modelling in google-datastore: use lists or separa

Question

I'm stuck trying to implement a data model for an app to be deployed on AppEngine.

Here's an example of the model:

EntityName
  id:  1
  author:       'me'
  description:  'whatever'
  datetime:     2011 etc.
  events: [
    {location: [1,2], phase: ['A', 'B']},
    {location: [3,4], phase: ['C', 'B']},
    ... more events ...
  ]

Explanation:

• EntityName has a few attributes, and many events
• each event has a list of numbers (its location) and a list of strings (phase)

How can this model be implemented in AppEngine? The data only needs to be searchable by id, author, datetime, and description.

---

What I've tried to figure out how to do (still going through the docs) (sticking point in italics):

1. one table, EntityName, with events as a ListProperty
   this would require nested lists in one table ... not sure if that's possible
2. two tables, EntityName and Event
   needs a join, which I understand isn't directly possible

Answer 1

Here's an alternative that only uses one model (i.e., Nick's idea):

import webapp2
from google.appengine.ext import db

class EntityName(db.Model):
    author = db.StringProperty()
    # ...
    event_locations = db.StringListProperty(indexed=False)
    event_phases    = db.StringListProperty(indexed=False)

class MainHandler(webapp2.RequestHandler):
    def get(self):
        e = EntityName(author='me',
            event_locations=['1,2', '3,4'], event_phases=['A,B', 'C,D'] )
        e.put()

        q = EntityName.all().filter('author =', 'me')
        text = ''
        for en in q:
            for locations, phases in zip(en.event_locations, en.event_phases):
                text += "location=%s, phase=%s<br>" % (locations, phases)
        self.response.out.write(text)

app = webapp2.WSGIApplication([('/', MainHandler)], debug=True)    

Sample output:

location=1,2, phase=A,B
location=3,4, phase=C,D

Since order is generally preserved in List and StringList properties, your locations and phases for an event can be matched up -- i.e., they have the same index in the two lists.

With one model you have to do a little more work splitting out each entry in the list, but you save on datastore reads and writes, since you only have one entity.

Answer 2

These aren't tables, exactly, but I think this should work for you

class EntityName(db.Model):
    author = db.StringProperty(...)

class Event(db.Model):
    entity_name = db.ReferenceProperty(EntityName, collection_name='events')
    location = db.StringListProperty()
    phase = db.StringListProperty()

Then...

e = EntityName(author='me', ...)
e.put()
event = Event(entity_name=e, location=...)
event.put()

for ev in e.events:
    print ev.location
    print ev.phase

See the documentation for details. You don't need to "do a join" because this isn't a SQL database. For your convenience, collection_name will create an iterator on the referenced side of a one-to-many relationship.