moviewyrm/bookwyrm/suggested_users.py
2021-04-23 16:34:04 -07:00

130 lines
4.3 KiB
Python

""" store recommended follows in redis """
import math
from django.dispatch import receiver
from django.db.models import signals, Q
from bookwyrm import models
from bookwyrm.redis_store import RedisStore, r
from bookwyrm.views.helpers import get_annotated_users
class SuggestedUsers(RedisStore):
""" suggested users for a user """
max_length = 10
def get_rank(self, obj):
""" get computed rank """
return obj.mutuals + (1.0 - (1.0 / (obj.shared_books + 1)))
def store_id(self, user): # pylint: disable=no-self-use
""" the key used to store this user's recs """
return "{:d}-suggestions".format(user.id)
def get_counts_from_rank(self, rank): # pylint: disable=no-self-use
""" calculate mutuals count and shared books count from rank """
return {
"mutuals": math.floor(rank),
"shared_books": int(1 / (-1 * (1 % rank - 1))),
}
def get_objects_for_store(self, store):
""" a list of potential follows for a user """
user = models.User.objects.get(id=store.split("-")[0])
return get_annotated_users(
user,
~Q(id=user.id),
~Q(followers=user),
~Q(follower_requests=user),
bookwyrm_user=True,
)
def get_stores_for_object(self, obj):
return [self.store_id(u) for u in self.get_users_for_object(obj)]
def get_users_for_object(self, obj): # pylint: disable=no-self-use
""" given a user, who might want to follow them """
return models.User.objects.filter(
local=True,
).exclude(following=obj)
def rerank_obj(self, obj):
""" update all the instances of this user with new ranks """
pipeline = r.pipeline()
for store_user in self.get_users_for_object(obj):
annotated_user = get_annotated_users(
store_user,
id=obj.id,
).first()
pipeline.zadd(
self.store_id(store_user),
self.get_value(annotated_user),
xx=True
)
pipeline.execute()
def rerank_user_suggestions(self, user):
""" update the ranks of the follows suggested to a user """
self.populate_store(self.store_id(user))
def get_suggestions(self, user):
""" get suggestions """
values = self.get_store(self.store_id(user), withscores=True)
results = []
# annotate users with mutuals and shared book counts
for user_id, rank in values[:5]:
counts = self.get_counts_from_rank(rank)
user = models.User.objects.get(id=user_id)
user.mutuals = counts["mutuals"]
user.shared_books = counts["shared_books"]
results.append(user)
return results
suggested_users = SuggestedUsers()
@receiver(signals.post_save, sender=models.UserFollows)
# pylint: disable=unused-argument
def update_suggestions_on_follow(sender, instance, created, *args, **kwargs):
""" remove a follow from the recs and update the ranks"""
if (
not created
or not instance.user_subject.local
or not instance.user_object.discoverable
):
return
suggested_users.bulk_remove_objects_from_store(
[instance.user_object], instance.user_subject
)
suggested_users.rerank_obj(instance.user_object)
@receiver(signals.post_save, sender=models.ShelfBook)
@receiver(signals.post_delete, sender=models.ShelfBook)
# pylint: disable=unused-argument
def update_rank_on_shelving(sender, instance, *args, **kwargs):
""" when a user shelves or unshelves a book, re-compute their rank """
if not instance.user.discoverable:
return
suggested_users.rerank_obj(instance.user)
@receiver(signals.post_save, sender=models.User)
# pylint: disable=unused-argument, too-many-arguments
def add_or_remove_on_discoverability_change(
sender, instance, created, raw, using, update_fields, **kwargs
):
""" make a user (un)discoverable """
if not update_fields or not "discoverable" in update_fields:
return
if created:
suggested_users.rerank_user_suggestions(instance)
if instance.discoverable:
suggested_users.rerank_obj(instance)
elif not created and not instance.discoverable:
suggested_users.remove_object_from_related_stores(instance)