optimise query

Author: Dylan-J

app/api/views/users.py

@@ -54,9 +54,15 @@ class TelegramUserViewSet(viewsets.ReadOnlyModelViewSet):
     ordering = ['-last_seen']
 
     def get_queryset(self):
+        # No .distinct() here: the channel/tag filters join rows that are
+        # unique per user (memberships are unique_together (user, channel);
+        # tags are an M2M through table), so they cannot duplicate users.
+        # DISTINCT forced DRF's pagination COUNT into
+        # COUNT(*) FROM (SELECT DISTINCT <all columns> ...), detoasting every
+        # profile_photo_base64 blob in the table — ~30 s per page load.
         return TelegramUser.objects.all().annotate(
             group_count_annotation=Count('memberships', distinct=True)
-        ).distinct()
+        )
 
     def get_serializer_class(self):
         if self.action == 'retrieve':