Toolbelt.EntityFrameworkCore.IndexAttribute.Attribute 5.2.0

IndexColumnAttribute for Entity Framework Core

What's this?

This library brings the [Index] attribute of EF 6 back to EF Core, as the [IndexColumn] attribute. It also includes extension methods that build the indexes in your model from these attributes.

The automated tests check that it works on SQL Server, PostgreSQL, and SQLite.

Attention

Since ver.5.0, EF Core has its own official [Index] attribute.

However, I will keep maintaining and improving these libraries, because they still have some advantages.

• You can create indexes with data annotations even on an older version of EF Core.
• You can create indexes with included columns on SQL Server.
• You can create a clustered index. This means you can also create a non-clustered primary key.

Supported Versions

The IsDescending property needs EF Core 7.0 or later, and package v.5.1.0 or later.

How to use?

1. Add the Toolbelt.EntityFrameworkCore.IndexAttribute package to your project.

> dotnet add package Toolbelt.EntityFrameworkCore.IndexAttribute

If you want to use the IsClustered or Includes features on SQL Server, add the Toolbelt.EntityFrameworkCore.IndexAttribute.SqlServer package to your project instead.

> dotnet add package Toolbelt.EntityFrameworkCore.IndexAttribute.SqlServer

2. Annotate your model with the [IndexColumn] attribute. It lives in the Toolbelt.ComponentModel.DataAnnotations.Schema.V5 namespace.

Notice - The attribute name is [IndexColumn], not [Index]. EF Core 5.0 and later already use the name [Index].

using Toolbelt.ComponentModel.DataAnnotations.Schema.V5;

public class Person
{
    public int Id { get; set; }

    [IndexColumn] // <- Here!
    public string Name { get; set; }
}

3. [Important] Override the OnModelCreating() method of your DbContext class, and call the BuildIndexesFromAnnotations() extension method. It lives in the Toolbelt.ComponentModel.DataAnnotations namespace.

using Microsoft.EntityFrameworkCore;
using Toolbelt.ComponentModel.DataAnnotations;

public class MyDbContext : DbContext
{
  ...
  // Override "OnModelCreating", ...
  protected override void OnModelCreating(ModelBuilder modelBuilder)
  {
    base.OnModelCreating(modelBuilder);

    // .. and invoke "BuildIndexesFromAnnotations"!
    modelBuilder.BuildIndexesFromAnnotations();
  }
}

If you use the IsClustered or Includes features on SQL Server, call the BuildIndexesFromAnnotationsForSqlServer() extension method instead of BuildIndexesFromAnnotations().

  ...
  // Override "OnModelCreating", ...
  protected override void OnModelCreating(ModelBuilder modelBuilder)
  {
    base.OnModelCreating(modelBuilder);

    // Invoke "BuildIndexesFromAnnotationsForSqlServer"
    // instead of "BuildIndexesFromAnnotations".
    modelBuilder.BuildIndexesFromAnnotationsForSqlServer();
  }

That's all!

The BuildIndexesFromAnnotations() (or BuildIndexesFromAnnotationsForSqlServer()) extension method scans the DbContext with .NET reflection, finds the [IndexColumn] attributes, and builds the indexes in the model.

As a result, the database that EF Core creates has the indexes that you specified with the [IndexColumn] attributes.

Options of the [IndexColumn] attribute

Name and column order

The first argument is the index name, and the second one is the column order. If you put the same index name on two or more properties of an entity, they become one multi-column index, sorted by the column order. If you don't specify a name, EF Core names the index by its convention, and each property gets its own single-column index.

public class Person
{
    public int Id { get; set; }

    [IndexColumn("IX_FullName", 0)]
    public string LastName { get; set; }

    [IndexColumn("IX_FullName", 1)]
    public string FirstName { get; set; }
}

Specify the column order on every column of a multi-column index. Without it, the order of the columns is not defined.

The columns of one index can also be spread over a base class and its derived class, as long as the base class is not an entity type of its own.

public abstract class TenantEntityBase
{
    [IndexColumn("IX_Tenant_Code", 0)]
    public int TenantId { get; set; }
}

public class Product : TenantEntityBase
{
    public int Id { get; set; }

    [IndexColumn("IX_Tenant_Code", 1)]
    public string Code { get; set; }
}

Properties

In a multi-column index, you only need to set IsUnique, IsClustered, and Includes on one of its columns. They apply to the whole index. If two or more columns have Includes, all of the given columns are included. (Package v.5.1.0 and earlier read these properties only from one of the columns, so set them on every column if you use those versions.)

public class Ranking
{
    public int Id { get; set; }

    [IndexColumn(IsUnique = true)]
    public string Code { get; set; }

    [IndexColumn("IX_Category_Score", 0, IsUnique = true)]
    public string Category { get; set; }

    [IndexColumn("IX_Category_Score", 1, IsDescending = true)]
    public int Score { get; set; }

    // SQL Server only.
    [IndexColumn(Includes = new[] { nameof(Score) })]
    public DateTime CreatedAt { get; set; }
}

The [PrimaryKey] attribute

The [PrimaryKey] attribute (in the same namespace as [IndexColumn]) defines the primary key. On SQL Server, its IsClustered property lets you create a non-clustered primary key.

To define a composite primary key, give the same key name and the column order to every column, like a multi-column index. Without a name, each [PrimaryKey] attribute defines a primary key of one column, so only one of them is used. You only need to set IsClustered on one of the columns.

public class OrderLine
{
    [PrimaryKey("PK_OrderLine", 0)]
    public int OrderId { get; set; }

    [PrimaryKey("PK_OrderLine", 1)]
    public int LineNumber { get; set; }
}

Appendix A - Suppress "NotSupportedException"

BuildIndexesFromAnnotations() throws a "NotSupportedException" in these cases.

• An [IndexColumn] attribute has IsClustered = true.
• An [IndexColumn] attribute has a non-empty Includes.
• An [IndexColumn] attribute has IsDescending = true, and the app runs on EF Core 6.0 or earlier.

BuildIndexesFromAnnotationsForSqlServer() doesn't throw it for IsClustered and Includes, because it supports them. But it still throws it for IsDescending on EF Core 6.0 or earlier.

If you still want to build the indexes in these cases (for example, to share one model between different database products), you can suppress the exception with options. Then the library ignores the unsupported settings. For example, an index with IsDescending = true is created in ascending order.

  ...
  protected override void OnModelCreating(ModelBuilder modelBuilder)
  {
    base.OnModelCreating(modelBuilder);

    // Suppress "NotSupportedException" for each feature that you need.
    modelBuilder.BuildIndexesFromAnnotations(options => {
      options.SuppressNotSupportedException.IsClustered = true;
      options.SuppressNotSupportedException.Includes = true;
      options.SuppressNotSupportedException.IsDescending = true;
    });
  }
}

Appendix B - Notes on "IsClustered=true"

Suppose you simply annotate the model with "IsClustered=true", like this.

public class Employee {
  public int Id { get; set; }

  [IndexColumn(IsClustered = true)]
  public string EmployeeCode { get; set; }
}

Then you will get a "SqlException" like this.

SqlException :
Cannot create more than one clustered index on table '(table name)'.
Drop the existing clustered index '(index name)' before creating another.

In this case, annotate the primary key property with the [PrimaryKey(IsClustered = false)] attribute explicitly. This stops the primary key from becoming a clustered index automatically.

public class Employee {
  [PrimaryKey(IsClustered = false)] // <- Add this line!
  public int Id { get; set; }

  [IndexColumn(IsClustered = true)]
  public string EmployeeCode { get; set; }
}

Appendix C - If you want to use only "IndexAttribute" without any dependencies...

If you want to use only the "IndexColumnAttribute" class without any dependencies, use the Toolbelt.EntityFrameworkCore.IndexAttribute.Attribute NuGet package.

Appendix D - Upgrade an existing project

For more details on this topic, see this link.

Appendix E - If you run into a compile error CS0104...

For more details on this topic, see this link.

For More Details...

This library uses the same syntax as the [Index] attribute of EF 6.x.

See the EF 6.x documents below for more details.

• MSDN Document - Entity Framework Code First Data Annotations
• MSDN Document - IndexAttribute class

Limitations

• The [IndexColumn] attribute with IsClustered=true works only on entity types that are not owned types.
• The [PrimaryKey] attribute doesn't work on owned types. Configure their keys with the Fluent API in OnModelCreating() instead.
• A multi-column index can't span two entity types of an inheritance hierarchy. When the base class is also an entity type (TPH, TPT, or TPC mapping), put all the columns of an index in the same class.
• Don't give a name to [PrimaryKey] on a base class property, such as [PrimaryKey("PK_Items")], when you use the table-per-type (TPT) or table-per-concrete-type (TPC) mapping. EF Core gives that one name to the primary key of every table in the hierarchy, so SQL Server and PostgreSQL fail to create the tables because of the duplicate name. Use [PrimaryKey] without a name instead. Then EF Core names each primary key after its own table.

Release Notes

• Toolbelt.EntityFrameworkCore.IndexAttribute.Attribute
• Toolbelt.EntityFrameworkCore.IndexAttribute
• Toolbelt.EntityFrameworkCore.IndexAttribute.SqlServer

License

MIT License