GAE datastore cache keys vs filter

Question

Lets say you have an Entity like this.

postid=db.StringProperty()
comment=db.StringProperty()

for storing comments on a certain post identified by post id. The comments can hit billions of records. Now if you want to get all comments belonging to a certain post you can do,

query=Comment.all()
query.filter('postid = ','id').

Or instead of doing that you can define post like

class Post(db.Model)
    commentids=db.StringListProperty()#store list of comment ids

This way you can directly get the comment by doing

comment=Comment.get_by_key_name('commentkey')

In the long run (when comments hit millions or even billions mark) which one is more efficient. In other words which one is more appropriate.

Answer 1

If you're planning to have billions of comments consider also using the newest NDB API, which, among other things, supports automatic caching.

Instead of filtering them by postid you should probably use a parent for your Comment entity. Here is an example (using DB, but it's very similar using NDB):

If you have model like this one:

class Post(db.Model):
  desc = db.StringProperty()

class Comment(db.Model):
  desc = db.TextProperty()

You can create posts and comments like:

post_db = Post(desc='Hello World')
post_db.put()

comment_db = Comment(parent=post_db, desc='Nice post')
comment_db.put()

And finally if you want to get all the comments from a particular post_db entity:

comment_dbs = Comment.all().ancestor(post_db)