Embedded document vs reference in mongoose design

问题:

Let's say I am building a discussion forum using Node.js, and mongoose. A user can have multiple forums, and a forum can have multiple comments. A user can also invite other users to join a forum too. Thus, my question is about the model design either using reference or embedded document !

If I go with embedded document, It would look like:

var Comment = new Schema({ ... });

var Forum = new Schema({
    title: {type: String},
    content: {type: String},
    comments: [Comment],
    attendees: [User]
});

var User = new Schema({
    name: {type: String},
    email: {type: String},
    forums: [Forum]
});

var Account = mongoose.model('Account', User);

Using the above design, I struggled with: when a user adds a comment to a forum, and that forum is in my forums, I don't think I would be able to get update of a new comment in my forum list. Do I ? Do you know how to get the embedded document to work in this case?

Thus, I was thinking of using reference in mongoose. In this case, I will have two collections: Account, and Forum. Adding a new comment to a forum is not a problem in this case. Am I right?

Would reference be better than embedded document for this app?

Thanks in advance,

回答1:

It depends mostly on how you're gonna query and update your data. Consistency and document size is also important in this case. Here's a good summary on when referencing or embedding documents:

Embedding:

• Small subdocuments
• Data that does not change regularly
• Eventual consistency is acceptable
• Document that grow by a small amout
• Data that you will often need to perform a second query to fetch
• Fast reads

Referencing:

• Large subdocuments
• Volatile data
• Immediate consistency is necessary
• Document that grow a large amount
• Data that you will often exclude from results
• Fast writes

This is an exctract from a book on mongo I read. These are just general rules but from my experience, using them makes it very clear wether to reference or embed most of the times.

I would rather reference Forum in this case. But please consider all your requirements. For example if you reference Forum from User and you need to query all User of a particular Forum the query might be slow in this case. If I were you I would compose a list of everything I need and then using general rules would find a balance between pros and cons of embeding and referencing.

Hope it helps!

回答2:

Personally I like to do references in situations like yours. In this way I can get a comment from a user, a user from a forum, a forum from a comment, a forum from a user, etc. without worrying about doing complicated embedded document queries. I don't even bother storing embedded reference documents. If there is a one to many relationship between a forum and comments then I would store a forum reference on the comment and no comment reference on the forum because when you add/remove comments from the comments collection you then also have to go remove the embedded reference document from the comments collection on the forum.

I can query for a forum from a comment using the forum reference and I can get all comments for a forum by querying the comments collection for that forum reference (which is just an ID number until mongoose populates it behind the scenes for you).