Question

I’m designing datastore models and trying to decide the best approach when thinking about how the query filters are going to work.

Best if I just write out two examples. The first example is if I have a fixed “gender” property with just a string which can be set to either “male” or “female”.

class Person(db.Model):
    name = db.StringProperty()
    gender = db.StringProperty()

p1 = Person(name="Steve", gender="male")
p2 = Person(name="Jane", gender="female")
p1.put()
p2.put()

males = db.GqlQuery("SELECT * FROM Person WHERE gender = :1", "male")

The second example is if the Person entity is an expando model, and I dynamically set a “is_male” or “is_female” dynamic property.

class Person(db.Expando):
    name = db.StringProperty()

p1 = Person(name="Steve")
p1.is_male = True
p1.put()
p2 = Person(name="Jane")
p2.is_female = True
p2.put()

males = db.GqlQuery("SELECT * FROM Person WHERE is_male = :1", True)

Now lets say that we gather millions of records and we want to do a query, which one of the two methods above would be faster in production Google App Engine running Python 2.7?

Answer 1

There is absolutely no difference – models look the same in the datastore regardless of whether the property was ‘dynamic’ or not. The only difference is that a standard property class with no value set will insert a field with value None in the datastore, which takes some extra space, but allows you to query for users with that value not set.