The MVC Music Store release came as a fortunate surprise, I wanted to write a sample application for Raven, and having someone else write all the hard bits (the UI :-)) for me and leave to just replace the data access is just about ideal. The overall goal is to touch absolutely no JS/HTML code in the app, replacing just the controllers. It would be interesting to see if I’ll manage to do that.

The database model for the application is:

You can see some interesting things in the schema immediately:

• The Cart table should probably be called CartLineItem, because it stores a row per item in the cart.
• The CartId is not an foreign key, but a reference to the user name or the session id.
• The Artist table only hold the artist name, there is no other data.

With this information, I think that the following data model would be appropriate:

Albums

• The album document contains a reference to the Genre as well as the Genre’s name. This allows us to display the album without having to refer to the Genre document.
• The album document contains the artist name and id, for much the same reason.
• There isn’t actually a set of Artist documents in the database. We don’t keep any information about an artist, except its name, so I didn’t see a reason to introduce it at this time.

Genre

The genre document is an exact duplicate of the Genre table, no surprises here.

Cart

The cart is a pretty common document format, we have a single document and it contains an array of elements where the relational model contained a set of rows. Note that the UserIdentifier is where we store the user id or the session id for the cart.

Order

Order is another pretty standard document format. We aggregate all the order information to a single document, and we aggregate related information (Address) to a specific node.

Artist

There is no artist document.

Why does Genre deserve a standalone document while Artist gets subsumed into Album?

For the simple reason that the application does something with Genres (if only just displaying their description) where the only thing that is done with Artists is displaying their name in the context of an album. At this point, I consider Artist to be wholly owned by album, so there is no reason to keep a separate document for it. The only reason that album’s artist have and artist id is that I am assuming (based on the data) that the source for artist is an external system that does something a bit more meaningful with it than keep the artist name around.

This is just one half of the work we need to do, we have defined the data model, now we need to consider how this is used in a broader context, and add the query model using indexes. That will be covered in a future post.