Optimizing appengine entity key usage

Question

Should I care about locality of entities on the Google App Engine datastore? Should I use custom entity key names for that?

For example, I could use "$article_uuid,$comment_id" as the key name of a Comment entity. Will it improve the speed of fetching all comments for an article? Or is it better to use shorter keys?

Is it a good practice to use the key in this way? I could use the "$article_uuid,$comment_id" key name also instead of an index:

def get_comments(article_uuid, limit=1000):
    key_prefix=db.Key.from_path('Comment', article_uuid)
    q = Comment.gql("where __key__ > :key_prefix and __key__ < :range_end",
        key_prefix=key_prefix, range_end=key_prefix+chr(ord(',')+1))
    return q.fetch(limit)

Answer 1

The locality of your data will be improved with your key_name scheme (ref, see slide 40) - since your key_name is prefixed with the corresponding article's ID, comments for a given article should be stored near each other.

The key_name you proposed doesn't seem like it would be too long. I don't think you'll see too much difference between that and shorter keys in terms of storage space or serialization/deserialization time. I expect that the size of the Comment entity will be dominated by the rest of the entity.