Breaking changes in EF Core 8 (EF8)
This page documents API and behavior changes that have the potential to break existing applications updating from EF Core 7 to EF Core 8. Make sure to review earlier breaking changes if updating from an earlier version of EF Core:
Target Framework
EF Core 8 targets .NET 8. Applications targeting older .NET, .NET Core, and .NET Framework versions will need to update to target .NET 8.
Summary
High-impact changes
Contains
in LINQ queries may stop working on older SQL Server versions
Old behavior
Previously, when the Contains
operator was used in LINQ queries with a parameterized value list, EF generated SQL that was inefficient but worked on all SQL Server versions.
New behavior
Starting with EF Core 8.0, EF now generates SQL that is more efficient, but is unsupported on SQL Server 2014 and below.
Note that newer SQL Server versions may be configured with an older compatibility level, also making them incompatible with the new SQL. This can also occur with an Azure SQL database which was migrated from a previous on-premises SQL Server instance, carrying over the old compatibility level.
Why
The previous SQL generated by EF Core for Contains
inserted the parameterized values as constants in the SQL. For example, the following LINQ query:
var names = new[] { "Blog1", "Blog2" };
var blogs = await context.Blogs
.Where(b => names.Contains(b.Name))
.ToArrayAsync();
... would be translated to the following SQL:
SELECT [b].[Id], [b].[Name]
FROM [Blogs] AS [b]
WHERE [b].[Name] IN (N'Blog1', N'Blog2')
Such insertion of constant values into the SQL creates many performance problems, defeating query plan caching and causing unneeded evictions of other queries. The new EF Core 8.0 translation uses the SQL Server OPENJSON
function to instead transfer the values as a JSON array. This solves the performance issues inherent in the previous technique; however, the OPENJSON
function is unavailable in SQL Server 2014 and below.
For more information about this change, see this blog post.
Mitigations
If your database is SQL Server 2016 (13.x) or newer, or if you're using Azure SQL, check the configured compatibility level of your database via the following command:
SELECT name, compatibility_level FROM sys.databases;
If the compatibility level is below 130 (SQL Server 2016), consider modifying it to a newer value (documentation).
Otherwise, if your database version really is older than SQL Server 2016, or is set to an old compatibility level which you cannot change for some reason, configure EF Core to revert to the older, less efficient SQL as follows:
protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)
=> optionsBuilder
.UseSqlServer(@"<CONNECTION STRING>", o => o.UseCompatibilityLevel(120));
Enums in JSON are stored as ints instead of strings by default
Old behavior
In EF7, enums mapped to JSON are, by default, stored as string values in the JSON document.
New behavior
Starting with EF Core 8.0, EF now, by default, maps enums to integer values in the JSON document.
Why
EF has always, by default, mapped enums to a numeric column in relational databases. Since EF supports queries where values from JSON interact with values from columns and parameters, it is important that the values in JSON match the values in the non-JSON column.
Mitigations
To continue using strings, configure the enum property with a conversion. For example:
protected override void OnModelCreating(ModelBuilder modelBuilder)
{
modelBuilder.Entity<User>().Property(e => e.Status).HasConversion<string>();
}
Or, for all properties of the enum type::
protected override void ConfigureConventions(ModelConfigurationBuilder configurationBuilder)
{
configurationBuilder.Properties<StatusEnum>().HaveConversion<string>();
}
Medium-impact changes
SQL Server date
and time
now scaffold to .NET DateOnly
and TimeOnly
Old behavior
Previously, when scaffolding a SQL Server database with date
or time
columns, EF would generate entity properties with types DateTime and TimeSpan.
New behavior
Starting with EF Core 8.0, date
and time
are scaffolded as DateOnly and TimeOnly.
Why
DateOnly and TimeOnly were introduced in .NET 6.0, and are a perfect match for mapping the database date and time types. DateTime notably contains a time component that goes unused and can cause confusion when mapping it to date
, and TimeSpan represents a time interval - possibly including days - rather than a time of day at which an event occurs. Using the new types prevents bugs and confusion, and provides clarity of intent.
Mitigations
This change only affects users which regularly re-scaffold their database into an EF code model ("database-first" flow).
It is recommended to react to this change by modifying your code to use the newly scaffolded DateOnly and TimeOnly types. However, if that isn't possible, you can edit the scaffolding templates to revert to the previous mapping. To do this, set up the templates as described on this page. Then, edit the EntityType.t4
file, find where the entity properties get generated (search for property.ClrType
), and change the code to the following:
var clrType = property.GetColumnType() switch
{
"date" when property.ClrType == typeof(DateOnly) => typeof(DateTime),
"date" when property.ClrType == typeof(DateOnly?) => typeof(DateTime?),
"time" when property.ClrType == typeof(TimeOnly) => typeof(TimeSpan),
"time" when property.ClrType == typeof(TimeOnly?) => typeof(TimeSpan?),
_ => property.ClrType
};
usings.AddRange(code.GetRequiredUsings(clrType));
var needsNullable = Options.UseNullableReferenceTypes && property.IsNullable && !clrType.IsValueType;
var needsInitializer = Options.UseNullableReferenceTypes && !property.IsNullable && !clrType.IsValueType;
#>
public <#= code.Reference(clrType) #><#= needsNullable ? "?" : "" #> <#= property.Name #> { get; set; }<#= needsInitializer ? " = null!;" : "" #>
<#
Boolean columns with a database generated value are no longer scaffolded as nullable
Old behavior
Previously, non-nullable bool
columns with a database default constraint were scaffolded as nullable bool?
properties.
New behavior
Starting with EF Core 8.0, non-nullable bool
columns are always scaffolded as non-nullable properties.
Why
A bool
property will not have its value sent to the database if that value is false
, which is the CLR default. If the database has a default value of true
for the column, then even though the value of the property is false
, the value in the database ends up as true
. However, in EF8, the sentinel used to determine whether a property has a value can be changed. This is done automatically for bool
properties with a database generated value of true
, which means that it is no longer necessary to scaffold the properties as nullable.
Mitigations
This change only affects users which regularly re-scaffold their database into an EF code model ("database-first" flow).
It is recommended to react to this change by modifying your code to use the non-nullable bool property. However, if that isn't possible, you can edit the scaffolding templates to revert to the previous mapping. To do this, set up the templates as described on this page. Then, edit the EntityType.t4
file, find where the entity properties get generated (search for property.ClrType
), and change the code to the following:
#>
var propertyClrType = property.ClrType != typeof(bool)
|| (property.GetDefaultValueSql() == null && property.GetDefaultValue() != null)
? property.ClrType
: typeof(bool?);
#>
public <#= code.Reference(propertyClrType) #><#= needsNullable ? "?" : "" #> <#= property.Name #> { get; set; }<#= needsInitializer ? " = null!;" : "" #>
<#
<#
Low-impact changes
SQLite Math
methods now translate to SQL
Old Behavior
Previously only the Abs, Max, Min, and Round methods on Math
were translated to SQL. All other members would be evaluated on the client if they appeared in the final Select expression of a query.
New behavior
In EF Core 8.0, all Math
methods with corresponding SQLite math functions are translated to SQL.
These math functions have been enabled in the native SQLite library that we provide by default (through our dependency on the SQLitePCLRaw.bundle_e_sqlite3 NuGet package). They have also been enabled in the library provided by SQLitePCLRaw.bundle_e_sqlcipher. If you're using one of these libraries, your application should not be affected by this change.
There is a chance, however, that applications including the native SQLite library by other means may not enable the math functions. In these cases, the Math
methods will be translated to SQL and encounter no such function errors when executed.
Why
SQLite added built-in math functions in version 3.35.0. Even though they're disabled by default, they've become pervasive enough that we decided to provide default translations for them in our EF Core SQLite provider.
We also collaborated with Eric Sink on the SQLitePCLRaw project to enable math functions in all of the native SQLite libraries provided as part of that project.
Mitigations
The simplest way to fix breaks is, when possible, to enable the math function is the native SQLite library by specifying the SQLITE_ENABLE_MATH_FUNCTIONS compile-time option.
If you don't control compilation of the native library, you can also fix breaks by create the functions yourself at runtime using the Microsoft.Data.Sqlite APIs.
sqliteConnection
.CreateFunction<double, double, double>(
"pow",
Math.Pow,
isDeterministic: true);
Alternatively, you can force client-evaluation by splitting the Select expression into two parts separated by AsEnumerable
.
// Before
var query = dbContext.Cylinders
.Select(
c => new
{
Id = c.Id
// May throw "no such function: pow"
Volume = Math.PI * Math.Pow(c.Radius, 2) * c.Height
});
// After
var query = dbContext.Cylinders
// Select the properties you'll need from the database
.Select(
c => new
{
c.Id,
c.Radius,
c.Height
})
// Switch to client-eval
.AsEnumerable()
// Select the final results
.Select(
c => new
{
Id = c.Id,
Volume = Math.PI * Math.Pow(c.Radius, 2) * c.Height
});
ITypeBase replaces IEntityType in some APIs
Old behavior
Previously, all mapped structural types were entity types.
New behavior
With the introduction of complex types in EF8, some APIs that were previously use an IEntityType
now use ITypeBase
so that the APIs can be used with either entity or complex types. This includes:
IProperty.DeclaringEntityType
is now obsolete andIProperty.DeclaringType
should be used instead.IEntityTypeIgnoredConvention
is now obsolete andITypeIgnoredConvention
should be used instead.IValueGeneratorSelector.Select
now accepts anITypeBase
which may be, but does not have to be anIEntityType
.
Why
With the introduction of complex types in EF8, these APIs can be used with either IEntityType
or IComplexType
.
Mitigations
The old APIs are obsoleted, but will not be removed until EF10. Code should be updated to use the new APIs ASAP.
ValueConverter and ValueComparer expressions must use public APIs for the compiled model
Old behavior
Previously, ValueConverter
and ValueComparer
definitions were not included in the compiled model, and so could contain arbitrary code.
New behavior
EF now extracts the expressions from the ValueConverter
and ValueComparer
objects and includes these C# in the compiled model. This means that these expressions must only use public API.
Why
The EF team is gradually moving more constructs into the compiled model to support using EF Core with AOT in the future.
Mitigations
Make the APIs used by the comparer public. For example, consider this simple converter:
public class MyValueConverter : ValueConverter<string, byte[]>
{
public MyValueConverter()
: base(v => ConvertToBytes(v), v => ConvertToString(v))
{
}
private static string ConvertToString(byte[] bytes)
=> ""; // ... TODO: Conversion code
private static byte[] ConvertToBytes(string chars)
=> Array.Empty<byte>(); // ... TODO: Conversion code
}
To use this converter in a compiled model with EF8, the ConvertToString
and ConvertToBytes
methods must be made public. For example:
public class MyValueConverter : ValueConverter<string, byte[]>
{
public MyValueConverter()
: base(v => ConvertToBytes(v), v => ConvertToString(v))
{
}
public static string ConvertToString(byte[] bytes)
=> ""; // ... TODO: Conversion code
public static byte[] ConvertToBytes(string chars)
=> Array.Empty<byte>(); // ... TODO: Conversion code
}
ExcludeFromMigrations no longer excludes other tables in a TPC hierarchy
Old behavior
Previously, using ExcludeFromMigrations
on a table in a TPC hierarchy would also exclude other tables in the hierarchy.
New behavior
Starting with EF Core 8.0, ExcludeFromMigrations
does not impact other tables.
Why
The old behavior was a bug and prevented migrations from being used to manage hierarchies across projects.
Mitigations
Use ExcludeFromMigrations
explicitly on any other table that should be excluded.
Non-shadow integer keys are persisted to Cosmos documents
Old behavior
Previously, non-shadow integer properties that match the criteria to be a synthesized key property would not be persisted into the JSON document, but were instead re-synthesized on the way out.
New behavior
Starting with EF Core 8.0, these properties are now persisted.
Why
The old behavior was a bug and prevented properties that match the synthesized key criteria from being persisted to Cosmos.
Mitigations
Exclude the property from the model if its value should not be persisted.
Additionally, you can disable this behavior entirely by setting Microsoft.EntityFrameworkCore.Issue31664
AppContext switch to true
, see AppContext for library consumers for more details.
AppContext.SetSwitch("Microsoft.EntityFrameworkCore.Issue31664", isEnabled: true);
Relational model is generated in the compiled model
Old behavior
Previously, the relational model was computed at run-time even when using a compiled model.
New behavior
Starting with EF Core 8.0, the relational model is part of the generated compiled model. However, for particularly large models the generated file may fail to compile.
Why
This was done to further improve startup time.
Mitigations
Edit the generated *ModelBuilder.cs
file and remove the line AddRuntimeAnnotation("Relational:RelationalModel", CreateRelationalModel());
as well as the method CreateRelationalModel()
.
Scaffolding may generate different navigation names
Old behavior
Previously when scaffolding a DbContext
and entity types from an existing database, the navigation names for relationships were sometimes derived from a common prefix of multiple foreign key column names.
New behavior
Starting with EF Core 8.0, common prefixes of column names from a composite foreign key are no longer used to generate navigation names.
Why
This is an obscure naming rule which sometimes generates very poor names like, S
, Student_
, or even just _
. Without this rule, strange names are no longer generated, and the naming conventions for navigations are also made simpler, thereby making it easier to understand and predict which names will be generated.
Mitigations
The EF Core Power Tools have an option to keep generating navigations in the old way. Alternatively, the code generated can be fully customized using T4 templates. This can be used to example the foreign key properties of scaffolding relationships and use whatever rule is appropriate for your code to generate the navigation names you need.
Discriminators now have a max length
Old behavior
Previously, discriminator columns created for TPH inheritance mapping were configured as nvarchar(max)
on SQL Server/Azure SQL, or the equivalent unbounded string type on other databases.
New behavior
Starting with EF Core 8.0, discriminator columns are created with a max length that covers all the known discriminator values. EF will generate a migration to make this change. However, if the discriminator column is constrained in some way -- for example, as part of an index -- then the AlterColumn
created by Migrations may fail.
Why
nvarchar(max)
columns are inefficient and unnecessary when the lengths of all possible values are known.
Mitigations
The column size can be made explicitly unbounded:
modelBuilder.Entity<Foo>()
.Property<string>("Discriminator")
.HasMaxLength(-1);
SQL Server key values are compared case-insensitively
Old behavior
Previously, when tracking entities with string keys with the SQL Server/Azure SQL database providers, the key values were compared using the default .NET case-sensitive ordinal comparer.
New behavior
Starting with EF Core 8.0, SQL Server/Azure SQL string key values are compared using the default .NET case-insensitive ordinal comparer.
Why
By default, SQL Server uses case-insensitive comparisons when comparing foreign key values for matches to principal key values. This means when EF uses case-sensitive comparisons it may not connect a foreign key to a principal key when it should.
Mitigations
Case-sensitive comparisons can be used by setting a custom ValueComparer
. For example:
protected override void OnModelCreating(ModelBuilder modelBuilder)
{
var comparer = new ValueComparer<string>(
(l, r) => string.Equals(l, r, StringComparison.Ordinal),
v => v.GetHashCode(),
v => v);
modelBuilder.Entity<Blog>()
.Property(e => e.Id)
.Metadata.SetValueComparer(comparer);
modelBuilder.Entity<Post>(
b =>
{
b.Property(e => e.Id).Metadata.SetValueComparer(comparer);
b.Property(e => e.BlogId).Metadata.SetValueComparer(comparer);
});
}