Installation:
pip install qfilters
Works both with Python 2 and Python 3.
What it is:
A lightweight package for django which does the filtering of django querysets. The central object - a filter - is a callable that takes a queryset as a parameter and returns a queryset:
filtered_queryset = filtr(Model.objects.all())
Django itself has a similar object - Q-object (django.db.models.Q
). Q-objects can be combined together (with |
, &
operations) or inverted (~
) and then passed to queryset.filter
.
With qfilters
, what you get in the most common case, is just a wrapper around the Q-object. However there are 2 features that may be the reasons to use the package:
1. Support for additional filter types.
For example, there is ValuesDictFilter
, which is constructed from a field list, that is passed to queryset.values
and retrieve a list of dictionaries, and a filtering function, which accepts that dict as a parameter.
This filters can be combined or inverted in the same way Q-objects do, so that using multiple filters would result in a single call to queryset.values
.
This is what it looks like in practice (all examples are taken from the qfilter
testsuite):
@ValuesDictFilter('@', fields_list=['traits__good_hunter'])
def nas_i_zdes_neploho_kormyat(obj):
return obj['traits__good_hunter'] is False # because it can be None
cats = nas_i_zdes_neploho_kormyat(self.CatsBreed.objects.all())
assert cats.exists()
There are also exotic variants (qfilters.exotic_types
) like QuerysetIterationHook
, which appends attributes to objects when queryset is iterated over. Another one is PropertyBasedFilter
, which can access object's attributes and even properties as if it were a regular django model object. The implementation is not very straightforward, still it passes the tests so far. Here is what it looks like:
class CatsBreed(models.Model):
# ...
traits = models.OneToOneField('Traits')
class Traits(models.Model):
#...
weight_min = models.FloatField(u'Вес от, кг', null=True)
weight_max = models.FloatField(u'Вес до, кг', null=True)
# if you can't specify min and max
weight = models.FloatField(u'Вес, кг', null=True)
@property
def kg(self):
return self.weight or self.weight_max
from qfilters import PropertyBasedFilter
@PropertyBasedFilter('@',
fields_list=['traits__weight', 'traits__weight_max'],
properties=['traits.kg'])
def light_cats(obj):
return not obj.traits.kg or obj.traits.kg < 3
assert light_cats(CatsBreed.objects.all()).exists()
2. Using class as a container: methods are filters
It is convenient to have an object, which can hold some context (for example, the view itself),
and let the methods be filters, and be able to access this context. qfilters.containers
provide this functionality, specifically, there is a MethodFilter
class:
from qfilter import MethodFilter
class ManyFilters(MethodFilter):
def filter__q(self):
return Q(name__in=[u'Сиамская', u'Норвежская лесная'])
@make_filter(PropertyBasedFilter,
fields_list=['traits__weight', 'traits__weight_max'],
properties=['traits.kg'])
def filter__big(self, cat):
return cat.traits.kg and cat.traits.kg > 5
def filter__q_yet_another(self):
return Q(name__in=[u'Персидская', u'Норвежская лесная'])
filters = ManyFilters()
cat = filters(CatsBreed.objects.all())[0]
The idea was born from the experience of using the django-rest-framework. There is a notion of filter backend (a class) which every view knows about. First I implemented a simple method-based filter backend, the possible return values for the methods were eiher a Q-object or a queryset. But then I got difficulties with debugging since the return value doesn't even know which method it came from. Thus, I decided it will be a good idea to have this filter object.
P.S. qfilters
does not provide a filter backend to use with django-rest-framework, but it's a piece of cake to write one.