Data Model

Optimizely has moved from NHibernate to Entity Framework in the 4.2 release of {user["product-name"]}. This topic covers a variety of topics to help developers acclimate to the new data layer.

The data layer supporting {user["product-name"]} has undergone a major shift from 4.1 to 4.2. There are two major topics to discuss when addressing this:

Whereas the previous versions of {user["product-name"]} had business logic encapsulated in methods and computed properties of data layer objects, 4.2 seeks to host all of that logic in the service layer, leaving the entities as simple as possible. This eliminates complexity in the data layer and allows for better control and extensibility throughout the platform. We encourage this paradigm be followed as the platform is extended.

Entity Framework is the default ORM for {user["product-name"]}. Although a great deal of the complexity of this switch should be abstracted from the application developer, there are a few differences in the development flows that will need to be considered. The bulk of this document will be focused on those.

The Insite.Data namespace is the root namespace for entity related components in the {user["product-name"]} framework. Models have moved from Insite.Model to Insite.Data.Entities, and many other data related components have made similar moves to this namespace. Insite.Data is the main dependency for all data layer related logic.

If you are familiar with previous versions of {user["product-name"]}, then most of your data use cases will not need any altering. The recommended way to access data is still to obtain a UnitOfWork object, get the appropriate Repository<T> object, and use that repository to access the entities.

To get an object by its primary key, simply call the Get method on the corresponding repository.

To retrieve a list of entities, call the GetTable method, which returns an IQueryable<T> that can be filtered (where), sorted (orderby) or mapped (select).

Also, if you would rather write your own optimized sql for retrieving the entities, you can do so and call GetList.

By default, lazy loading is enabled and can be used to retrieve entities and collections of entities related to your entity in the database. A separate sql call will be made once the collection is accessed in the code to retrieve these entities. In the example below, there are two sql calls made, one for the Get and one when Job.CustomProperties is accessed.

Although lazy loading is great for a number of use cases, it is sometimes advantageous to load collections ahead of time if you know that you are planning to access them. This optimizes the process by making only one SQL call.

In our last example, we always access the CustomProperties collection, so instead of lazy loading that collection, we can eagerly load it via the Include method provided by Insite.Data.Extensions.

Creating an entity consists of constructing a new instance of the desired entity type, and calling the Insert method on the corresponding repository.

Updating an attached entity requires an update to the instance, followed by a call to the Save method on the UnitOfWork instance.

You can also perform a Save asynchronously by calling SaveAsync on the UnitOfWork instance.

Deleting a job requires you invoke the Delete method on the Repository, followed by Save or SaveAsync on the UnitOfWork.

The officially supported method of changing the data model is to add entities to the model. Removing {user["product-name"]} tables or extending existing tables is not supported.

To add to the data model, you should first create a POCO (Plain Old CLR Object) that extends from Insite.Data.Entities.EntityBase. For this example we have created the Job class.

In the previous image, you will notice that a Table attribute/annotation was added to this class. This is superfluous as convention assumes the name of the class. A list of common annotations for EF can be found here

https://msdn.microsoft.com/en-us/data/jj591583.aspxhttps://msdn.microsoft.com/en-us/data/jj591583.aspx

Once your class is created, you can begin to add properties. Properties can be of varying types, including other entities, but must have "virtual" as part of their signatures. Keeping in mind the principle of a "thin" data layer, it is best practice to leave out computed properties or methods. Our example adds the following:

Once you have created your entity class, you must create a mapping class. This class defines any custom relationships that are outside of standard entity framework conventions. Since our example follows convention, we don't require any logic. Still, we need to create the class, and it must extend from Insite.Data.Providers.EntityFramework.EntityMappings.EntityBaseTypeConfiguration<T> in order for your entity to be bootstrapped to the context correctly.

Primarily, you will only need to add custom mapping logic to your mapping class if there is a many-to-many relationship between your entity and another. If this is the case, then you can follow standard entity framework guidance on creating this mapping.

Migrations are not currently supported as of {user["product-name"]} 4.2. Any changes made to the database schema need to be done via SQL files embedded in your custom dll's.

Creating a script consists of adding a new sql file to your Visual Studio Project. The file should have the format of <year>.<month>.<day>.<increment>.<description>.sql. These can be added anywhere in the project's folder structure.

Including a script into the build consists of setting the Build Action property of the file to Embedded Resource and the Copy to Output Directory property to Do not copy.

Sql scripts that are embedded resources are run automatically during the bootstrap operation. If the sql script executes successfully, it is added to the DatabaseScript table in the database. This is done for auditing purposes, and to ensure that a script is not run twice. You can query this table to determine if your script has been run.

There is a known issue using Linq with Entity Framework in conjunction with ICollection<T>.Contains() from within a Where clause. An example of this would be as follows:

Using this approach can severely degrade the performance of your application. To alleviate this issue, {user["product-name"]} has provided an extension method, WhereContains, which gives the same result in a much more performant manner. The same method as above should be refactored to the following:

Another optimization that can be made is to work with a set of data that does not have change tracking enabled. This improves read performance and can also improve the performance for batch operations.

Entity Framework accomplishes this via the AsNoTracking extension method, found here.

To limit dependencies on Entity Framework, {user["product-name"]} exposes the same functionality on all of their repository instances. To get a collection without tracking, invoke the GetTableAsNoTracking method on the Repository. An example would look like the following:

Typical usage would be scenarios where no updates to the entity are expected, or optimized sections where ObjectState and attachment to the context will be actively managed.