Further queryset filter parameter for similar_objects method

[askowronek]

Hi, as far as I can see the current implementation of TaggableManager::similar_objects lacks the ability to specify further (exclude) filters for the result set. Thanks for investigating this issue.

[alex]

That's correct. As the documentation (http://packages.python.org/django-taggit/api.html#similar-objects) says, this method returns a list, not a queryset. I don't think there's anything we can do about this.

[alex]

I haven't thought of anything, so I'm closing this.

[askowronek]

sorry for being late to explain my feature request.

The current implementation requires me to call similar_objects once to get a list of all objects which look the same according to their tags. When I want to apply further filter on that result set, I need to get the IDs of all returned objects first and then do a second query with my filters and the ID list.

I owe you a beer if you could improve the similar_objects method to accept a filter queryset that applies on the queryset to retrieve similar objects ;D.

[alex]

I've reopened this to think about it some more.

[askowronek]

Bump I really don't want to hurry you since a good job needs the time it requires but I need that feature soon and I believe you can do the job much better here then me.

[joshmaker]

I have a branch of Taggit in which I have implemented what seems to be a simple fix for this problem.

Here is how my similar_objects method looks:

def similar_objects(self, num=None, **filters):
    lookup_kwargs = self._lookup_kwargs()
    lookup_keys = sorted(lookup_kwargs)
    qs = self.through.objects.values(*lookup_kwargs.keys())
    qs = qs.annotate(n=models.Count('pk'))
    qs = qs.exclude(**lookup_kwargs)
    subq = self.all()
    qs = qs.filter(tag__in=list(subq))
    qs = qs.order_by('-n')

    if filters is not None:
        qs = qs.filter(**filters)

    if num is not None:
        qs = qs[:num]

    // etc... (the rest is the same)

Now a user can easily specify both a count and any filter methods that they need when doing a similar_objects query.

[vdboor]

I'd love to have some extra filtering in place.

Right now I'm writing a FAQ app, and would love to have it return "similar questions". That means creating a "through table" right now, because otherwise other objects will also be returned.

[melinath]

+1 on this.

[b-jam]

+1, perhaps order_by would be good too

[avorio]

+1

Also, the ability to retrieve only objects of a specific given model would be great!

[legshort]

+1

[vdboor]

FYI, this is what I've eventually written back then: https://github.com/django-fluent/django-fluent-utils/blob/master/fluent_utils/softdeps/taggit.py#L73 this code could clearly be optimized, but it could be a good starting point.

[rtpg]

#709 is a dupe of this issue

[rtpg]

looking at this now (after many extra years), I think it might be possible for us to solve this. Django's ORM has made many strides, though the query itself might end up being extra gnarly