Optimizely has moved from NHibernate to Entity Framework in the 4.2 release of <<product-name>>. This topic covers a variety of topics to help developers acclimate to the new data layer.
## Work with models/ORM (EF)
The data layer supporting <<product-name>> has undergone a major shift from 4.1 to 4.2. There are two major topics to discuss when addressing this:
Changes to the technology stack, which swaps out NHibernate for Entity Framework.
Changes in coding standards to encourage a "thin" data layer, or a data layer whose entities have little to no business logic.
### Skinny data model
Whereas the previous versions of <<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
Entity Framework is the default ORM for <<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 <<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.
## Basic usage
### Unit of work and repositories
If you are familiar with previous versions of <<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.
### Retrieve Entities
#### Get an entity by Id
To get an object by its primary key, simply call the **Get** method on the corresponding repository.
#### Get a list of entities
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**.
This will return the values immediately. This method should be used with care and only when necessary.
### Retrieve related data
#### Lazy loading
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.
#### Eager loading
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 **Expand** method provided by **Insite.Data.Extensions**.
You must return an **IQueryable** to use this method.
### Update entities
#### Create an entity
Creating an entity consists of constructing a new instance of the desired entity type, and calling the Insert method on the corresponding repository.
To actually save the entity to the database, you will also need to call either the **Save** or **SaveAsync** method on the **UnitOfWork**. That will be covered in the next SubTopic.
#### Update an entity
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.
You do not have to call **Save** or **SaveAsync** immediately after updating the entity
#### Delete an entity
Deleting a job requires you invoke the **Delete** method on the Repository, followed by **Save** or **SaveAsync** on the **UnitOfWork**.
You do not have to call **Save** or **SaveAsync** immediately after updating the entity
## Extend the data model
### Alter the data model
#### Best practices
The officially supported method of changing the data model is to add entities to the model. Removing <<product-name>> tables or extending existing tables is not supported.
### Create an entity class
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.
#### Convention and annotations
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
<a href="https://msdn.microsoft.com/en-us/data/jj591583.aspx" class="Hyperlink">https\://msdn.microsoft.com/en-us/data/jj591583.aspx</a>
### Add properties
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:
### Map the entity
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.
#### Custom mappings
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.
### Alter the database schema
Migrations are not currently supported as of <<product-name>> 4.2. Any changes made to the database schema need to be done via SQL files embedded in your custom dll's.
#### Create a script
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.
#### Include a script
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**.
#### Run a Sql script
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.
## Performance considerations
### Retrieve records
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, <<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 doesn't 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
<a href="https://msdn.microsoft.com/en-us/library/gg679352(v=vs.103).aspx" class="Hyperlink">https\://msdn.microsoft.com/en-us/library/gg679352(v=vs.103).aspx</a>
To limit dependencies on Entity Framework, <<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.