[英]Updating fields via subquery in Django
I have an app with models and db schema that looks as shown below.我有一个带有模型和数据库架构的应用程序,如下所示。 I am trying to add field
r
to L2 in order to be able to access the related objects from model R. The new field is not shown in the schema figure.我正在尝试将字段
r
添加到 L2,以便能够从模型 R 访问相关对象。新字段未显示在架构图中。
Retrieving the desired value of field r
using a subquery and an annotation works as expected.使用子查询和注释检索字段
r
的所需值按预期工作。 However, populating/updating the field with an update()
call does not work.但是,使用
update()
调用填充/更新字段不起作用。 Do I have to modify my subquery?我必须修改我的子查询吗? Or is this not possible at all in Django without resorting to raw SQL?
或者这在 Django 中根本不可能而不诉诸原始 SQL?
from django.db import models
class L1(models.Model):
desc = models.CharField(max_length=16)
m1 = models.ForeignKey('M1', on_delete=models.CASCADE)
class L2(models.Model):
desc = models.CharField(max_length=16)
l1 = models.ForeignKey('L1', on_delete=models.CASCADE)
m2 = models.ForeignKey('M2', on_delete=models.CASCADE)
# r is the field added
r = models.ForeignKey('R', null=True, default=None, on_delete=models.SET_NULL)
class M1(models.Model):
desc = models.CharField(max_length=16)
class M2(models.Model):
desc = models.CharField(max_length=16)
class R(models.Model):
desc = models.CharField(max_length=16)
m1 = models.ForeignKey('M1', on_delete=models.CASCADE)
m2 = models.ForeignKey('M2', on_delete=models.CASCADE)
from random import randint
from django.db import connection, reset_queries
from django.db.models import F, OuterRef, Subquery
from myapp.models import L1, L2, M1, M2, R
# create random data
for m in range(10):
M1.objects.create(desc=f'M1_{m:02d}')
M2.objects.create(desc=f'M2_{m:02d}')
for r in range(40):
R.objects.create(desc=f'R_{r:02d}', m1_id=randint(1,10), m2_id=randint(1,10))
for l1 in range(20):
L1.objects.create(desc=f'L1_{l1:02d}', m1_id=randint(1,10))
for l2 in range(100):
L2.objects.create(desc=f'L2_{l2:02d}', l1_id=randint(1,20), m2_id=randint(1,10))
# use subquery to annotate model - success
reset_queries()
subquery = Subquery(R.objects.filter(m2_id=OuterRef('m2_id'), m1_id=OuterRef('l1__m1_id')).values('id')[:1])
annotated = L2.objects.all().annotate(_r_id=subquery)
annotated_l=list(annotated)
print(connection.queries[-1])
# query SQL-1
# use subquery to annotate and update model - failure
reset_queries()
annotated.update(r_id=F('_r_id'))
# ...
# django.db.utils.ProgrammingError: missing FROM-clause entry for table "myapp_l1"
# LINE 1: ...ECT U0."id" FROM "myapp_r" U0 WHERE (U0."m1_id" = "myapp_l1"...
# ^
print(connection.queries[-1])
# produces SQL-2
SELECT
"myapp_l2"."id",
"myapp_l2"."desc",
"myapp_l2"."l1_id",
"myapp_l2"."m2_id",
"myapp_l2"."r_id",
(
SELECT
U0."id"
FROM
"myapp_r" U0
WHERE (U0."m1_id" = "myapp_l1"."m1_id"
AND U0."m2_id" = "myapp_l2"."m2_id")
LIMIT 1) AS "_r_id"
FROM
"myapp_l2"
INNER JOIN "myapp_l1" ON ("myapp_l2"."l1_id" = "myapp_l1"."id")
UPDATE
"myapp_l2"
SET
"r_id" = (
SELECT
U0."id"
FROM
"myapp_r" U0
WHERE (U0."m1_id" = "myapp_l1"."m1_id"
AND U0."m2_id" = "myapp_l2"."m2_id")
LIMIT 1)
WHERE
"myapp_l2"."id" IN (
SELECT
V0."id"
FROM
"myapp_l2" V0
INNER JOIN "myapp_l1" V1 ON (V0."l1_id" = V1."id"))
The following finally did the trick.以下终于成功了。 This was inspired by the answer here .
这是受到这里答案的启发。 Only, for this case, a nested subquery has to be used.
只有在这种情况下,才必须使用嵌套子查询。
For the record, performance was pretty good.根据记录,性能相当不错。 It took around 50sec to update 1.5M L2 objects from 830K L1 objects, and 12K R objects.
从 830K L1 对象和 12K R 对象更新 1.5M L2 对象大约需要 50 秒。
from django.db import connection, reset_queries
from django.db.models import OuterRef, Subquery
from myapp.models import L1, L2, M1, M2, R
# create queryset with annotation
subquery = Subquery(R.objects.filter(m2_id=OuterRef('m2_id'), m1_id=OuterRef('l1__m1_id')).values('id')[:1])
annotated = L2.objects.annotate(_r_id=subquery)
# use the queryset in a subquery to get the annotation value
reset_queries()
L2.objects.update(r_id=Subquery(annotated.filter(id=OuterRef('id')).values('_r_id')[:1]))
print(connection.queries[-1])
# produces SQL-good
# verify results with a loop
for l2 in L2.objects.all():
r = R.objects.filter(m1_id=l2.l1.m1_id, m2=l2.m2_id).first()
print(f'{str(r == l2.r):5s} {str(r):10s} {str(l2.r):10s}')
UPDATE
"myapp_l2"
SET
"r_id" = (
SELECT
(
SELECT
U0."id"
FROM
"myapp_r" U0
WHERE (U0."m1_id" = V1."m1_id"
AND U0."m2_id" = V0."m2_id")
LIMIT 1) AS "_r_id"
FROM
"myapp_l2" V0
INNER JOIN "myapp_l1" V1 ON (V0."l1_id" = V1."id")
WHERE
V0."id" = "myapp_l2"."id"
LIMIT 1) '
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.