Querysets now utilises a local cache

Changed __len__ behavour in the queryset (#247, #311)

docs/changelog.rst

@@ -4,6 +4,8 @@ Changelog
 
 Changes in 0.8.0
 ================
+- Querysets now utilises a local cache
+- Changed __len__ behavour in the queryset (#247, #311)
 - Fixed querying string versions of ObjectIds issue with ReferenceField (#307)
 - Added $setOnInsert support for upserts (#308)
 - Upserts now possible with just query parameters (#309)
@@ -25,7 +27,6 @@ Changes in 0.8.0
 - Added SequenceField.set_next_value(value) helper (#159)
 - Updated .only() behaviour - now like exclude it is chainable (#202)
 - Added with_limit_and_skip support to count() (#235)
-- Removed __len__ from queryset (#247)
 - Objects queryset manager now inherited (#256)
 - Updated connection to use MongoClient (#262, #274)
 - Fixed db_alias and inherited Documents (#143)

docs/upgrade.rst

@@ -235,12 +235,15 @@ update your code like so: ::
     mammals = Animal.objects(type="mammal").filter(order="Carnivora")  # The final queryset is assgined to mammals
     [m for m in mammals]                                               # This will return all carnivores
 
-No more len
------------
+Len iterates the queryset
+--------------------------
 
-If you ever did len(queryset) it previously did a count() under the covers, this
-caused some unusual issues - so now it has been removed in favour of the
-explicit `queryset.count()` to update::
+If you ever did `len(queryset)` it previously did a `count()` under the covers,
+this caused some unusual issues.  As `len(queryset)` is most often used by
+`list(queryset)` we now cache the queryset results and use that for the length.
+
+This isn't as performant as a `count()` and if you aren't iterating the
+queryset you should upgrade to use count::
 
     # Old code
     len(Animal.objects(type="mammal"))