使用旧数据库在 django 项目中使用复合主键
我有一个旧数据库,其中一些表包含复合主键.我通过运行 manage.py inspectdb
命令得到的模型如下所示.
class MyTable(models.Model):field1_id = models.IntegerField(db_column='field1id', primary_key=True)is_favorite = models.BooleanField(db_column='isfavorite', default=False, null=False)is_admin = models.BooleanField(db_column='isadmin', default=False, null=False)role = models.IntegerField(默认=USER_GUEST,null=False)field2 = models.BooleanField(null=False, default=True)field3 = models.BooleanField(null=False, default=True)is_active = models.BooleanField(db_column='isactive', null=False, default=True)用户 = 模型.ForeignKey(自定义用户,models.DO_NOTHING,db_column='userid',primary_key=True)元类:管理 = 错误db_table = '我的表'unique_together = (('user', 'field1_id'),)
我可以正常获取数据.但是,当我想在某个模型实例上运行 save()
命令时会出现问题.django 执行的查询不正确.例如:
但我需要:
{'time': '0.000', 'sql': 'UPDATE `mytable` SET `isfavorite` = 0, `isadmin` = 1, `role` = 3, `field2` = 1, `field3` = 1, `isactive` = 1 WHERE `mytable`.`field1id` = 12' AND `mytable`.`userid` = 1'}
由于 django 不支持复合主键,那么克服该问题的最佳解决方案是什么?请注意,它是旧数据库表,没有 AutoField
.
添加旧表结构
+------------+------------+------+-----+---------+-----+|领域 |类型 |空 |钥匙 |默认 |额外 |+------------+------------+------+-----+---------+-------+|用户名 |整数(11) |否 |PRI |空 |||field1id |整数(11) |否 |PRI |空 |||最喜欢 |整数(11) |否 ||0 |||isadmin |整数(11) |否 ||0 |||角色|整数(11) |是 ||空 |||场2 |tinyint(1) |否 ||1 |||字段3 |tinyint(1) |否 ||1 |||活跃 |tinyint(4) |否 ||1 ||+------------+------------+------+-----+---------+-------+
解决方案 不幸的是 django 没有 复合主键(复合主键也叫多列主键)
有一个第三方库,不出所料地将其命名为 django-compositekey 这使得它成为可能创建一个具有多列主键的模型.但是,该项目并未得到维护,并且与最新版本的 django 不兼容.
最好的办法是在表中添加一个新列,该列是一个 AutoField,并使其成为新的主键.目前构成主键的字段可以标记为 unique_together.虽然这对于一些查询来说可能并不理想.对大多数人来说已经足够了.
如果您使用当前的表结构更新您的问题(如您的 sql 控制台中所示),我将更新我的答案以显示模型的外观.
更新有很多不同的方法可以删除旧的复合主键并用新的自动增量主键替换它.这是最简单的,它只涉及 SQL
CREATE TABLE newtable LIKE mytable;ALTER TABLE newtable DROP PRIMARY KEY;ALTER TABLE newtable ADD `id` INT NOT NULL AUTO_INCREMENT PRIMARY KEY;将表 mytable 重命名为 mytable_old;将 TABLE newtable 重命名为 mytable;INSERT INTO mytable(userid,field1id, 其余字段)SELECT * FROM mytable_old;
然后编辑您的模型并从这些字段中删除 primary_key=True 标志.
脚注:
某些数据库(例如 sqlite)不支持 CREATE TABLE
中的 LIKE
子句.在这些情况下,您必须查找原始表的 create table 语句,并在编辑表名后复制粘贴.从您的表结构来看,您似乎正在使用 mysql 并且它支持 LIKE
子句.
I have a legacy database, where some table contains composite primary key. The model I get by running manage.py inspectdb
command looks like this.
class MyTable(models.Model):
field1_id = models.IntegerField(db_column='field1id', primary_key=True)
is_favorite = models.BooleanField(db_column='isfavorite', default=False, null=False)
is_admin = models.BooleanField(db_column='isadmin', default=False, null=False)
role = models.IntegerField(default=USER_GUEST, null=False)
field2 = models.BooleanField(null=False, default=True)
field3 = models.BooleanField(null=False, default=True)
is_active = models.BooleanField(db_column='isactive', null=False, default=True)
user = models.ForeignKey(
CustomUser, models.DO_NOTHING, db_column='userid', primary_key=True)
class Meta:
managed = False
db_table = 'mytable'
unique_together = (('user', 'field1_id'),)
I can fetch data normally. However, the problem arises when I want to run save()
command on some model instance. The query django executes is not correct one. For example:
>>> from web_services.apps.my_app.models import MyTable
>>> g = MyTable.objects.get(field1_id=12)
>>> g.is_active = True
>>> g.save()
>>> connection.queries[-1]
{'time': '0.000', 'sql': 'UPDATE `mytable` SET `isfavorite` = 0, `isadmin` = 1, `role` = 3, `field2` = 1, `field3` = 1, `isactive` = 1 WHERE `mytable`.`field1id` = 12'}
But I need:
{'time': '0.000', 'sql': 'UPDATE `mytable` SET `isfavorite` = 0, `isadmin` = 1, `role` = 3, `field2` = 1, `field3` = 1, `isactive` = 1 WHERE `mytable`.`field1id` = 12' AND `mytable`.`userid` = 1'}
Since django does not support composite primary key, what would be the best solution to overcome that problem? Note, it's legacy database table and it didn't have AutoField
.
EDIT: adds legacy table structure
+------------+------------+------+-----+---------+-------+
| Field | Type | Null | Key | Default | Extra |
+------------+------------+------+-----+---------+-------+
| userid | int(11) | NO | PRI | NULL | |
| field1id | int(11) | NO | PRI | NULL | |
| isfavorite | int(11) | NO | | 0 | |
| isadmin | int(11) | NO | | 0 | |
| role | int(11) | YES | | NULL | |
| field2 | tinyint(1) | NO | | 1 | |
| field3 | tinyint(1) | NO | | 1 | |
| isactive | tinyint(4) | NO | | 1 | |
+------------+------------+------+-----+---------+-------+
解决方案
Unfortunately django does not yet have a composite primary key (composite primary keys are also called multi column primary keys)
There is a third party library, unsurprisingly name django-compositekey that makes it possible to create a model that has a multi column primary key. However that project is not maintained and not compatible with the latest versions of django.
Your best bet, is to add a new column into your table which is a an AutoField and make it the new primary key. The fields that make up the primary key at present can then be marked as unique_together. While this may not be ideally for a few queries. It's good enough for most.
If you update your question with the current table structure (as shown in your sql console) I will update my answer to show what the model will look like.
Update There are lots of different ways in which you can drop the old composite primary key and replace it with a new auto increment primary key. Here is the easiest, it involves just SQL
CREATE TABLE newtable LIKE mytable;
ALTER TABLE newtable DROP PRIMARY KEY;
ALTER TABLE newtable ADD `id` INT NOT NULL AUTO_INCREMENT PRIMARY KEY;
RENAME TABLE mytable to mytable_old;
RENAME TABLE newtable to mytable;
INSERT INTO mytable(userid,field1id, rest of the fields)
SELECT * FROM mytable_old;
Then edit your model and remove the primary_key=True flag from those fields.
Footnote:
Some databases like sqlite for example does not support the LIKE
clause in a CREATE TABLE
. In these cases, you will have to look up the create table statement for the original table, and copy paste after editing the table name. Judging by your table structure you seem to be using mysql and it supports the LIKE
clause.
相关文章