Collection: further improve stoken performance.

We merged the two queries into one and we made it like in the view, so we
can now merge the two instead of having two implementations.

django_etebase/models.py

@@ -17,7 +17,8 @@ from pathlib import Path
 from django.db import models, transaction
 from django.conf import settings
 from django.core.validators import RegexValidator
-from django.db.models import Max
+from django.db.models import Max, Value as V
+from django.db.models.functions import Coalesce, Greatest
 from django.utils.functional import cached_property
 from django.utils.crypto import get_random_string
 
@@ -39,6 +40,8 @@ class Collection(models.Model):
     main_item = models.OneToOneField("CollectionItem", related_name="parent", null=True, on_delete=models.SET_NULL)
     owner = models.ForeignKey(settings.AUTH_USER_MODEL, on_delete=models.CASCADE)
 
+    stoken_id_fields = ["items__revisions__stoken", "members__stoken"]
+
     def __str__(self):
         return self.uid
 
@@ -56,9 +59,15 @@ class Collection(models.Model):
 
     @cached_property
     def stoken(self):
-        stoken1 = self.items.aggregate(stoken=Max("revisions__stoken"))["stoken"] or 0
-        stoken2 = self.members.aggregate(stoken=Max("stoken"))["stoken"] or 0
-        stoken_id = max(stoken1, stoken2)
+        aggr_fields = [Coalesce(Max(field), V(0)) for field in self.stoken_id_fields]
+        max_stoken = Greatest(*aggr_fields) if len(aggr_fields) > 1 else aggr_fields[0]
+        stoken_id = (
+            self.__class__.objects.filter(main_item=self.main_item)
+            .annotate(max_stoken=max_stoken)
+            .values("max_stoken")
+            .first()["max_stoken"]
+        )
+
         if stoken_id == 0:
             raise Exception("stoken is None. Should never happen")