Note that drop_collection is not modified since there is no
guarantee that a GridFS collection holds files for only one
Document class. Otherwise you could drop files for other fields
or documents accidentally.
No longer demands FileField be an actual instance of file, but instead
checks whether object has a 'read' attribute. Fixes read() on
GridFSProxy to return an empty string on read failure, or None if file
does not exist.
No change required to upgrade to multiple databases. Aliases are used
to describe the database and these can be manually registered or fall
through to a default alias using connect.
Made get_connection and get_db first class members of the connection class.
Old style _get_connection and _get_db still supported.
Refs: #84#87#93#215
For documents that do not have a value set for a given field, most field types
return None (or [] in the case of ListField). This makes it easy to test
whether a field has been set using "if doc.field". FileFields, on the other
hand, always return a GridFSProxy. Adding GridFSProxy.__nonzero__ which simply
checks for a grid_id allows the same boolean-test pattern for FileFields, as
well.
All saves on exisiting items do set / unset operations only on changed fields.
* Note lists and dicts generally do set operations for things like pop() del[key]
As there is no easy map to unset and explicitly matches the new list / dict
fixes#18
* Handles the efficient lazy dereferencing of DBrefs.
* Handles complex nested values in ListFields and DictFields
* Allows for both strictly declared ListFields and DictFields where the embedded
value must be of a field type or no restrictions where the values can be a mix
of field types / values.
* Handles DBrefences of documents where allow_inheritance = False.
Allow searching multiple levels deep in DictFields
Allow DictField entries containing strings to use matching operators
Thanks again to @theojulien for the initial code #108
