This allows an index spec to reference arbitrary keys of a DictField:
class MyDoc(Document):
frobs = DictField()
meta = {
"indexes": ["frobs.fmep", "frobs.gorp"],
}
Using a cursos in an if statement:
cursor = Collection.objects
if cursor:
(...)
Will open all documents, because there are not an __nonzero__ method.
This change check only one document (if present) and returns True or False.
In the latest released version of dateutil, there's a bug whereby a TypeError
can be raised whilst parsing a date. This is because it calls a method which
it expects to return 2 arguments, however it can return 1 depending upon the
input, which results in a TypeError: ArgType not iterable exception. Since
this is equivalent to a failed parse anyway, we can treat it the same as a
ValueError.
Migrate changes to include updating single elements of ListFields as
well as MapFields by adding the same changes to BaseList. This is
done by ensuring all BaseDicts and BaseLists have the correct name
from the base of the nearest (Embedded)Document, then marking changes
with their key or index when they are changed.
Tests also all fixed up.
Since v2.5, PyMongo has supported a "source" option, to specify a
particular database to authenticate against. This adds support for that
option, in the form of a "authentication_source" option to
register_connection.
When using nested MapFields from a document loaded from the database, the
nested dictionaries aren't converted to BaseDict, so changes aren't
marked.
This also includes a change when marking a field as changed to ensure that
nested fields aren't included in a $set query if a parent is already marked
as changed. Not sure if this could occur but it prevents breakage if it does.
