A field that returns a python object.
pip install django-register-fieldThere have been breaking changes made between V1 and V2. Please read the migration docs.
The RegisterField allows storing specific objects in the database and retrieving them directly from the model. They will be stored as strings, with a Register that is used to keep track of which string maps to what object. It cannot be used to store objects dynamically, but can be very useful for separating some logic into their related classes, reducing the number of if/else required in the model methods.
There are 2 ways to implement the fields: through the RegisterChoices, or by manually implementing the register. The former makes it easy to implement and use, while the latter offers more flexibility.
The implementation of the RegisterChoices is very similar to django's default Choices, but even simpler. It looks like this:
from dataclasses import dataclass
from django_register import RegisterChoices
@dataclass(unsafe_hash=True)
class MyOptions:
some_field: str
some_description: str
class SomeRegisterChoices(RegisterChoices):
OPTION_1 = MyOptions(some_field='field_name', some_description='field_description')
OPTION_2 = MyOptions(some_field='field_name_2', some_description='field_description_2')SomeRegisterChoices is now usable in a RegisterField on a model:
from django_register import RegisterField
class SomeModel(models.Model):
my_field = RegisterField(choices=SomeRegisterChoices)Note that the objects passed to RegisterChoices must be hashable. This is because the register keeps track of the relationship between the label and the object in both directions, so the object needs to be usable as a key in a dictionary.
By default, the label used in the database will be the same as the variable name of the choices, in lower case. This can be changed by having a key attribute on the object. If one is set, that is what will be used database side. Similarly, the verbose field used in the .choices to be displayed in Django admin will be the variable name with all underscores replaced by a space, and .title applied to it. This can be changed by setting the label attribute on the object.
In the background, RegisterChoices takes care of setting and handling the Register for you. You can also create it and set it manually, if using the Choices is not an option.
The method with the Choices is very good when what you want to keep is information. However, if there is logic that changes as well, you can quickly end up with circular dependencies,
where the RegisterField needs the RegisterChoices, which in turn needs you objects. If your code calling the model then is in those same objects, you get the circularity. In that situation,
it might be better to keep a structure where the RegisterField has no knowledge of the objects it holds. This can be done by creating the Register manually and then adding the objects
to it. This register can then be passed directly to the RegisterField.
To set the register manually, you first need to instantiate a Register:
from django_register import Register
register = Register()With that done, you can register objects:
register.register(some_object, db_key='some_label')db_key is optional. If not set, the key must be set on the object, otherwise a ValueError is raised. You can then pass the register to the RegisterField directly:
from django_register import RegisterField
class SomeModel(models.Model):
my_field = RegisterField(register=register)You can also have it be part of the Model class:
from django_register import RegisterField
class SomeModel(models.Model):
register = Register()
register.register(some_object, db_key='some_label')
my_field = RegisterField(register=register)You can even use it as a decorator:
@register.register
class SomeClass:
key = "some_class"
@register.register(db_key="some_other_class")
class SomeOtherClass:
passNote that if using that technique, you are responsible for keeping track of your object. The RegisterChoices make it easier to keep your objects for comparison and use them outside of the model, but both methods will give the same results database side.
If you need to set the register values dynamically, you can do so after the fact by using the register directly. However in that case, you need to provide a max_length if your database does not support having a CharField without a max_length. That is because in the background, a CharField is used to store the key in the database.
You can set the values dynamically like this:
class SomeApp(AppConfig):
def ready(self):
register.register(some_object, db_key='some_label')It does not have to be in the ready method, values can be added to the register anywhere, however you should be very careful about where you allow adding values and when. If the value is not available somewhere in the code, it will return the unknown_item_class instead of the expected object.
Removing items from the register requires some consideration. The string in the database is still there unless you create a migration, and it is possible it will cause issues due to the class linked to it not existing anymore. Before version 1.0.8, this would fail dramatically, giving a ValidationError and stopping anyone from interacting with the database items it was linked to, not even to delete (in most cases). In that case, the only solution would be to add the item back, delete or edit the affected database rows, then remove the item again.
From v1.0.8 onward, it will no longer fail dramatically, but rather return a default object. This default object will only contain the label. If you want to define other defaults for fields that are accessed, you can pass an unknown_item_class parameter to the RegisterField, to the register itself, or set an _UNKNOWN_ attribute in the RegisterChoices. All these methods will give the same result: defining the default object to be used when the item can no longer be found. Note that the label will not be passed, but rather set after the creation of the object, so make sure that the __init__ takes no arguments. It is also recommended to use a different class from the one used to set the options, if only to allow checking if the object is an instance of said class later.
class SomeRegisterChoices(RegisterChoices):
OPTION_1 = MyOptions(some_field='field_name', some_description='field_description')
OPTION_2 = MyOptions(some_field='field_name_2', some_description='field_description_2')
_UNKNOWN_ = UnknownOptionclass SomeModel(models.Model):
my_field = RegisterField(register=register, unknown_item_class=UnknownOption)register = Register(unknown_item_class=UnknownOption)If using with rest_framework, there is a Serializer Field already built in to be used by the Serializer. You simply need to set the field as such:
from django_register.rest_framework import RegisterField
class SomeModelSerializer(serializers.ModelSerializer):
some_register_field = RegisterField()
class Meta:
model = SomeModel
fields = ('some_register_field',)Behind the scenes, the serializer field goes and gets the register from the model field to get the work done.
Since v1.0.7, it is now possible to pass keys to the serializer RegisterField to tell it to return an object with those keys. This is useful if you want to expose certain values on the object to the API.
It would look something like this:
from django_register.rest_framework import RegisterField
class SomeModelSerializer(serializers.ModelSerializer):
some_register_field = RegisterField(keys=['key', 'some_value', 'some_other_value'])
class Meta:
model = SomeModel
fields = ('some_register_field',)The example above would return the following JSON:
{
"some_register_field": {
"key": "obj_label",
"some_value": "value",
"some_other_value": "other_value"
}
}Note that if the key or label is not set on the object directly, the default value that is set automatically will be returned, so they can always be used this way.
This library is tested against the following versions:
- Python 3.10, 3.11, 3.12 and 3.13
- Django 4.2 and 5.2
- Django REST Framework 3.14, 3.15 and 3.16
While it probably works with some other versions, it is not garanteed.
Huge thanks to Tim Schilling from Better Simple for his article that was the catalyst behind the idea for this library.