表数据
class UserInfo(AbstractUser):
"""
用户信息
"""
nid = models.BigAutoField(primary_key=True)
nickname = models.CharField(verbose_name='昵称', max_length=32)
telephone = models.CharField(max_length=11, blank=True, null=True, unique=True, verbose_name='手机号码')
avatar = models.FileField(verbose_name='头像',upload_to = 'avatar/',default="/avatar/default.png")
create_time = models.DateTimeField(verbose_name='创建时间', auto_now_add=True)
fans = models.ManyToManyField(verbose_name='粉丝们',
to='UserInfo',
through='UserFans',
related_name='f',
through_fields=('user', 'follower'))
def __str__(self):
return self.username
class UserFans(models.Model):
"""
互粉关系表
"""
nid = models.AutoField(primary_key=True)
user = models.ForeignKey(verbose_name='博主', to='UserInfo', to_field='nid', related_name='users')
follower = models.ForeignKey(verbose_name='粉丝', to='UserInfo', to_field='nid', related_name='followers')
class Blog(models.Model):
"""
博客信息
"""
nid = models.BigAutoField(primary_key=True)
title = models.CharField(verbose_name='个人博客标题', max_length=64)
site = models.CharField(verbose_name='个人博客后缀', max_length=32, unique=True)
theme = models.CharField(verbose_name='博客主题', max_length=32)
user = models.OneToOneField(to='UserInfo', to_field='nid')
def __str__(self):
return self.title
class Category(models.Model):
"""
博主个人文章分类表
"""
nid = models.AutoField(primary_key=True)
title = models.CharField(verbose_name='分类标题', max_length=32)
blog = models.ForeignKey(verbose_name='所属博客', to='Blog', to_field='nid')
class Article(models.Model):
nid = models.BigAutoField(primary_key=True)
title = models.CharField(max_length=50, verbose_name='文章标题')
desc = models.CharField(max_length=255, verbose_name='文章描述')
read_count = models.IntegerField(default=0)
comment_count= models.IntegerField(default=0)
up_count = models.IntegerField(default=0)
down_count = models.IntegerField(default=0)
category = models.ForeignKey(verbose_name='文章类型', to='Category', to_field='nid', null=True)
create_time = models.DateField(verbose_name='创建时间')
blog = models.ForeignKey(verbose_name='所属博客', to='Blog', to_field='nid')
tags = models.ManyToManyField(
to="Tag",
through='Article2Tag',
through_fields=('article', 'tag'),
)
class ArticleDetail(models.Model):
"""
文章详细表
"""
nid = models.AutoField(primary_key=True)
content = models.TextField(verbose_name='文章内容', )
article = models.OneToOneField(verbose_name='所属文章', to='Article', to_field='nid')
class Comment(models.Model):
"""
评论表
"""
nid = models.BigAutoField(primary_key=True)
article = models.ForeignKey(verbose_name='评论文章', to='Article', to_field='nid')
content = models.CharField(verbose_name='评论内容', max_length=255)
create_time = models.DateTimeField(verbose_name='创建时间', auto_now_add=True)
parent_comment = models.ForeignKey('self', blank=True, null=True, verbose_name='父级评论')
user = models.ForeignKey(verbose_name='评论者', to='UserInfo', to_field='nid')
up_count = models.IntegerField(default=0)
def __str__(self):
return self.content
class ArticleUpDown(models.Model):
"""
点赞表
"""
nid = models.AutoField(primary_key=True)
user = models.ForeignKey('UserInfo', null=True)
article = models.ForeignKey("Article", null=True)
models.BooleanField(verbose_name='是否赞')
class CommentUp(models.Model):
"""
点赞表
"""
nid = models.AutoField(primary_key=True)
user = models.ForeignKey('UserInfo', null=True)
comment = models.ForeignKey("Comment", null=True)
class Tag(models.Model):
nid = models.AutoField(primary_key=True)
title = models.CharField(verbose_name='标签名称', max_length=32)
blog = models.ForeignKey(verbose_name='所属博客', to='Blog', to_field='nid')
class Article2Tag(models.Model):
nid = models.AutoField(primary_key=True)
article = models.ForeignKey(verbose_name='文章', to="Article", to_field='nid')
tag = models.ForeignKey(verbose_name='标签', to="Tag", to_field='nid')
select_related
简单使用
对于一对一字段(OneToOneField)和外键字段(ForeignKey),可以使用select_related 来对QuerySet进行优化。
select_related 返回一个QuerySet,当执行它的查询时它沿着外键关系查询关联的对象的数据。它会生成一个复杂的查询并引起性能的损耗,但是在以后使用外键关系时将不需要数据库查询。
简单说,在对QuerySet使用select_related()函数后,Django会获取相应外键对应的对象,从而在之后需要的时候不必再查询数据库了。
下面的例子解释了普通查询和select_related() 查询的区别。
查询id=2的文章的分类名称,下面是一个标准的查询:
# Hits the database.
article=models.Article.objects.get(nid=2)
# Hits the database again to get the related Blog object.
print(article.category.title)
SELECT
"blog_article"."nid",
"blog_article"."title",
"blog_article"."desc",
"blog_article"."read_count",
"blog_article"."comment_count",
"blog_article"."up_count",
"blog_article"."down_count",
"blog_article"."category_id",
"blog_article"."create_time",
"blog_article"."blog_id",
"blog_article"."article_type_id"
FROM "blog_article"
WHERE "blog_article"."nid" = 2; args=(2,)
SELECT
"blog_category"."nid",
"blog_category"."title",
"blog_category"."blog_id"
FROM "blog_category"
WHERE "blog_category"."nid" = 4; args=(4,)
如果我们使用select_related()函数:
articleList=models.Article.objects.select_related("category").all()
for article_obj in articleList:
# Doesn't hit the database, because article_obj.category
# has been prepopulated in the previous query.
print(article_obj.category.title)
SELECT
"blog_article"."nid",
"blog_article"."title",
"blog_article"."desc",
"blog_article"."read_count",
"blog_article"."comment_count",
"blog_article"."up_count",
"blog_article"."down_count",
"blog_article"."category_id",
"blog_article"."create_time",
"blog_article"."blog_id",
"blog_article"."article_type_id",
"blog_category"."nid",
"blog_category"."title",
"blog_category"."blog_id"
FROM "blog_article"
LEFT OUTER JOIN "blog_category" ON ("blog_article"."category_id" = "blog_category"."nid");
多外键查询
这是针对category的外键查询,如果是另外一个外键呢?让我们一起看下:
article=models.Article.objects.select_related("category").get(nid=1)
print(article.articledetail)
观察logging结果,发现依然需要查询两次,所以需要改为:
article=models.Article.objects.select_related("category","articledetail").get(nid=1)
print(article.articledetail)
或者:
article=models.Article.objects
.select_related("category")
.select_related("articledetail")
.get(nid=1) # django 1.7 支持链式操作
print(article.articledetail)
SELECT
"blog_article"."nid",
"blog_article"."title",
......
"blog_category"."nid",
"blog_category"."title",
"blog_category"."blog_id",
"blog_articledetail"."nid",
"blog_articledetail"."content",
"blog_articledetail"."article_id"
FROM "blog_article"
LEFT OUTER JOIN "blog_category" ON ("blog_article"."category_id" = "blog_category"."nid")
LEFT OUTER JOIN "blog_articledetail" ON ("blog_article"."nid" = "blog_articledetail"."article_id")
WHERE "blog_article"."nid" = 1; args=(1,)
深层查询
# 查询id=1的文章的用户姓名
article=models.Article.objects.select_related("blog").get(nid=1)
print(article.blog.user.username)
依然需要查询两次:
SELECT
"blog_article"."nid",
"blog_article"."title",
......
"blog_blog"."nid",
"blog_blog"."title",
FROM "blog_article" INNER JOIN "blog_blog" ON ("blog_article"."blog_id" = "blog_blog"."nid")
WHERE "blog_article"."nid" = 1;
SELECT
"blog_userinfo"."password",
"blog_userinfo"."last_login",
......
FROM "blog_userinfo"
WHERE "blog_userinfo"."nid" = 1;
这是因为第一次查询没有query到userInfo表,所以,修改如下:
article=models.Article.objects.select_related("blog__user").get(nid=1)
print(article.blog.user.username)
SELECT
"blog_article"."nid", "blog_article"."title",
......
"blog_blog"."nid", "blog_blog"."title",
......
"blog_userinfo"."password", "blog_userinfo"."last_login",
......
FROM "blog_article"
INNER JOIN "blog_blog" ON ("blog_article"."blog_id" = "blog_blog"."nid")
INNER JOIN "blog_userinfo" ON ("blog_blog"."user_id" = "blog_userinfo"."nid")
WHERE "blog_article"."nid" = 1;
总结
- select_related主要针一对一和多对一关系进行优化。
- 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。
3.可以通过可变长参数指定需要select_related的字段名。也可以通过使用双下划线“__”连接字段名来实现指定的递归查询。 - 没有指定的字段不会缓存,没有指定的深度不会缓存,如果要访问的话Django会再次进行SQL查询。
也可以通过depth参数指定递归的深度,Django会自动缓存指定深度内所有的字段。如果要访问指定深度外的字段,Django会再次进行SQL查询。 - 也接受无参数的调用,Django会尽可能深的递归查询所有的字段。但注意有Django递归的限制和性能的浪费。
- Django >= 1.7,链式调用的select_related相当于使用可变长参数。Django < 1.7,链式调用会导致前边的select_related失效,只保留最后一个。
prefetch_related()
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。
prefetch_related()和select_related()的设计目的很相似,都是为了减少SQL查询的数量,但是实现的方式不一样。后者是通过JOIN语句,在SQL查询内解决问题。但是对于多对多关系,使用SQL语句解决就显得有些不太明智,因为JOIN得到的表将会很长,会导致SQL语句运行时间的增加和内存占用的增加。若有n个对象,每个对象的多对多字段对应Mi条,就会生成Σ(n)Mi 行的结果表。
prefetch_related()的解决方法是,分别查询每个表,然后用Python处理他们之间的关系。
# 查询所有文章关联的所有标签
article_obj=models.Article.objects.all()
for i in article_obj:
print(i.tags.all()) #4篇文章: hits database 5
改为prefetch_related:
# 查询所有文章关联的所有标签
article_obj=models.Article.objects.prefetch_related("tags").all()
for i in article_obj:
print(i.tags.all()) #4篇文章: hits database 2
SELECT "blog_article"."nid",
"blog_article"."title",
......
FROM "blog_article";
SELECT
("blog_article2tag"."article_id") AS "_prefetch_related_val_article_id",
"blog_tag"."nid",
"blog_tag"."title",
"blog_tag"."blog_id"
FROM "blog_tag"
INNER JOIN "blog_article2tag" ON ("blog_tag"."nid" = "blog_article2tag"."tag_id")
WHERE "blog_article2tag"."article_id" IN (1, 2, 3, 4);