Database specific features for Microsoft SQL Server

This section illustrates the database specific features for Microsoft SQL Server which are available to you through configuration, either through the .config file of your application or through code.

See the Application configuration through .config files section for the details about the configuration settings defined through .config files.

See the RuntimeConfiguration system section for details about configuring the settings through code.

SQL Server 2016+ Always Encrypted support

Enabling Always Encrypted is done through the connection string. Registration of key store providers is done by calling the static method SqlConnection.RegisterColumnEncryptionKeyStoreProviders once, e.g. at the start of your app.

For usage of key provider structures, .NET 4.6.1 or higher is required. If you're using .NET Core/NET Standard, please use the Microsoft.Data.SqlClient SqlClient package instead.

Info

Per-query parameter/field encryption directives are not currently supported due to limitations in how SqlClient's SqlCommand class deals with the parameter/field specifications. If Microsoft changes the way this is implemented in a future .NET version, we'll add support for it.

SQL Server 2016+ Temporal table support

On SQL Server 2016 or higher, Temporal tables are supported for select / fetch queries in Linq and QuerySpec. For SQL Server, System Time predicates are supported. Below is an example using Linq where the Employee entities are queried using a BETWEEN System Time predicate.

var q = from e in metaData.Employee
                                .ForSystemTime("BETWEEN {0} AND {1}", 
                                                fromDate, toDate)
        where e.EmployeeId == 1
        select e;
Important!

It's not recommended to use Temporal tables with entity inheritance, as FOR SYSTEM_TIME clauses are added to all tables targeted by an entity.

Target hint support

On all supported SQL Server versions, target hints are supported through Linq and QuerySpec. These hints are applied to elements in the FROM clause in a select query. Below is an example using Linq where a join is performed between two entities where one of them receives two hint directives.

var q = from c in metaData.Customer
                                .WithHint("NOLOCK")
                                .WithHint("FORCESEEK")
        join o in metaData.Order on c.CustomerId equals o.CustomerId 
        where o.EmployeeId > 4
        select c;

NEWSEQUENTIALID() support

When you're using unique_identifier types for primary keys on SQL Server 2005 or higher, you can benefit from a feature called NEWSEQUENTIALID(). This feature allows you to auto-generate new GUIDs for your primary keys which are sequential, so they are friendly for clustered indexes. To use this feature in LLBLGen Pro, you have to specify as the default for the primary key field in the table definition: NEWSEQUENTIALID().

Furthermore, you shouldn't set the PK field to a new GUID value when you're saving the entity. The DQE will then let the database insert the NEWSEQUENTIALID() produced value and it's automatically retrieved for you into the entity's PK.

LocalDB support

The LLBLGen Pro Runtime Framework supports LocalDB at runtime. To make sure the generated code works with the LocalDB database, the connection string has to be adjusted to match the requirements for LocalDB. Please consult the LocalDB documentation for this.

SQL Server 2012+ sequence support

The runtime supports SQL Server 2012+ sequence objects on pk and non-pk fields. There's no check for SQL Server 2012+ compatibility, however the queries will fail on SQL Server instances which are not capable of working with SQL Server 2012+ sequences. The generated INSERT queries look like (in the example, Field2 is sequenced)

SELECT @p2=NEXT VALUE FOR [schema].[sequence];
INSERT INTO table (Field1, Field2, .., Fieldn) VALUES (@p1, @p2, .. @pn); 

The schema name is taking schema name overwriting into account, so you can use the normal schema name overwriting (per call, equal to Oracle).

SQL Server compatibility mode

The SQL Server DQE is usable for a variety of different SQL Server versions, from 7 to the latest version, and from CE 3.1 to CE 4.0. The SQL Server DQE uses a compatibility mode to determine for which database to generate the SQL so the SQL matches the supported features for the SQL Server version used.

You can set the SQL Server DQE's compatibility mode in two different ways: using the application's .config file or use a code statement. For the application's .config file method, please see Generated code - Application configuration through .config files. For using the code method, please see for SelfServicing: Generated code - CommonDaoBase functionality and for Adapter: Generated code - DataAccessAdapter functionality. For the RuntimeConfiguration method, please see Runtime Configuration - SQL Server - Set default compatibility level.

The SqlServerCompatibilityLevel is defined as follows:

  • 0 for SqlServerCompatibilityLevel.SqlServer7
  • 1 for SqlServerCompatibilityLevel.SqlServer2000
  • 2 for SqlServerCompatibilityLevel.SqlServer2005 (also for 2008/2008 R2, which doesn't require a separate compatibility level)
  • 3 for SqlServerCompatibilityLevel.SqlServerCE3x (Sqlserver CE 3.0 and 3.1)
  • 4 for SqlServerCompatibilityLevel.SqlServerCE35
  • 5 for SqlServerCompatibilityLevel.SqlServerCE40
  • 6 for SqlServerCompatibilityLevel.SqlServer2012 (also for 2014/2016, which don't require a separate compatibility level)

By default the compatibility mode is set to SqlServerCompatibilityLevel.SqlServer2005.

The SQL differences per compatibility mode

Setting the compatibility level has the following effect on the generated SQL

  • the sequence retrieval logic to use by default:
    • @@IDENTITY with compatibility level SqlServer7, SqlServerCE3x, SqlServerCE35, SqlServerCE40
    • SCOPE_IDENTITY() with compatibility level SqlServer2000, SqlServer2005, SqlServer2012
  • the ability to use NEWSEQUENTIALID(), with compatibility level SqlServer2005, SqlServer2012
  • the SQL produced for a paging query
    • using a temp table approach with compatibility level SqlServer7, SqlServer2000
    • a CTE approach is used with compatibility level SqlServer2005
    • an OFFSET/FETCH using query with compatibility level SqlServer2012
    • a client-side paging approach (skip rows on datareader/close datareader when page is read) with compatibility level SqlServerCE3x, SqlServerCE35, SqlServerCE40
  • the TOP value in a SELECT query:
    • a parameter is used with compatibility level SqlServer2005, SqlServer2012
    • an in-line constant without brackets (e.g. SELECT TOP 5 ... ) with compatibility level SqlServer7, SqlServer2000
    • an in-line constant with brackets (e.g. SELECT TOP(5) ... ) with compatibility level SqlServerCE35, SqlServerCE40
    • a client-side limiter approach (read TOP rows from datareader, close datareader afterwards) with compatibility level SqlServerCE3x

ArithAbort support

When you're using indexed views in your database, and you're inserting data into tables which are used in these indexed views, you'll run into the problem that you have to set ARITHABORT ON before the particular insert statement is executed.

To signal that the SQL Server DQE has to emit the ARITHABORT statement prior to an insert statement, you can use the ArithAbort flag implemented on the CommonDaoBase class (SelfServicing) or DataAccessAdapter class (Adapter). Please see for SelfServicing: Generated code - CommonDaoBase functionality and for Adapter: Generated code - DataAccessAdapter functionality.

Spatial types / User Defined Types support

SQL Server 2005 or higher supports User Defined Types (UDTs) written in a CLR language like C# or VB.NET. The SQL Server driver can read these fields and if you're using UDTs in your tables, the fields which have a UDT as their type will be read by the driver and their UDT type is considered their valid type.

Entities mapped onto these tables (or views) have fields which .NET type is equal to the UDT of the target field in the table/view they're mapped on. The generated entity classes will have properties which refer to the UDT type as the type of the property, as the UDT is a normal CLR type.

In these situations you have to reference the assembly which contains the UDT in your generated Visual Studio project. (For Adapter, the database generic project). Usage of the field in .NET code is like any other code: you can set the field to an instance of the UDT type and normally save it and load it. Saving a UDT requires that the UDT is serializable to string, which is a requirement for SQL Server as well.

The same mechanism is used for loading and persisting Spatial Typed data.

SQL Server CE Desktop support

Important!

SQL Server CE isn't supported in the Netstandard 2.0+ build of the LLBLGen Pro runtime framework.

LLBLGen Pro supports SQL Server CE Desktop v3.1 or higher. SQL Server CE Desktop is the win32 runnable version of the same database known from the compact framework, SQL Server CE 3.0. SQL Server CE Desktop is SQL Server CE v3.1 or higher, but embeds roughly the same features as SQL Server CE 3.0 or higher for the compact framework: no stored procedures, a single schema and no meta-data retrieval.

It's recommended that you use the latest CE Desktop version, as it contains more features. To be able to target SQL Server CE Desktop, you first has to create a SQL Server project.

Stored procedures aren’t supported on CE Desktop, although they might be generated into the generated code. LLBLGen Pro uses the normal SQL Server DQE assembly for query production for CE Desktop. You also have to specify the compatibility level for the DQE, to signal it that it has to generate queries for SQL Server CE. For more information about this compatibility level, please see Generated code - Application configuration through .config files.

To be able to connect to a SqlServerCE desktop database, one has to adjust the connection string, as this connection string is the one used to connect to the SQL Server catalog from which the LLBLGen Pro project was created. It has to have the format shown by the following example:

<add key="Main.ConnectionString"  value="data source=c:\path\to\dbdatabasename.sdf;"/>

As SQL Server CE Desktop doesnt support multiple catalogs nor multiple schemas, these features arent available. Also System.Transactions transactions aren’t supported. All other LLBLGen Pro native features, like dependency injection, validation, authorization etc. are supported.

Linq is supported with SQL Server CE Desktop v3.5 or higher, however there are limitations in SQL Server CE Desktop which make it a bit of a struggle. For example the lack of scalar query support can lead to a lot of errors at runtime because a scalar query in a projection or WHERE clause isn't supported by SQL Server CE Desktop.

SqlServerCe provider registration

It's not necessary to reference System.Data.SqlServerCe.dll, however on the machine the application is ran which uses compatibility level 3 or 4, this dll has to be installed as documented in the SQL Server CE Desktop documentation about deployment: via the .msi shipped with SQL Server CE Desktop.

If you can’t run this .msi installer, be sure your application’s .config file contains the appropriate provider registration for the DbProviderFactory. (this information is installed in the machine.config file by the .msi installer of SQL Server CE Desktop). More details about this are available in the SQL Server CE Desktop documentation (the 'Books online' documents of SQL Server CE Desktop)

SQL Azure support

LLBLGen Pro supports SQL Azure out of the box, at runtime. To make your application run on SQL Azure you have to do the following:

  • Use Catalog name overwriting to overwrite your catalog's name to "". See Catalog Name Overwriting in application config files for details how to do this. This works for Selfservicing and adapter. As 'old' catalog name you specify the name of the catalog you're using in the project, e.g. Northwind. As 'new' catalog you specify the empty string: "".
  • Use only one catalog in your project. If you use multiple catalogs in your project, they'll be seen as one catalog on SQL Azure, as there's no catalog name allowed in SQL queries.
  • Make sure you utilize the LLBLGen Pro Runtime Framework feature Transient Error Recovery on code which can fail on Azure due to long running connections/transactions.