Separate Binary Data Out Of An Entity For Effective Loading

I received an email this morning asking about loading binary data related to an entity, here is the excerpt:

“Hello, 

I was reading the following post on your blog and checking the code in Google Code. Entity Framework 4 POCO, Repository and Specification Pattern [Upgraded to EF 4.1]

I found it really interesting and realized that the performance issues where solved.

But how do you handle binary data?

Consider you have an entity with 2 Byte[] fields (varbinary or filestream in the database).

public class Product
{
    public Int32 Id { get; set; }

    public Boolean Available { get; set; }

    public String Name { get; set; }

    public Byte[] Image { get; set; }

    public Byte[] Brochure { get; set; }
}

When N records are loaded you will get N * 2 files being loaded even if you want to display only the names.

One solution to this problem is to apply a projection. For example:

context.Products.Where(x => x.Available).Select(x => new ProductView { Available = x.Available, Name = x.Name, Image = x.Image });

In this case I am leaving out the brochure but including the image … How do you address this with your code?”

—————————–

Here is my answer:

“In my experience regarding the binary data, I would create another entity, says, BinaryData which is somewhat like below (of course there will be a corresponding BinaryData table in the database):

public class BinaryData
{
    public Int32 Id { get; set; }
    public Byte[] Data { get; set; }
}

Then within an entity that has binary data, like your Product with Image or Brochure properties, I just add dependencies (or relationship) to the BinaryData entity:

public class Product
{
    // ... other properties
    public BinaryData Image { get; set; }
    public BinaryData Brochure { get; set; }
}

With lazy-loading enabled, you won’t worry about loading binary data (Image, Brochure) when loading Product, but it will be loaded when you access to it like: product.Image”

I myself think it is a good design practice when working with ORM framework, like Entity Framework or NHibernate.

What do you think? If there is a better solution for this, please drop a comment.