tldr;
In EF Core 2.0, you can now rip out your per-Entity Fluent API table customizations into separate classes instead of having them all in the OnModelCreating method of your DbContext. You do this by inheriting from IEntityTypeConfiguration<T>.
Before In EF Core 1.x, note lines 13-33:
After in EF Core 2.0+, note lines 13, 14, and the 2 extra files:
The latter scales better to a big application with lots of entities.
The Problem
Let’s say you want to make a customization to one of your EF Core 2.0 properties and how it maps to your underlying database. Some customization scenarios include changing the default table name, changing the max length of the column, etc.
There are two ways of doing this. You could either choose the Data Annotations way or using the Fluent API. I personally like the Fluent API, so that my models aren’t littered with attributes. It also decouples my models from Entity Framework, in case I choose to use Dapper or another ORM down the road.
Prior to EF Core 2.0, to use the Fluent API you would have to “inline” any customizations in your OnModelCreating method. This works ok for small projects, but as you add tons of tables to your DbContext, this becomes a little hard to manage.
This would look something like this:
It looks ugly just for two tables… imagine 10, 20, or even 100! …Although maybe you should have multiple DbContexts at that point, but that’s beside the point.
This obviously doesn’t scale great. You could rip out each configuration into your own custom classes or static methods, but now there is a built-in solution in EF Core 2.0+.
The Solution
With EF Core 2.0 and above, you can now implement the interface IEntityTypeConfiguration<T> where T is the Model you’re configuring. After creating that class with the necessary customizations, you need to wire that up to your DbContext’s OnModelCreating method by calling builder.ApplyConfiguration(new ConfigClass). It’s important to make sure that your customizations come after the base.OnModelCreating(builder) call.
It looks something like this, note lines 13, 14, and the two extra files:
That looks much better and scales much nicer.
If you use the “inline” way today, you should be able to convert to the IEntityTypeConfiguration way without impacting the rest of your app or changing any schema. The API is the exact same in both formats (i.e. it’s HasKey in both spots to change the Primary Key).
Reverse Engineering from Existing Database
While you could handle your customization in EF Core 1 via custom static methods or your own hand rolled classes for maintainability purposes, now we have a consistent way of doing this across all EF Core Projects. That enables awesome things.
Such as the EF team is starting to work on adding this as an option when you Reverse Engineer classes from an existing database and have it auto-generate your customizations for you into these separate IEntityTypeConfiguration<T> classes. It’s currently slated for the EF Core 2.1 release that is likely to be released late this year or early next year. Obviously no guarantees that EF Core 2.1 lands at that time or that it includes this feature. However, it’s awesome that this is coming down the pipe.
This is also possible in EF 6
One final thing I’d like to mention is that you could do this same thing in EF 6. You just have to inherit from the EntityTypeConfiguration<T> abstract class (not an interface like EF Core), and then put your configuration in the constructor. There are many examples out there on how to do that and is outside of the scope of this post, but just thought I’d pass along in case you like this pattern and are using EF 6 today.
Hope this helps!