python - How does Django's Meta class work?


I am using Django which allows people to add extra parameters to a class by using class Meta.

class FooModel(models.Model):
    class Meta:

The only thing I found in Python's documentation was:

class FooMetaClass(type):

class FooClass:
    __metaclass__ = FooMetaClass

However, I don't think this is the same thing.

All Answers
  • Translate

    You are asking a question about two different things:

    1. Meta inner class in Django models:

      This is just a class container with some options (metadata) attached to the model. It defines such things as available permissions, associated database table name, whether the model is abstract or not, singular and plural versions of the name etc.

      Short explanation is here: Django docs: Models: Meta options

      List of available meta options is here: Django docs: Model Meta options

    2. Metaclass in Python:

      The best description is here: What is a metaclass in Python?

  • Translate

    Extending on Tadeck's Django answer above, the use of 'class Meta:' in Django is just normal Python too.

    The internal class is a convenient namespace for shared data among the class instances (hence the name Meta for 'metadata' but you can call it anything you like). While in Django it's generally read-only configuration stuff, there is nothing to stop you changing it:

    In [1]: class Foo(object):
       ...:     class Meta:
       ...:         metaVal = 1
    In [2]: f1 = Foo()
    In [3]: f2 = Foo()
    In [4]: f1.Meta.metaVal
    Out[4]: 1
    In [5]: f2.Meta.metaVal = 2
    In [6]: f1.Meta.metaVal
    Out[6]: 2
    In [7]: Foo.Meta.metaVal
    Out[7]: 2

    You can explore it in Django directly too e.g:

    In [1]: from django.contrib.auth.models import User
    In [2]: User.Meta
    Out[2]: django.contrib.auth.models.Meta
    In [3]: User.Meta.__dict__
    {'__doc__': None,
     '__module__': 'django.contrib.auth.models',
     'abstract': False,
     'verbose_name': <django.utils.functional.__proxy__ at 0x26a6610>,
     'verbose_name_plural': <django.utils.functional.__proxy__ at 0x26a6650>}

    However, in Django you are more likely to want to explore the _meta attribute which is an Options object created by the model metaclass when a model is created. That is where you'll find all of the Django class 'meta' information. In Django, Meta is just used to pass information into the process of creating the _meta Options object.

  • Translate

    Django's Model class specifically handles having an attribute named Meta which is a class. It's not a general Python thing.

    Python metaclasses are completely different.

  • Translate

    Answers that claim Django model's Meta and metaclasses are "completely different" are misleading answers.

    The construction of Django model class objects (that is to say the object that stands for the class definition itself; yes, classes are also objects) are indeed controlled by a metaclass called ModelBase, you can see that code here:

    And one of the things that ModelBase does is to create the _meta attribute on every Django model which contains validation machinery, field details, saving machinery and so forth. And, during this operation, anything that is specified in the model's inner Meta class is read and used within that process.

    So, while yes, in a sense Meta and metaclasses are different 'things', within the mechanics of Django model construction they are intimately related; understanding how they work together will deepen your insight into both at once.

    This might be a helpful source of information to better understand how Django models employ metaclasses.

    And this might help too if you want to better understand how objects work in general.

  • Translate

    Inner Meta Class Document This document of django Model metadata is “anything that’s not a field”, such as ordering options (ordering), database table name (db_table), or human-readable singular and plural names (verbose_name and verbose_name_plural). None are required, and adding class Meta to a model is completely optional.