Document Design: RaccoonBlog.Comments

[Tim Murphy]

I'm reading RaccoonBlog source code to educate myself in the use of RavenDB.

How does the design and updating of https://github.com/ayende/RaccoonBlog/blob/master/RaccoonBlog.Web/Models/PostComments.cs not lead to loss of comments. If I understand this document and RavenDB correctly then this is what I think could happen if two comments are submitted at the same time:

Comment A loads PostComments for post 1.

Comment B loads PostComments for post 1.

Comment A and B have the same copy of PostComments.

Comment A is added to PostComments and saved.

Comment B is added to PostComments and saved. This save does not know about Comment A and therefore Comment A is effectively removed from the database.

Have I got this completely wrong? 

In such a situation does RavenDB throw an exception?

Why not save comments to a "standard" RavenDB collection? 

What benefit does PostComments give over saving comments to a standard RavenDB collection?

[Itamar Syn-Hershko]

It uses Optimistic Concurrency to make the comment is added to the latest version of the PostComments object, and fetches that doc again if it gets an indication of a change

See  https://github.com/ayende/RaccoonBlog/blob/master/RaccoonBlog.Web/Infrastructure/Tasks/AddCommentTask.cs  and https://github.com/ayende/RaccoonBlog/blob/master/HibernatingRhinos.Loci.Common/Tasks/BackgroundTask.cs#L17

It is done in the background so not to block the request

[Oren Eini (Ayende Rahien)]

Tim,

We use optimistic concurrency and reties to avoid this issue.

And putting this in a dedicated document means that loading the post page is just loading two documents by their ids, which is very cheap.

[Tim Murphy]

Oren & Itamar,

Thanks the explanation. I now understand whats happening.

Tim