Get best of the week

Django ORM para principiantes | Optimizamos solicitudes



Django ORM (Object Relational Mapping) es una de las características más poderosas de Django. Esto nos permite interactuar con la base de datos usando código Python, no SQL.

Para demostrarlo, describiré el siguiente modelo:

from django.db import models

class Blog(models.Model):
    name = models.CharField(max_length=250)
    url = models.URLField()

    def __str__(self):
        return self.name

class Author(models.Model):
    name = models.CharField(max_length=250)

    def __str__(self):
        return self.name

class Post(models.Model):
    title = models.CharField(max_length=250)
    content = models.TextField()
    published = models.BooleanField(default=True)
    blog = models.ForeignKey(Blog, on_delete=models.CASCADE)
    authors = models.ManyToManyField(Author, related_name="posts")

Voy a utilizar Django extensiones para obtener información útil con:

python manage.py shell_plus --print-sql

Vamos a empezar:

>>> post = Post.objects.all()
>>> post
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
 LIMIT 21
Execution time: 0.000172s [Database: default]
<QuerySet [<Post: Post object (1)>]>

1. Use los valores ForeignKey directamente


>>> Post.objects.first().blog.id
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
 ORDER BY "blog_post"."id" ASC
 LIMIT 1
Execution time: 0.000225s [Database: default]
SELECT "blog_blog"."id",
       "blog_blog"."name",
       "blog_blog"."url"
  FROM "blog_blog"
 WHERE "blog_blog"."id" = 1
 LIMIT 21
Execution time: 0.000144s [Database: default]
1

Y así obtenemos 1 consulta en la base de datos:

>>> Post.objects.first().blog_id
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
 ORDER BY "blog_post"."id" ASC
 LIMIT 1
Execution time: 0.000155s [Database: default]
1

2. Relaciones OneToMany


Si usamos la relación OneToMany usamos el campo ForeignKey y la consulta se ve así:

>>> post = Post.objects.get(id=1)
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
 WHERE "blog_post"."id" = 1
 LIMIT 21
Execution time: 0.000161s [Database: default]

Y si queremos acceder al objeto del blog desde el objeto de publicación, podemos hacer:

>>> post.blog
SELECT "blog_blog"."id",
       "blog_blog"."name",
       "blog_blog"."url"
  FROM "blog_blog"
 WHERE "blog_blog"."id" = 1
 LIMIT 21
Execution time: 0.000211s [Database: default]
<Blog: Django tutorials>

Sin embargo, esto desencadenó una nueva solicitud para obtener información del blog. Entonces use select_related para evitar esto. Para usarlo, podemos actualizar nuestra solicitud original:

>>> post = Post.objects.select_related("blog").get(id=1)
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id",
       "blog_blog"."id",
       "blog_blog"."name",
       "blog_blog"."url"
  FROM "blog_post"
 INNER JOIN "blog_blog"
    ON ("blog_post"."blog_id" = "blog_blog"."id")
 WHERE "blog_post"."id" = 1
 LIMIT 21
Execution time: 0.000159s [Database: default]

Tenga en cuenta que Django está utilizando JOIN ahora! Y el tiempo de ejecución de la consulta es más corto que antes. Además, ahora post.blog se almacenará en caché.

>>> post.blog
<Blog: Django tutorials>

select_related también funciona con QurySets:

>>> posts = Post.objects.select_related("blog").all()
>>> for post in posts:
...     post.blog
...
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id",
       "blog_blog"."id",
       "blog_blog"."name",
       "blog_blog"."url"
  FROM "blog_post"
 INNER JOIN "blog_blog"
    ON ("blog_post"."blog_id" = "blog_blog"."id")
Execution time: 0.000241s [Database: default]
<Blog: Django tutorials>

3. Relaciones ManyToMany


Para obtener autores de publicaciones, usamos algo como esto:

>>> for post in Post.objects.all():
...     post.authors.all()
...
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
Execution time: 0.000242s [Database: default]
SELECT "blog_author"."id",
       "blog_author"."name"
  FROM "blog_author"
 INNER JOIN "blog_post_authors"
    ON ("blog_author"."id" = "blog_post_authors"."author_id")
 WHERE "blog_post_authors"."post_id" = 1
 LIMIT 21
Execution time: 0.000125s [Database: default]
<QuerySet [<Author: Dmytro Parfeniuk>, <Author: Will Vincent>, <Author: Guido van Rossum>]>
SELECT "blog_author"."id",
       "blog_author"."name"
  FROM "blog_author"
 INNER JOIN "blog_post_authors"
    ON ("blog_author"."id" = "blog_post_authors"."author_id")
 WHERE "blog_post_authors"."post_id" = 2
 LIMIT 21
Execution time: 0.000109s [Database: default]
<QuerySet [<Author: Dmytro Parfeniuk>, <Author: Will Vincent>]>

Parece que tenemos una solicitud para cada objeto de publicación. Por lo tanto, debemos usar prefetch_related . Esto es similar a select_related pero se usa con ManyToMany Fields:

>>> for post in Post.objects.prefetch_related("authors").all():
...     post.authors.all()
...
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."blog_id"
  FROM "blog_post"
Execution time: 0.000300s [Database: default]
SELECT ("blog_post_authors"."post_id") AS "_prefetch_related_val_post_id",
       "blog_author"."id",
       "blog_author"."name"
  FROM "blog_author"
 INNER JOIN "blog_post_authors"
    ON ("blog_author"."id" = "blog_post_authors"."author_id")
 WHERE "blog_post_authors"."post_id" IN (1, 2)
Execution time: 0.000379s [Database: default]
<QuerySet [<Author: Dmytro Parfeniuk>, <Author: Will Vincent>, <Author: Guido van Rossum>]>
<QuerySet [<Author: Dmytro Parfeniuk>, <Author: Will Vincent>]>

Lo que acaba de suceder ??? Redujimos el número de consultas de 2 a 1 para obtener 2 QuerySet-a!

4. Objeto de captación previa


prefetch_related es suficiente para la mayoría de los casos, pero no siempre ayuda a evitar solicitudes adicionales. Por ejemplo, si usamos filtros, Django no puede usar nuestras publicaciones en caché , ya que no se filtraron cuando se solicitaron en la primera solicitud. Y obtenemos:

>>> authors = Author.objects.prefetch_related("posts").all()
>>> for author in authors:
...     print(author.posts.filter(published=True))
...
SELECT "blog_author"."id",
       "blog_author"."name"
  FROM "blog_author"
Execution time: 0.000580s [Database: default]
SELECT ("blog_post_authors"."author_id") AS "_prefetch_related_val_author_id",
       "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."published",
       "blog_post"."blog_id"
  FROM "blog_post"
 INNER JOIN "blog_post_authors"
    ON ("blog_post"."id" = "blog_post_authors"."post_id")
 WHERE "blog_post_authors"."author_id" IN (1, 2, 3)
Execution time: 0.000759s [Database: default]
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."published",
       "blog_post"."blog_id"
  FROM "blog_post"
 INNER JOIN "blog_post_authors"
    ON ("blog_post"."id" = "blog_post_authors"."post_id")
 WHERE ("blog_post_authors"."author_id" = 1 AND "blog_post"."published" = 1)
 LIMIT 21
Execution time: 0.000299s [Database: default]
<QuerySet [<Post: Post object (1)>, <Post: Post object (2)>]>
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."published",
       "blog_post"."blog_id"
  FROM "blog_post"
 INNER JOIN "blog_post_authors"
    ON ("blog_post"."id" = "blog_post_authors"."post_id")
 WHERE ("blog_post_authors"."author_id" = 2 AND "blog_post"."published" = 1)
 LIMIT 21
Execution time: 0.000336s [Database: default]
<QuerySet [<Post: Post object (1)>, <Post: Post object (2)>]>
SELECT "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."published",
       "blog_post"."blog_id"
  FROM "blog_post"
 INNER JOIN "blog_post_authors"
    ON ("blog_post"."id" = "blog_post_authors"."post_id")
 WHERE ("blog_post_authors"."author_id" = 3 AND "blog_post"."published" = 1)
 LIMIT 21
Execution time: 0.000412s [Database: default]
<QuerySet [<Post: Post object (1)>]>

Es decir, utilizamos prefetch_related para reducir el número de solicitudes, pero en realidad lo aumentamos. Para evitar esto, podemos personalizar la solicitud utilizando el objeto Prefetch :

>>> authors = Author.objects.prefetch_related(
...     Prefetch(
...             "posts",
...             queryset=Post.objects.filter(published=True),
...             to_attr="published_posts",
...     )
... )
>>> for author in authors:
...     print(author.published_posts)
...
SELECT "blog_author"."id",
       "blog_author"."name"
  FROM "blog_author"
Execution time: 0.000183s [Database: default]
SELECT ("blog_post_authors"."author_id") AS "_prefetch_related_val_author_id",
       "blog_post"."id",
       "blog_post"."title",
       "blog_post"."content",
       "blog_post"."published",
       "blog_post"."blog_id"
  FROM "blog_post"
 INNER JOIN "blog_post_authors"
    ON ("blog_post"."id" = "blog_post_authors"."post_id")
 WHERE ("blog_post"."published" = 1 AND "blog_post_authors"."author_id" IN (1, 2, 3))
Execution time: 0.000404s [Database: default]
[<Post: Post object (1)>, <Post: Post object (2)>]
[<Post: Post object (1)>, <Post: Post object (2)>]
[<Post: Post object (1)>]

Utilizamos una solicitud específica para recibir publicaciones a través del parámetro de solicitud y guardamos los mensajes filtrados en un nuevo atributo. Como podemos ver, ahora solo tenemos 2 consultas a la base de datos.

More articles:


All Articles