Alonso Robles

Following my recent post about getting started with RavenDB, I decided to work on a small shopping cart demo application based on the sample depot application that is written in part 2 of Agile Web Development with Rails when I started running into some trouble with the RavenDB auto generated document identifiers and the MVC routes. Like any other programmer, I quickly performed a Google search to see what others had done. The search led me to a few Stack Overflow well voted answers all leading me to a blog post on this topic. Problem solved right? Not exactly. Two solutions were proposed. But before getting into those, I want to describe the problem a little more.

The Problem Described

Take a look at the ASP.NET MVC 3 default route found in the Global.asax.cs file:

public static void RegisterRoutes(RouteCollection routes)
{
    routes.IgnoreRoute("{resource}.axd/{*pathInfo}");

    routes.MapRoute(
        "Default", // Route name
        "{controller}/{action}/{id}", // URL with parameters
        new { controller = "Home", action = "Index", id = UrlParameter.Optional } // Parameter defaults
    );
}

This default route will basically allow you access to a resource using the basic URL convention that may look something this: http://localhost/products/edit/1 . However, the default RavenDB auto generated document identifier for the product document ends up looking something like this: products/1. (More information about the RavenDB document identifiers one of the documentation pages. ) So the URL to access that product resource for editing would really have to look like this: http://localhost/products/edit/products/1. Well that doesn’t work for two reasons:

1. The default route will not be matched so the URL will not be routed the correct controller action.
2. The URL is not SEO friendly.

Typically, this can be solved by using a custom route that access this resource by a natural key where the URL would probably end up looking like this: http://localhost/products/product-name/edit. This makes the URL SEO friendly, easy to read, and easy to route. However, there are some situations where the document identifier makes sense and I will continue this post on the assumption that this is the case.

The blog post mentioned above presents two solutions:

• Solution 1 – Change the identity parts separator convention for the RavenDB document store.
• Solution 2 – Modify the ASP.NET default route

Issues with Solution 1

Let’s look at the first solution. The suggestion is to change the identity separator convention from a slash (/) to a dash (-). The convention is really easy to change during the initialization of the document store object.

documentStore = new DocumentStore { Url = "http://localhost:8080/" };
documentStore.Initialize();
documentStore.Conventions.IdentityPartsSeparator = "-";

This makes RavenDB generate document identifiers that now look like: products-1. Now we can generate a URL that will match the default route and may look like: http://localhost/products/edit/products-1. I still find some issues this this. Namely, the URL is still not really SEO friendly and not really human readable. I don’t like the idea of exposing an identifier that was forced on me by RavenDB since it essentially has coupled my URL to the RavenDB identifier convention. It just doesn’t feel right.

Issues with Solution 2

This solution suggests the change of the default route to something like:

public static void RegisterRoutes(RouteCollection routes)
{
    routes.IgnoreRoute("{resource}.axd/{*pathInfo}");

    routes.MapRoute(
        "Default", // Route name
        "{controller}/{action}/{*id}", // URL with parameters
        new { controller = "Home", action = "Index", id = UrlParameter.Optional } // Parameter defaults
    );
}

This will allow route to match a URL that looked like the first one: http://localhost/products/edit/products/1. I don’t like this either. While it does allow the default route to match and do its job. The URL is just ugly. It also clobbers the id making the route also match something like this: http://localhost/products/edit/products/more-junk/1. Obviously this is not SEO friendly or human readable. Again, it just doesn’t feel right.

Another Option

After some digging and learning about more about the document identity field, identity type converters, custom identifiers with the document key generator convention, I realized there was a pretty simple alternative to the two options above. But to see it and to understand the simplicity, you have to look at the source of the problem.

The reason that MVC routes are choking on the preconfigured RavenDB document identity convention is because our models included a string based identity property. I realized I had done this on purpose to ease the mapping of the RavenDB document identifier and my object identifier. Take a look at my product object definition:

public class Product
{
    public string Id { get; set; }
    public string Title { get; set; }
    public string Description { get; set; }
    public string ImageUrl { get; set; }
    public decimal Price { get; set; }
}

The shopping cart identifier is a string. Why would it be a string? It could have been an integer or a Guid or something else. It could have even been a natural key of some sort. Or even a compound key. It doesn’t have to be a string. Taking another look at the  IDocumentSession interface we can see we have four load methods available.

namespace Raven.Client
{
  public interface IDocumentSession : IDisposable
  {
    ISyncAdvancedSessionOperation Advanced { get; }
    void Delete<T>(T entity);
    T Load<T>(string id);
    T[] Load<T>(params string[] ids);
    T[] Load<T>(IEnumerable<string> ids);
    T Load<T>(ValueType id);
    IRavenQueryable<T> Query<T>(string indexName);
    IRavenQueryable<T> Query<T>();
    IRavenQueryable<T> Query<T, TIndexCreator>() where TIndexCreator : AbstractIndexCreationTask, new();
    ILoaderWithInclude<object> Include(string path);
    ILoaderWithInclude<T> Include<T>(Expression<Func<T, object>> path);
    void SaveChanges();
    void Store(object entity, Guid etag);
    void Store(object entity, Guid etag, string id);
    void Store(object entity);
    void Store(object entity, string id);
  }
}

After thinking about the difference between the method that accepts a string identifier and the one that accepts a value type identifier, I realized that all I had to do was change my identifier type from a string to a value type and I’m done. No fancy mapping or parsing the document identifier from RavenDB to the web browser and back. No changes to the default MVC route. No changes to the default identity separator convention. It just worked. My URLs are SEO friendly and more readable.

All I had to do was change the object identifier type.

public class Product
{
    public int Id { get; set; }
    public string Title { get; set; }
    public string Description { get; set; }
    public string ImageUrl { get; set; }
    public decimal Price { get; set; }
}

That was easy.

Of course this doesn’t solve everything. What if I really do need a string based identifier? This will probably require tweaking the behavior of the RavenDB client. I also don’t have an answer for compound keys and other natural keys which may not be a value type. Some of these may require more thought from both the RavenDB side and the MVC side of my project. But that is another problem to think about some other time.