 models.UserInfo.objects. filter (...).only( 'username' , 'id' ) def using( self , alias): 指定使用的数据库,参数为别名(setting中的设置) ################################################## # PUBLIC METHODS THAT RETURN A QUERYSET SUBCLASS # ################################################## def raw( self , raw_query, params = None , translations = None , using = None ): # 执行原生SQL models.UserInfo.objects.raw( 'select * from userinfo' ) # 如果SQL是其他表时,必须将名字设置为当前UserInfo对象的主键列名 models.UserInfo.objects.raw( 'select id as nid from 其他表' ) # 为原生SQL设置参数 models.UserInfo.objects.raw( 'select id as nid from userinfo where nid>%s' , params = [ 12 ,]) # 将获取的到列名转换为指定列名 name_map = { 'first' : 'first_name' , 'last' : 'last_name' , 'bd' : 'birth_date' , 'pk' : 'id' } Person.objects.raw( 'SELECT * FROM some_other_table' , translations = name_map) # 指定数据库 models.UserInfo.objects.raw( 'select * from userinfo' , using = "default" ) ################### 原生SQL ################### from django.db import connection, connections cursor = connection.cursor() # cursor = connections['default'].cursor() cursor.execute( """SELECT * from auth_user where id = %s""" , [ 1 ]) row = cursor.fetchone() # fetchall()/fetchmany(..) def values( self , * fields): # 获取每行数据为字典格式 def values_list( self , * fields, * * kwargs): # 获取每行数据为元祖 def dates( self , field_name, kind, order = 'ASC' ): # 根据时间进行某一部分进行去重查找并截取指定内容 # kind只能是:"year"(年), "month"(年-月), "day"(年-月-日) # order只能是:"ASC" "DESC" # 并获取转换后的时间 - year : 年 - 01 - 01 - month: 年 - 月 - 01 - day : 年 - 月 - 日 models.DatePlus.objects.dates( 'ctime' , 'day' , 'DESC' ) def datetimes( self , field_name, kind, order = 'ASC' , tzinfo = None ): # 根据时间进行某一部分进行去重查找并截取指定内容,将时间转换为指定时区时间 # kind只能是 "year", "month", "day", "hour", "minute", "second" # order只能是:"ASC" "DESC" # tzinfo时区对象 models.DDD.objects.datetimes( 'ctime' , 'hour' ,tzinfo = pytz.UTC) models.DDD.objects.datetimes( 'ctime' , 'hour' ,tzinfo = pytz.timezone( 'Asia/Shanghai' )) """ pip3 install pytz import pytz pytz.all_timezones pytz.timezone(‘Asia/Shanghai’) """ def none( self ): # 空QuerySet对象 #################################### # METHODS THAT DO DATABASE QUERIES # #################################### def aggregate( self , * args, * * kwargs): # 聚合函数,获取字典类型聚合结果 from django.db.models import Count, Avg, Max , Min , Sum result = models.UserInfo.objects.aggregate(k = Count( 'u_id' , distinct = True ), n = Count( 'nid' )) = = = > { 'k' : 3 , 'n' : 4 } def count( self ): # 获取个数 def get( self , * args, * * kwargs): # 获取单个对象 def create( self , * * kwargs): # 创建对象 def bulk_create( self , objs, batch_size = None ): # 批量插入 # batch_size表示一次插入的个数 objs = [ models.DDD(name = 'r11' ), models.DDD(name = 'r22' ) ] models.DDD.objects.bulk_create(objs, 10 ) def get_or_create( self , defaults = None , * * kwargs): # 如果存在,则获取,否则,创建 # defaults 指定创建时,其他字段的值 obj, created = models.UserInfo.objects.get_or_create(username = 'root1' , defaults = { 'email' : '1111111' , 'u_id' : 2 , 't_id' : 2 }) def update_or_create( self , defaults = None , * * kwargs): # 如果存在,则更新,否则,创建 # defaults 指定创建时或更新时的其他字段 obj, created = models.UserInfo.objects.update_or_create(username = 'root1' , defaults = { 'email' : '1111111' , 'u_id' : 2 , 't_id' : 1 }) def first( self ): # 获取第一个 def last( self ): # 获取最后一个 def in_bulk( self , id_list = None ): # 根据主键ID进行查找 id_list = [ 11 , 21 , 31 ] models.DDD.objects.in_bulk(id_list) def delete( self ): # 删除 def update( self , * * kwargs): # 更新 def exists( self ): # 是否有结果 |
Django终端打印SQL语句
在Django项目的settings.py文件中,在最后复制粘贴如下代码:即为你的Django项目配置上一个名为django.db.backends的logger实例即可查看翻译后的SQL语句。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
|
LOGGING = { 'version' : 1 , 'disable_existing_loggers' : False , 'handlers' : { 'console' :{ 'level' : 'DEBUG' , 'class' : 'logging.StreamHandler' , }, }, 'loggers' : { 'django.db.backends' : { 'handlers' : [ 'console' ], 'propagate' : True , 'level' : 'DEBUG' , }, } } |
在Python脚本中调用Django环境
1
2
3
4
5
6
7
8
9
10
11
|
import os if __name__ = = '__main__' : os.environ.setdefault( "DJANGO_SETTINGS_MODULE" , "BMS.settings" ) import django django.setup() from app01 import models books = models.Book.objects. all () print (books) |