Walkthrough: Accessing a SQL Database by Using Type Providers (F#)

This walkthrough explains how to use the SqlDataConnection (LINQ to SQL) type provider that is available in F# 3.0 to generate types for data in a SQL database when you have a live connection to a database. If you do not have a live connection to a database, but you do have a LINQ to SQL schema file (DBML file), see Walkthrough: Generating F# Types from a DBML File (F#).

This walkthrough illustrates the following tasks. These tasks must be performed in this order for the walkthrough to succeed:

  • Preparing a test database

  • Creating the project

  • Setting up a type provider

  • Querying the data

  • Working with nullable fields

  • Calling a stored procedure

  • Updating the database

  • Executing Transact-SQL code

  • Using the full data context

  • Deleting data

  • Create a test database (as needed)

Preparing a Test Database

On a server that's running SQL Server, create a database for testing purposes. You can use the database create script at the bottom of this page in the section MyDatabase Create Script to do this.

To prepare a test database

  • To run the MyDatabase Create Script, open the View menu, and then choose SQL Server Object Explorer or choose the Ctrl+\, Ctrl+S keys. In SQL Server Object Explorer window, open the shortcut menu for the appropriate instance, choose New Query, copy the script at the bottom of this page, and then paste the script into the editor. To run the SQL script, choose the toolbar icon with the triangular symbol, or choose the Ctrl+Q keys. For more information about SQL Server Object Explorer, see Connected Database Development.

Creating the project

Next, you create an F# application project.

To create and set up the project

  1. Create a new F# Application project.

  2. Add references to .FSharp.Data.TypeProviders, as well as System.Data, and System.Data.Linq.

  3. Open the appropriate namespaces by adding the following lines of code to the top of your F# code file Program.fs.

    open System
    open System.Data
    open System.Data.Linq
    open Microsoft.FSharp.Data.TypeProviders
    open Microsoft.FSharp.Linq
  4. As with most F# programs, you can execute the code in this walkthrough as a compiled program, or you can run it interactively as a script. If you prefer to use scripts, open the shortcut menu for the project node, select Add New Item, add an F# script file, and add the code in each step to the script. You will need to add the following lines at the top of the file to load the assembly references.

    #r "System.Data.dll"
    #r "FSharp.Data.TypeProviders.dll"
    #r "System.Data.Linq.dll"

    You can then select each block of code as you add it and press Alt+Enter to run it in F# Interactive.

Setting up a type provider

In this step, you create a type provider for your database schema.

To set up the type provider from a direct database connection

  • There are two critical lines of code that you need in order to create the types that you can use to query a SQL database using the type provider. First, you instantiate the type provider. To do this, create what looks like a type abbreviation for a SqlDataConnection with a static generic parameter. SqlDataConnection is a SQL type provider, and should not be confused with SqlConnection type that is used in ADO.NET programming. If you have a database that you want to connect to, and have a connection string, use the following code to invoke the type provider. Substitute your own connection string for the example string given. For example, if your server is MYSERVER and the database instance is INSTANCE, the database name is MyDatabase, and you want to use Windows Authentication to access the database, then the connection string would be as given in the following example code.

    type dbSchema = SqlDataConnection<"Data Source=MYSERVER\INSTANCE;Initial Catalog=MyDatabase;Integrated Security=SSPI;">
    let db = dbSchema.GetDataContext()
    // Enable the logging of database activity to the console.
    db.DataContext.Log <- System.Console.Out

    Now you have a type, dbSchema, which is a parent type that contains all the generated types that represent database tables. You also have an object, db, which has as its members all the tables in the database. The table names are properties, and the type of these properties is generated by the F# compiler. The types themselves appear as nested types under dbSchema.ServiceTypes. Therefore, any data retrieved for rows of these tables is an instance of the appropriate type that was generated for that table. The name of the type is ServiceTypes.Table1.

    To familiarize yourself with how the F# language interprets queries into SQL queries, review the line that sets the Log property on the data context.

    To further explore the types created by the type provider, add the following code.

    let table1 = db.Table1

    Hover over table1 to see its type. Its type is System.Data.Linq.Table<dbSchema.ServiceTypes.Table1> and the generic argument implies that the type of each row is the generated type, dbSchema.ServiceTypes.Table1. The compiler creates a similar type for each table in the database.

Querying the data

In this step, you write a query using F# query expressions.

To query the data

  1. Now create a query for that table in the database. Add the following code.

    let query1 =
            query {
                for row in db.Table1 do
                select row
    query1 |> Seq.iter (fun row -> printfn "%s %d" row.Name row.TestData1)

    The appearance of the word query indicates that this is a query expression, a type of computation expression that generates a collection of results similar of a typical database query. If you hover over query, you will see that it is an instance of Linq.QueryBuilder Class (F#), a type that defines the query computation expression. If you hover over query1, you will see that it is an instance of IQueryable<T>. As the name suggests, IQueryable<T> represents data that may be queried, not the result of a query. A query is subject to lazy evaluation, which means that the database is only queried when the query is evaluated. The final line passes the query through Seq.iter. Queries are enumerable and may be iterated like sequences. For more information, see Query Expressions (F#).

  2. Now add a query operator to the query. There are a number of query operators available that you can use to construct more complex queries. This example also shows that you can eliminate the query variable and use a pipeline operator instead.

    query {
            for row in db.Table1 do
            where (row.TestData1 > 2)
            select row
    |> Seq.iter (fun row -> printfn "%d %s" row.TestData1 row.Name)
  3. Add a more complex query with a join of two tables.

    query {
            for row1 in db.Table1 do
            join row2 in db.Table2 on (row1.Id = row2.Id)
            select (row1, row2)
    |> Seq.iteri (fun index (row1, row2) ->
         if (index = 0) then printfn "Table1.Id TestData1 TestData2 Name Table2.Id TestData1 TestData2 Name"
         printfn "%d %d %f %s %d %d %f %s" row1.Id row1.TestData1 row1.TestData2 row1.Name
             row2.Id (row2.TestData1.GetValueOrDefault()) (row2.TestData2.GetValueOrDefault()) row2.Name)
  4. In real-world code, the parameters in your query are usually values or variables, not compile-time constants. Add the following code that wraps a query in a function that takes a parameter, and then calls that function with the value 10.

    let findData param =
        query {
            for row in db.Table1 do
            where (row.TestData1 = param)
            select row
    findData 10 |> Seq.iter (fun row -> printfn "Found row: %d %d %f %s" row.Id row.TestData1 row.TestData2 row.Name)

Working with nullable fields

In databases, fields often allow null values. In the .NET type system, you cannot use the ordinary numerical data types for data that allows nulls because those types do not have null as a possible value. Therefore, these values are represented by instances of Nullable<T> type. Instead of accessing the value of such fields directly with the name of the field, you need to add some extra steps. You can use the Value property to access the underlying value of a nullable type. The Value property throws an exception if the object is null rather than having a value. You can use the HasValue Boolean method to determine if a value exists, or use GetValueOrDefault to ensure that you have an actual value in all cases. If you use GetValueOrDefault and there is a null in the database, then it is replaced with a value such as an empty string for string types, 0 for integral types or 0.0 for floating point types.

When you need to perform equality tests or comparisons on nullable values in a where clause in a query, you can use the nullable operators found in the Linq.NullableOperators Module (F#). These are like the regular comparison operators =, >, <=, and so on, except that a question mark appears on the left or right of the operator where the nullable values are. For example, the operator >? is a greater-than operator with a nullable value on the right. The way these operators work is that if either side of the expression is null, the expression evaluates to false. In a where clause, this generally means that the rows that contain null fields are not selected and not returned in the query results.

To work with nullable fields

  • The following code shows working with nullable values; assume that TestData1 is an integer field that allows nulls.

    query {
            for row in db.Table2 do
            where (row.TestData1.HasValue && row.TestData1.Value > 2)
            select row
    |> Seq.iter (fun row -> printfn "%d %s" row.TestData1.Value row.Name)
    query {
            for row in db.Table2 do
            // Use a nullable operator ?>
            where (row.TestData1 ?> 2)
            select row
    |> Seq.iter (fun row -> printfn "%d %s" (row.TestData1.GetValueOrDefault()) row.Name)

Calling a stored procedure

Any stored procedures on the database can be called from F#. You must set the static parameter StoredProcedures to true in the type provider instantiation. The type provider SqlDataConnection contains several static methods that you can use to configure the types that are generated. For a complete description of these, see SqlDataConnection Type Provider (F#). A method on the data context type is generated for each stored procedure.

To call a stored procedure

  • If the stored procedures takes parameters that are nullable, you need to pass an appropriate Nullable<T> value. The return value of a stored procedure method that returns a scalar or a table is ISingleResult<T>, which contains properties that enable you to access the returned data. The type argument for ISingleResult<T> depends on the specific procedure and is also one of the types that the type provider generates. For a stored procedure named Procedure1, the type is Procedure1Result. The type Procedure1Result contains the names of the columns in a returned table, or, for a stored procedure that returns a scalar value, it represents the return value.

    The following code assumes that there is a procedure Procedure1 on the database that takes two nullable integers as parameters, runs a query that returns a column named TestData1, and returns an integer.

    type schema = SqlDataConnection<"Data Source=MYSERVER\INSTANCE;Initial Catalog=MyDatabase;Integrated Security=SSPI;",
                                    StoredProcedures = true>
    let testdb = schema.GetDataContext()
    let nullable value = new System.Nullable<_>(value)
    let callProcedure1 a b =
        let results = testdb.Procedure1(nullable a, nullable b)
        for result in results do
            printfn "%d" (result.TestData1.GetValueOrDefault())
        results.ReturnValue :?> int
    printfn "Return Value: %d" (callProcedure1 10 20)

Updating the database

The LINQ DataContext type contains methods that make it easier to perform transacted database updates in a fully typed fashion with the generated types.

To update the database

  1. In the following code, several rows are added to the database. If you are only adding a row, you can use InsertOnSubmit to specify the new row to add. If you are inserting multiple rows, you should put them into a collection and call InsertAllOnSubmit<TSubEntity>. When you call either of these methods, the database is not immediately changed. You must call SubmitChanges to actually commit the changes. By default, everything that you do before you call SubmitChanges is implicitly part of the same transaction.

    let newRecord = new dbSchema.ServiceTypes.Table1(Id = 100,
                                                     TestData1 = 35, 
                                                     TestData2 = 2.0,
                                                     Name = "Testing123")
    let newValues =
        [ for i in [1 .. 10] ->
              new dbSchema.ServiceTypes.Table3(Id = 700 + i,
                                               Name = "Testing" + i.ToString(),
                                               Data = i) ]
    // Insert the new data into the database.
        printfn "Successfully inserted new rows."
       | exn -> printfn "Exception:\n%s" exn.Message
  2. Now clean up the rows by calling a delete operation.

    // Now delete what was added.
        printfn "Successfully deleted all pending rows."
       | exn -> printfn "Exception:\n%s" exn.Message

Executing Transact-SQL code

You can also specify Transact-SQL directly by using the ExecuteCommand method on the DataContext class.

To execute custom SQL commands

  • The following code shows how to send SQL commands to insert a record into a table and also to delete a record from a table.

       db.DataContext.ExecuteCommand("INSERT INTO Table3 (Id, Name, Data) VALUES (102, 'Testing', 55)") |> ignore
       | exn -> printfn "Exception:\n%s" exn.Message
    try //AND Name = 'Testing' AND Data = 55
       db.DataContext.ExecuteCommand("DELETE FROM Table3 WHERE Id = 102 ") |> ignore
       | exn -> printfn "Exception:\n%s" exn.Message

Using the full data context

In the previous examples, the GetDataContext method was used to get what is called the simplified data context for the database schema. The simplified data context is easier to use when you are constructing queries because there are not as many members available. Therefore, when you browse the properties in IntelliSense, you can focus on the database structure, such as the tables and stored procedures. However, there is a limit to what you can do with the simplified data context. A full data context that provides the ability to perform other actions. is also available This is located in the ServiceTypes and has the name of the DataContext static parameter if you provided it. If you did not provide it, the name of the data context type is generated for you by SqlMetal.exe based on the other input. The full data context inherits from DataContext and exposes the members of its base class, including references to ADO.NET data types such as the Connection object, methods such as ExecuteCommand and ExecuteQuery that you can use to write queries in SQL, and also a means to work with transactions explicitly.

To use the full data context

  • The following code demonstrates getting a full data context object and using it to execute commands directly against the database. In this case, two commands are executed as part of the same transaction.

    let dbConnection = testdb.Connection
    let fullContext = new dbSchema.ServiceTypes.MyDatabase(dbConnection)
    let transaction = dbConnection.BeginTransaction()
    fullContext.Transaction <- transaction
        let result1 = fullContext.ExecuteCommand("INSERT INTO Table3 (Id, Name, Data) VALUES (102, 'A', 55)")
        printfn "ExecuteCommand Result: %d" result1
        let result2 = fullContext.ExecuteCommand("INSERT INTO Table3 (Id, Name, Data) VALUES (103, 'B', -2)")
        printfn "ExecuteCommand Result: %d" result2
        if (result1 <> 1 || result2 <> 1) then
            printfn "Rolled back creation of two new rows."
            printfn "Successfully committed two new rows."
        | exn -> transaction.Rollback()
                 printfn "Rolled back creation of two new rows due to exception:\n%s" exn.Message

Deleting data

This step shows you how to delete rows from a data table.

To delete rows from the database

  • Now, clean up any added rows by writing a function that deletes rows from a specified table, an instance of the Table<TEntity> class. Then write a query to find all the rows that you want to delete, and pipe the results of the query into the deleteRows function. This code takes advantage of the ability to provide partial application of function arguments.

    let deleteRowsFrom (table:Table<_>) rows =
    query {
        for rows in db.Table3 do
        where (rows.Id > 10)
        select rows
    |> deleteRowsFrom db.Table3
    printfn "Successfully deleted rows with Id greater than 10 in Table3."

Creating a test database

This section shows you how to set up the test database to use in this walkthrough.

Note that if you alter the database in some way, you will have to reset the type provider. To reset the type provider, rebuild or clean the project that contains the type provider.

To create the test database

  1. In Server Explorer, open the shortcut menu for the Data Connections node, and choose Add Connection. The Add Connection dialog box appears.

  2. In the Server name box, specify the name of an instance of SQL Server that you have administrative access to, or if you do not have access to a server, specify (localdb\v11.0). SQL Express LocalDB provides a lightweight database server for development and testing on your machine. A new node is created in Server Explorer under Data Connections. For more information about LocalDB, see Walkthrough: Creating a SQL Server Express LocalDB Database.

  3. Open the shortcut menu for the new connection node, and select New Query.

  4. Copy the following SQL script, paste it into the query editor, and then choose the Execute button on the toolbar or choose the Ctrl+Shift+E keys.

    USE [master];
    IF EXISTS (SELECT * FROM sys.databases WHERE name = 'MyDatabase')
                    DROP DATABASE MyDatabase;
    -- Create the MyDatabase database.
    -- Specify a simple recovery model 
    -- to keep the log growth to a minimum.
    ALTER DATABASE MyDatabase 
                    SET RECOVERY SIMPLE;
    USE MyDatabase;
    -- Create the Table1 table.
    CREATE TABLE [dbo].[Table1] (
        [Id]        INT        NOT NULL,
        [TestData1] INT        NOT NULL,
        [TestData2] FLOAT (53) NOT NULL,
        [Name]      NTEXT      NOT NULL,
    --Create Table2.
    CREATE TABLE [dbo].[Table2] (
        [Id]        INT        NOT NULL,
        [TestData1] INT        NULL,
        [TestData2] FLOAT (53) NULL,
        [Name]      NTEXT      NOT NULL,
    --     Create Table3.
    CREATE TABLE [dbo].[Table3] (
        [Id]   INT           NOT NULL,
        [Name] NVARCHAR (50) NOT NULL,
        [Data] INT           NOT NULL,
    CREATE PROCEDURE [dbo].[Procedure1]
           @param1 int = 0,
           @param2 int
           SELECT TestData1 FROM Table1
    RETURN 0
    -- Insert data into the Table1 table.
    USE MyDatabase
    INSERT INTO Table1 (Id, TestData1, TestData2, Name)
    VALUES(1, 10, 5.5, 'Testing1');
    INSERT INTO Table1 (Id, TestData1, TestData2, Name)
    VALUES(2, 20, -1.2, 'Testing2');
    --Insert data into the Table2 table.
    INSERT INTO Table2 (Id, TestData1, TestData2, Name)
    VALUES(1, 10, 5.5, 'Testing1');
    INSERT INTO Table2 (Id, TestData1, TestData2, Name)
    VALUES(2, 20, -1.2, 'Testing2');
    INSERT INTO Table2 (Id, TestData1, TestData2, Name)
    VALUES(3, NULL, NULL, 'Testing3');
    INSERT INTO Table3 (Id, Name, Data)
    VALUES (1, 'Testing1', 10);
    INSERT INTO Table3 (Id, Name, Data)
    VALUES (2, 'Testing2', 100);

See Also


Walkthrough: Generating F# Types from a DBML File (F#)


SqlDataConnection Type Provider (F#)

Query Expressions (F#)

SqlMetal.exe (Code Generation Tool)

Other Resources

Type Providers