An identity mapper for the Django ORM
© 2014-2017 Thomas Khyn © 2009 David Cramer
An identity mapper for the Django ORM. This is a fork of django-idmapper, which is no longer maintained.
django-idmap has been tested against django 1.8, 1.11 and 2.0 and the latest minor versions of Python 2 and 3 (Django 2.0 only supports Python 3).
If you like django-gm2m and are looking for a way to thank me and/or encourage future development, here is my BTC or BCH donation address: 1EwENyR8RV6tMc1hsLTkPURtn5wJgaBfG9.
What is it?
django-idmap is a Django application which:
- loads only once the instances in memory the first time they are needed
- shares them throughout your interpreter until the request is finished
Indeed, the default Django behavior is to expose different instances for the same database entry between the start and the end of the request. It has one main consequence: the temporary attributes you may set are lost if you want to access the same database object in another place in your code.
Deserialization (such as from the cache) will not use the identity mapper.
Tested with django 1.8+ and latest minor versions of python 2 and 3.
As straightforward as it can be, using pip:
pip install django-idmap
You also need to add 'idmap' to the INSTALLED_APPS setting.
To enable the identity mapper for a model, you simply need to make it inherit from idmap.models.IdMapModel instead of django.db.models.Model.
You can import idmap.models as you would import django.db.models. idmap.models exposes all that is exposed by django.db.models plus the IdMapModel model class.
You may of course mix and match IdMapModel and Model:
from idmap import models class MyModel(models.IdMapModel): name = models.CharField(...) fkey = models.ForeignKey('Other', on_delete=models.CASCADE) class Other(models.Model): name = models.CharField(...)
2 caching modes are available:
- Weak references mode: the instance will be removed from the cache once there are no more references to it. This is the default behavior
- Strong references mode: the instance will be loaded only once from the database and will be removed from the cache when it is flushed
If you want to use strong references for a particular model, simply set use_strong_refs to True in the derived model class’ Meta:
from idmap import models class MyModel(models.IdMapModel): class Meta: use_strong_refs = True [...]
In most cases, that’s all there is to do with django-idmap. Sometimes, you may need to flush the cache manually before the request is finished.
You can use:
- idmap.flush() to erase the whole cache
- IdMapModel.flush_instance_cache() to erase the cache for one model
- IdMapModel.flush_cached_instance(instance) to erase one instance from the cache
idmap.signals.pre_flush and idmap.signals.post_flush are sent before - respectively after - the cache is flushed. Connect handlers to these if you need to run code at these moments.
django-idmap flushes the cache when the request_finished or post_migrate signal are sent. This default behavior may be modified (at your own risk!) by disconnecting the idmap.signals.flush_idmap handler from these signals.
Multiple database support
In some cases, you may need to store instances of the same model in several databases. It is possible to tell django-idmap to also take the database into account when creating or getting instances:
class MyModel(models.IdMapModel): class Meta: multi_db = True [...]
This way, instance1_1 with primary key 1 in database db1 will be different from instance2_1 with primary key 1 in database db2:
>>> MyModel.objects.using('db1').create(pk=1) >>> MyModel.objects.using('db2').create(pk=1) >>> idmap.flush() >>> instance1_1 = MyModel.objects.using('db1').get(pk=1) >>> instance2_1 = MyModel.objects.using('db2').get(pk=1) >>> assert instance1_1 is instance 2_1 AssertionError
When using multiple databases, you may also flush only one database by providing its name to idmap.flush():
will only flush instances that were retrieved using the database db1. IdMapModel.flush_instance_cache can also take a db argument.
Similarly, a keyword-argument db is provided when the pre_flush and post_flush signals are sent. db is None if all databases are flushed (i.e. if no database alias was provided).
All instances of models and proxy models using the same base concrete class (let’s call this a proxy family) are stored in the same cache, and are accessible through all the members of the proxy family:
>>> class MyProxyModel(MyModel): >>> class Meta: >>> proxy = True >>> original = MyModel.objects.create(pk=1) >>> proxy = MyProxyModel.objects.create(pk=2) >>> assert original is MyProxyModel.get(pk=1)
In case you need to use a custom manager or custom metaclass on a model based on IdMapModel, you need to derive them from idmap’s own manager and metaclass:
>>> class MyModelBase(models.IdMapModelBase): >>> [...] >>> class MyManager(models.IdMapManager): >>> [...] >>> class MyModel(models.IdMapModel, metaclass=MyModelBase): # on python 3 >>> __metaclass__ = MyModelBase # on python 2 >>> objects = MyManager()