datastore: default to empty dataset

[pcostell]

When working with datasets, the Datastore backend will normalize any empty datasets to the dataset of the calling application. This should be preferred to prepending "s~" to dataset ids as that is not always a valid operation.

[tseaver]

The 'to_protobuf' method of gcloud.datastore.key.Key jams a 's~' prefix onto the key's dataset_id, if the ID does not already have a 's~' or 'e~' prefix:

https://github.com/GoogleCloudPlatform/gcloud-python/blob/1f2aaf3606802f4f02eeb30e22e1082ef7f4594e/gcloud/datastore/key.py#L60

@pcostell Is that the normalization you are thinking of?

[pcostell]

Yes that's the normalization that is not needed. The datastore backend is smart about looking up the correct dataset for the application.

[tseaver]

The fix in #259 is not correct / complete. See:

• my report on breakage to the demo
• @pcostell's follow up