Possible to do an `in` `lookup_type` through the django-filter URL parser?

Question

I'm using django-filter with django-rest-framework and I'm trying to instantiate a filter that accepts lists of numbers for filtering the query set down

class MyFilter(django_filters.FilterSet):   
    ids = django_filters.NumberFilter(name='id',lookup_type='in')
    class Meta:
        model = MyModel
        fields = ('ids',)

class MyModelViewSet(viewsets.ModelViewSet):
    queryset = MyModel.objects.all()
    serializer_class = MyModelSerializer
    filter_class = MyFilter

If I pass in a comma separated list of integers, the filter is ignored altogether.

If I pass in a single integer, it gets through django-filter into django's form validator and complains:

'Decimal' object is not iterable

Is there a way to create a django-filter object which can handle a list of integers and properly filter down the queryset?

Answer 1

For better or worse, I created a custom filter for this:

class IntegerListFilter(django_filters.Filter):
    def filter(self,qs,value):
        if value not in (None,''):
            integers = [int(v) for v in value.split(',')]
            return qs.filter(**{'%s__%s'%(self.name, self.lookup_type):integers})
        return qs

Which is used like:

class MyFilter(django_filters.FilterSet):   
    ids = IntegerListFilter(name='id',lookup_type='in')
    class Meta:
        model = MyModel
        fields = ('ids',)

class MyModelViewSet(viewsets.ModelViewSet):
    queryset = MyModel.objects.all()
    serializer_class = MyModelSerializer
    filter_class = MyFilter

Now my interface accepts comma-delimited lists of integers.

Answer 2

I know this is an old post, but there is now a better solution. The change that makes it correct is posted here.

They added a BaseInFilter and a BaseRangeFilter. The documentation is here.

Big picture, BaseFilter checks for CSV, and then when mixed with another filter it does what you are asking. Your code can now be written like:

class NumberInFilter(filters.BaseInFilter, filters.NumberFilter):
    pass

class MyModelViewSet(viewsets.ModelViewSet):
    ids = NumberInFilter(name='id', lookup_expr='in')

    class Meta:
        model = MyModel
        fields = ['ids']

Answer 3

Here's a complete solution:

from django_filters import Filter, FilterSet
from rest_framework.filters import DjangoFilterBackend
from rest_framework.viewsets import ModelViewSet
from .models import User
from .serializers import UserSerializer


class ListFilter(Filter):

    def filter(self, qs, value):
        if not value:
            return qs

        self.lookup_type = 'in'
        values = value.split(',')
        return super(ListFilter, self).filter(qs, values)


class UserFilter(FilterSet):
    ids = ListFilter(name='id')

    class Meta:
        model = User
        fields = ['ids']


class UserViewSet(viewsets.ModelViewSet):
    serializer_class = UserSerializer
    queryset = User.objects.all()
    filter_backends = (DjangoFilterBackend,)
    filter_class = UserFilter

Answer 4

According to a post in the django-filter issues:

from django_filters import Filter
from django_filters.fields import Lookup

class ListFilter(Filter):
    def filter(self, qs, value):
        return super(ListFilter, self).filter(qs, Lookup(value.split(u","), "in"))

I have personally used this without any issue in my projects, and it works without having to create a per-type filter.

Answer 5

Based on @yndolok answer I have come to a general solution. I think filtering by a list of ids is a very common task and therefore should be included in the FilterBackend:

class ListFilter(django_filters.Filter):

    """Class to filter from list of integers."""

    def filter(self, qs, value):
        """Filter function."""
        if not value:
            return qs
        self.lookup_type = 'in'
        try:
            map(int, value.split(','))
            return super(ListFilter, self).filter(qs, value.split(','))
        except ValueError:
            return super(ListFilter, self).filter(qs, [None])


class FilterBackend(filters.DjangoFilterBackend):

    """A filter backend that includes ListFilter."""

    def get_filter_class(self, view, queryset=None):
        """Append ListFilter to AutoFilterSet."""
        filter_fields = getattr(view, 'filter_fields', None)

        if filter_fields:
            class AutoFilterSet(self.default_filter_set):
                ids = ListFilter(name='id')

                class Meta:
                    model = queryset.model
                    fields = list(filter_fields) + ["ids"]

            return AutoFilterSet

        else:
            return super(FilterBackend, self).get_filter_class(view, queryset)

Answer 6

Uptodate solution:

from django_filters import rest_framework as filters

name-->field_name

lookup_type-->lookup_expr

class IntegerListFilter(filters.Filter):
    def filter(self,qs,value):
        if value not in (None,''):
            integers = [int(v) for v in value.split(',')]
            return qs.filter(**{'%s__%s'%(self.field_name, self.lookup_expr):integers})
        return qs

class MyFilter(filters.FilterSet):   
    ids = IntegerListFilter(field_name='id',lookup_expr='in')
    class Meta:
        model = MyModel
        fields = ('ids',)

class MyModelViewSet(viewsets.ModelViewSet):
    queryset = MyModel.objects.all()
    serializer_class = MyModelSerializer
    filter_class = MyFilter

Answer 7

As I have answered here DjangoFilterBackend with multiple ids, it is now pretty easy to make a filter that accepts list and validates the contents

For Example:

from django_filters import rest_framework as filters


class NumberInFilter(filters.BaseInFilter, filters.NumberFilter):
    pass

class MyFilter(filters.FilterSet):
    id_in = NumberInFilter(field_name='id', lookup_expr='in')

    class Meta:
        model = MyModel
        fields = ['id_in', ]

This will accept a list of integers from a get parameter. For example /endpoint/?id_in=1,2,3