Schema-less design guidelines for Google App Engine Datastore and other NoSQL DBs
Coming from a relational database background, as I'm sure many others are, I'm looking for some solid guidelines for setting up / designing my datastore on Google App Engine. Are there any good rules of thumb people have for setting up these kinds of schema-less data stores? I understand some of the basics such as denormalizing since you can't do joins, but I was wondering what other recommendations people had.
The particular simple example I am working with concerns storing searches and their results. For example I have the following two models defined in my Google App Engine app using Python:
class Search(db.Model):
who = db.StringProperty()
what = db.StringProperty()
where = db.StringProperty()
createDate = db.DateTimeProperty(auto_now_add=True)
class SearchResult(db.Model):
title = db.StringProperty()
content = db.StringProperty()
who = db.StringProperty()
what = db.StringProperty()
where = db.StringProperty()
createDate = db.DateTimeProperty(auto_now_add=True)
I'm duplicating a bunch of 开发者_如何学Goproperties between the models for the sake of denormalization since I can't join Search
and SearchResult
together. Does this make sense? Or should I store a search ID in the SearchResult
model and effectively "join" the two models in code when I retrieve them from the datastore? Please keep in mind that this is a simple example. Both models will have a lot more properties and the way I'm approaching this right now, I would put any property I put in the Search model in the SearchResult
model as well.
Don't duplicate the properties if they'll always be the same between the SearchResult
and a Search
. If a SearchResult
should have a reference to a Search
, keep a ReferenceProperty
pointing to the Search. This basically stores the related Search
's Key
in the model.
class SearchResult(db.Model):
search = db.ReferenceProperty(Search, required=True)
# other stuff...
I also highly recommend you watch some of the App Engine videos from last year's Google I/O (and from 2008), in particular this one by Brett Slatkin, and this one by Ryan Barrett. They're all pretty helpful videos if you have the time, but I found those two in particular to be really great.
精彩评论