I've been trying to save and restore my dataset including keys using the bulkloader as docuemnted. There are a few undocumented things that are not mentioned and quite important. First is how to save your model keys. If we had the following model
name = db.StringProperty(required=True)
Then the following is defined to export the model
[('__key__', str, None),
('name', str, None),
Notice the '__key__' Field. This saves the key in the csv for that model.
The following should load the model back in with the key, updating it if the model already exists.
[('key', lambda x: x.decode('utf-8')),
('name', lambda x: x.decode('utf-8')),
Notice the special field key. This is Used in the Model.__init__ constructor in file "google/appengine/ext/db/__init__.py" Line 655 to recreate the Key from the string. This is created in bulkloader.py line 1260.
This key is then used when _populate_entity "google/appengine/ext/db/__init__.py" is called from bulkload at line 1264
_populate_entity then fills out the 'id' field with using the id() call on the key. The Issue with this is that key.id() returns a long
The Entity is then Created but line 422 of file "google/appengine/api/datastore.py" checks to see if the id is an int but it's a long so the creation fails. Can the id be a long?
My solution is to cast the return from "self._key.id()" at "google/appengine/ext/db/__init__.py" line 823 to an int.
This then fixs stuff nicly.