I have a SQL Server 2005 database. In a few procedures I have table parameters that I pass to a stored proc as an nvarchar
(separated by commas) and internally divide into single values. I add it to the SQL command parameters list like this:
cmd.Parameters.Add("@Logins", SqlDbType.NVarchar).Value = "jim18,jenny1975,cosmo";
I have to migrate the database to SQL Server 2008. I know that there are table value parameters, and I know how to use them in stored procedures. But I don't know how to pass one to the parameters list in an SQL command.
Does anyone know correct syntax of the Parameters.Add
procedure? Or is there another way to pass this parameter?
DataTable
, DbDataReader
, or IEnumerable<SqlDataRecord>
objects can be used to populate a table-valued parameter per the MSDN article Table-Valued Parameters in SQL Server 2008 (ADO.NET).
The following example illustrates using either a DataTable
or an IEnumerable<SqlDataRecord>
:
SQL Code:
CREATE TABLE dbo.PageView
(
PageViewID BIGINT NOT NULL CONSTRAINT pkPageView PRIMARY KEY CLUSTERED,
PageViewCount BIGINT NOT NULL
);
CREATE TYPE dbo.PageViewTableType AS TABLE
(
PageViewID BIGINT NOT NULL
);
CREATE PROCEDURE dbo.procMergePageView
@Display dbo.PageViewTableType READONLY
AS
BEGIN
MERGE INTO dbo.PageView AS T
USING @Display AS S
ON T.PageViewID = S.PageViewID
WHEN MATCHED THEN UPDATE SET T.PageViewCount = T.PageViewCount + 1
WHEN NOT MATCHED THEN INSERT VALUES(S.PageViewID, 1);
END
C# Code:
private static void ExecuteProcedure(bool useDataTable,
string connectionString,
IEnumerable<long> ids)
{
using (SqlConnection connection = new SqlConnection(connectionString))
{
connection.Open();
using (SqlCommand command = connection.CreateCommand())
{
command.CommandText = "dbo.procMergePageView";
command.CommandType = CommandType.StoredProcedure;
SqlParameter parameter;
if (useDataTable) {
parameter = command.Parameters
.AddWithValue("@Display", CreateDataTable(ids));
}
else
{
parameter = command.Parameters
.AddWithValue("@Display", CreateSqlDataRecords(ids));
}
parameter.SqlDbType = SqlDbType.Structured;
parameter.TypeName = "dbo.PageViewTableType";
command.ExecuteNonQuery();
}
}
}
private static DataTable CreateDataTable(IEnumerable<long> ids)
{
DataTable table = new DataTable();
table.Columns.Add("ID", typeof(long));
foreach (long id in ids)
{
table.Rows.Add(id);
}
return table;
}
private static IEnumerable<SqlDataRecord> CreateSqlDataRecords(IEnumerable<long> ids)
{
SqlMetaData[] metaData = new SqlMetaData[1];
metaData[0] = new SqlMetaData("ID", SqlDbType.BigInt);
SqlDataRecord record = new SqlDataRecord(metaData);
foreach (long id in ids)
{
record.SetInt64(0, id);
yield return record;
}
}
Further to Ryan's answer you will also need to set the DataColumn
's Ordinal
property if you are dealing with a table-valued parameter
with multiple columns whose ordinals are not in alphabetical order.
As an example, if you have the following table value that is used as a parameter in SQL:
CREATE TYPE NodeFilter AS TABLE (
ID int not null
Code nvarchar(10) not null,
);
You would need to order your columns as such in C#:
table.Columns["ID"].SetOrdinal(0);
// this also bumps Code to ordinal of 1
// if you have more than 2 cols then you would need to set more ordinals
If you fail to do this you will get a parse error, failed to convert nvarchar to int.
Generic
public static DataTable ToTableValuedParameter<T, TProperty>(this IEnumerable<T> list, Func<T, TProperty> selector)
{
var tbl = new DataTable();
tbl.Columns.Add("Id", typeof(T));
foreach (var item in list)
{
tbl.Rows.Add(selector.Invoke(item));
}
return tbl;
}
The cleanest way to work with it. Assuming your table is a list of integers called "dbo.tvp_Int" (Customize for your own table type)
Create this extension method...
public static void AddWithValue_Tvp_Int(this SqlParameterCollection paramCollection, string parameterName, List<int> data)
{
if(paramCollection != null)
{
var p = paramCollection.Add(parameterName, SqlDbType.Structured);
p.TypeName = "dbo.tvp_Int";
DataTable _dt = new DataTable() {Columns = {"Value"}};
data.ForEach(value => _dt.Rows.Add(value));
p.Value = _dt;
}
}
Now you can add a table valued parameter in one line anywhere simply by doing this:
cmd.Parameters.AddWithValueFor_Tvp_Int("@IDValues", listOfIds);
if(paramCollection != null)
check at the top of the method will be fine
IEnumerable
instead of List
in the signature, that way you can pass anything that is IEnumerable
, not just lists, Since you're not using any function specific to List
, I don't really see a reason not to us IEnumerable
Use this code to create suitable parameter from your type:
private SqlParameter GenerateTypedParameter(string name, object typedParameter)
{
DataTable dt = new DataTable();
var properties = typedParameter.GetType().GetProperties().ToList();
properties.ForEach(p =>
{
dt.Columns.Add(p.Name, Nullable.GetUnderlyingType(p.PropertyType) ?? p.PropertyType);
});
var row = dt.NewRow();
properties.ForEach(p => { row[p.Name] = (p.GetValue(typedParameter) ?? DBNull.Value); });
dt.Rows.Add(row);
return new SqlParameter
{
Direction = ParameterDirection.Input,
ParameterName = name,
Value = dt,
SqlDbType = SqlDbType.Structured
};
}
Success story sharing
DataTable
as the parameter value, setSqlDbType
toStructured
andTypeName
to the database UDT name.null
.CreateSqlDataRecords
will never returnnull
if given an emptyids
parameter.DataTable
(orDataSet
) only implement it because they have to suppiort drag&drop capabilities in Visual-Studio, so they implementIComponent
which implementsIDisposable
. If you don't use the designer but create it manually there's no reason to dispose it (or to use theusing
-statement). So this is one of the exceptions of the golden rule "dispose everything that implementsIDisposable
".DataTable
approach consumes the hell of a memory, whereas theIEnumerable
one simply does not.