If the table receiving new data and schema from a merge has a primary key, new rows from the incoming data are matched with existing rows that have the same Original primary key values as those in the incoming data. If the columns from the incoming schema match those of the existing schema, the data in the existing rows is modified. Columns that do not match the existing schema are either ignored or added based on the MissingSchemaAction parameter. New rows with primary key values that do not match any existing rows are appended to the existing table.

If incoming or existing rows have a row state of Added, their primary key values are matched using the Current primary key value of the Added row because no Original row version exists.

If an incoming table and an existing table contain a column with the same name but different data types, an exception is thrown and the MergeFailed event of the DataSet is raised. If an incoming table and an existing table both have defined keys, but the primary keys are for different columns, an exception is thrown and the MergeFailed event of the DataSet is raised.

If the table receiving new data from a merge does not have a primary key, new rows from the incoming data cannot be matched to existing rows in the table and are instead appended to the existing table.

DataTable objects can optionally be assigned a Namespace property value. When Namespace values are assigned, a DataSet can contain multiple DataTable objects with the same TableName value. During merge operations, both TableName and Namespace are used to identify the target of a merge. If no Namespace has been assigned, only the TableName is used to identify the target of a merge.

Note:

This behavior changed in version 2.0 of the .NET Framework. In version 1.1, namespaces were supported but were ignored during merge operations. For this reason, a DataSet that uses Namespace property values will have different behaviors depending on which version of the .NET Framework you are running. For example, suppose you have two DataSets containing DataTables with the same TableName property values but different Namespace property values. In version 1.1 of the .NET Framework, the different Namespace names will be ignored when merging the two DataSet objects. However, starting with version 2.0, merging causes two new DataTables to be created in the target DataSet. The original DataTables will be unaffected by the merge.

When you pass a DataSet, DataTable, or DataRow array to the Merge method, you can include optional parameters that specify whether or not to preserve changes in the existing DataSet, and how to handle new schema elements found in the incoming data. The first of these parameters after the incoming data is a Boolean flag, PreserveChanges, which specifies whether or not to preserve the changes in the existing DataSet. If the PreserveChanges flag is set to true, incoming values do not overwrite existing values in the Current row version of the existing row. If the PreserveChanges flag is set to false, incoming values do overwrite the existing values in the Current row version of the existing row. If the PreserveChanges flag is not specified, it is set to false by default. For more information about row versions, see Row States and Row Versions.

When PreserveChanges is true, the data from the existing row is maintained in the Current row version of the existing row, while the data from the Original row version of the existing row is overwritten with the data from the Original row version of the incoming row. The RowState of the existing row is set to Modified. The following exceptions apply:

• If the existing row has a RowState of Deleted, this RowState remains Deleted and is not set to Modified. In this case, the data from the incoming row will still be stored in the Original row version of the existing row, overwriting the Original row version of the existing row (unless the incoming row has a RowState of Added).

• If the incoming row has a RowState of Added, the data from the Original row version of the existing row will not be overwritten with data from the incoming row, because the incoming row does not have an Original row version.

When PreserveChanges is false, both the Current and Original row versions in the existing row are overwritten with the data from the incoming row, and the RowState of the existing row is set to the RowState of the incoming row. The following exceptions apply:

• If the incoming row has a RowState of Unchanged and the existing row has a RowState of Modified, Deleted, or Added, the RowState of the existing row is set to Modified.

• If the incoming row has a RowState of Added, and the existing row has a RowState of Unchanged, Modified, or Deleted, the RowState of the existing row is set to Modified. Also, the data from the Original row version of the existing row is not overwritten with data from the incoming row, because the incoming row does not have an Original row version.

You can use the optional MissingSchemaAction parameter of the Merge method to specify how Merge will handle schema elements in the incoming data that are not part of the existing DataSet.

The following table describes the options for MissingSchemaAction.

With the Merge method, constraints are not checked until all new data has been added to the existing DataSet. Once the data has been added, constraints are enforced on the current values in the DataSet. You must ensure that your code handles any exceptions that might be thrown due to constraint violations.

Consider a case where an existing row in a DataSet is an Unchanged row with a primary key value of 1. During a merge operation with a Modified incoming row with an Original primary key value of 2 and a Current primary key value of 1, the existing row and the incoming row are not considered matching because the Original primary key values differ. However, when the merge is completed and constraints are checked, an exception will be thrown because the Current primary key values violate the unique constraint for the primary key column.

Note:
When rows are inserted into a database table containing an auto incrementing column such as an identity column, the identity column value returned by the insert may not match the value in the DataSet, causing the returned rows to be appended instead of merged. For more information, see Retrieving Identity or Autonumber Values (ADO.NET).

The following code example merges two DataSet objects with differents schemas into one DataSet with the combined schemas of the two incoming DataSet objects.

Using connection As SqlConnection = New SqlConnection( _
   connectionString)

    Dim adapter As SqlDataAdapter = New SqlDataAdapter( _
      "SELECT CustomerID, CompanyName FROM Customers", connection)

    connection.Open()

    Dim customers As DataSet = New DataSet()
    adapter.FillSchema(customers, SchemaType.Source, "Customers")
    adapter.Fill(customers, "Customers")

    Dim orders As DataSet = New DataSet()
    orders.ReadXml("Orders.xml", XmlReadMode.ReadSchema)
    orders.AcceptChanges()

    customers.Merge(orders, True, MissingSchemaAction.AddWithKey)
End Using

using (SqlConnection connection =
           new SqlConnection(connectionString))
{
    SqlDataAdapter adapter = 
        new SqlDataAdapter(
        "SELECT CustomerID, CompanyName FROM dbo.Customers", 
        connection);

    connection.Open();

    DataSet customers = new DataSet();
    adapter.FillSchema(customers, SchemaType.Source, "Customers");
    adapter.Fill(customers, "Customers");

    DataSet orders = new DataSet();
    orders.ReadXml("Orders.xml", XmlReadMode.ReadSchema);
    orders.AcceptChanges();

    customers.Merge(orders, true, MissingSchemaAction.AddWithKey);

The following code example takes an existing DataSet with updates and passes those updates to a DataAdapter to be processed at the data source. The results are then merged into the original DataSet. After rejecting changes that resulted in an error, the merged changes are committed with AcceptChanges.

Dim customers As DataTable = dataSet.Tables("Customers")

' Make modifications to the Customers table.

' Get changes to the DataSet.
Dim dataSetChanges As DataSet = dataSet.GetChanges()

' Add an event handler to handle the errors during Update.
AddHandler adapter.RowUpdated, New SqlRowUpdatedEventHandler( _
  AddressOf OnRowUpdated)

connection.Open()
adapter.Update(dataSetChanges, "Customers")
connection.Close()

' Merge the updates.
dataSet.Merge(dataSetChanges, True, MissingSchemaAction.Add)

' Reject changes on rows with errors and clear the error.
Dim errRows() As DataRow = dataSet.Tables("Customers").GetErrors()
Dim errRow As DataRow
For Each errRow In errRows
    errRow.RejectChanges()
    errRow.RowError = Nothing
Next

' Commit the changes.
dataSet.AcceptChanges()

DataTable customers = dataSet.Tables["Customers"];

// Make modifications to the Customers table.

// Get changes to the DataSet.
DataSet dataSetChanges = dataSet.GetChanges();

// Add an event handler to handle the errors during Update.
adapter.RowUpdated += new SqlRowUpdatedEventHandler(OnRowUpdated);

connection.Open();
adapter.Update(dataSetChanges, "Customers");
connection.Close();

// Merge the updates.
dataSet.Merge(dataSetChanges, true, MissingSchemaAction.Add);

// Reject changes on rows with errors and clear the error.
DataRow[] errRows = dataSet.Tables["Customers"].GetErrors();
foreach (DataRow errRow in errRows)
{
    errRow.RejectChanges();
    errRow.RowError = null;
}

// Commit the changes.
dataSet.AcceptChanges();

Private Sub OnRowUpdated( _
    ByVal sender As Object, ByVal args As SqlRowUpdatedEventArgs)
    If args.Status = UpdateStatus.ErrorsOccurred Then
        args.Row.RowError = args.Errors.Message
        args.Status = UpdateStatus.SkipCurrentRow
    End If
End Sub

protected static void OnRowUpdated(
    object sender, SqlRowUpdatedEventArgs args)
{
    if (args.Status == UpdateStatus.ErrorsOccurred)
    {
        args.Row.RowError = args.Errors.Message;
        args.Status = UpdateStatus.SkipCurrentRow;
    }
}

Concepts

Other Resources