在SQL Server 中插入一条数据使用Insert语句,可是若是想要批量插入一堆数据的话,循环使用Insert不只效率低,并且会致使SQL一系统性能问题。下面介绍SQL Server支持的两种批量数据插入方法:Bulk和表值参数(Table-Valued Parameters)。sql
运行下面的脚本,创建测试数据库和表值参数。数据库
- --Create DataBase
- create database BulkTestDB;
- go
- use BulkTestDB;
- go
- --Create Table
- Create table BulkTestTable(
- Id int primary key,
- UserName nvarchar(32),
- Pwd varchar(16))
- go
- --Create Table Valued
- CREATE TYPE BulkUdt AS TABLE
- (Id int,
- UserName nvarchar(32),
- Pwd varchar(16))
缓存
下面咱们使用最简单的Insert语句来插入100万条数据,代码以下:性能
- Stopwatch sw = new Stopwatch();
-
- SqlConnection sqlConn = new SqlConnection(
- ConfigurationManager.ConnectionStrings["ConnStr"].ConnectionString);
-
- SqlCommand sqlComm = new SqlCommand();
- sqlComm.CommandText = string.Format("insert into BulkTestTable(Id,UserName,Pwd)values(@p0,@p1,@p2)");
- sqlComm.Parameters.Add("@p0", SqlDbType.Int);
- sqlComm.Parameters.Add("@p1", SqlDbType.NVarChar);
- sqlComm.Parameters.Add("@p2", SqlDbType.VarChar);
- sqlComm.CommandType = CommandType.Text;
- sqlComm.Connection = sqlConn;
- sqlConn.Open();
- try
- {
-
- for (int multiply = 0; multiply < 10; multiply++)
- {
- for (int count = multiply * 100000; count < (multiply + 1) * 100000; count++)
- {
-
- sqlComm.Parameters["@p0"].Value = count;
- sqlComm.Parameters["@p1"].Value = string.Format("User-{0}", count * multiply);
- sqlComm.Parameters["@p2"].Value = string.Format("Pwd-{0}", count * multiply);
- sw.Start();
- sqlComm.ExecuteNonQuery();
- sw.Stop();
- }
-
- Console.WriteLine(string.Format("Elapsed Time is {0} Milliseconds", sw.ElapsedMilliseconds));
- }
- }
- catch (Exception ex)
- {
- throw ex;
- }
- finally
- {
- sqlConn.Close();
- }
-
- Console.ReadLine();
测试
耗时图以下:spa

因为运行过慢,才插入10万条就耗时72390 milliseconds,因此我就手动强行中止了。.net
下面看一下使用Bulk插入的状况:pwa
bulk方法主要思想是经过在客户端把数据都缓存在Table中,而后利用SqlBulkCopy一次性把Table中的数据插入到数据库code
代码以下:orm
- public static void BulkToDB(DataTable dt)
- {
- SqlConnection sqlConn = new SqlConnection(
- ConfigurationManager.ConnectionStrings["ConnStr"].ConnectionString);
- SqlBulkCopy bulkCopy = new SqlBulkCopy(sqlConn);
- bulkCopy.DestinationTableName = "BulkTestTable";
- bulkCopy.BatchSize = dt.Rows.Count;
-
- try
- {
- sqlConn.Open();
- if (dt != null && dt.Rows.Count != 0)
- bulkCopy.WriteToServer(dt);
- }
- catch (Exception ex)
- {
- throw ex;
- }
- finally
- {
- sqlConn.Close();
- if (bulkCopy != null)
- bulkCopy.Close();
- }
- }
-
- public static DataTable GetTableSchema()
- {
- DataTable dt = new DataTable();
- dt.Columns.AddRange(new DataColumn[]{
- new DataColumn("Id",typeof(int)),
- new DataColumn("UserName",typeof(string)),
- new DataColumn("Pwd",typeof(string))});
-
- return dt;
- }
-
- static void Main(string[] args)
- {
- Stopwatch sw = new Stopwatch();
- for (int multiply = 0; multiply < 10; multiply++)
- {
- DataTable dt = Bulk.GetTableSchema();
- for (int count = multiply * 100000; count < (multiply + 1) * 100000; count++)
- {
- DataRow r = dt.NewRow();
- r[0] = count;
- r[1] = string.Format("User-{0}", count * multiply);
- r[2] = string.Format("Pwd-{0}", count * multiply);
- dt.Rows.Add(r);
- }
- sw.Start();
- Bulk.BulkToDB(dt);
- sw.Stop();
- Console.WriteLine(string.Format("Elapsed Time is {0} Milliseconds", sw.ElapsedMilliseconds));
- }
-
- Console.ReadLine();
- }
耗时图以下:

可见,使用Bulk后,效率和性能明显上升。使用Insert插入10万数据耗时72390,而如今使用Bulk插入100万数据才耗时17583。
最后再看看使用表值参数的效率,会另你大为惊讶的。
表值参数是SQL Server 2008新特性,简称TVPs。对于表值参数不熟悉的朋友,能够参考最新的book online,我也会另外写一篇关于表值参数的博客,不过这次不对表值参数的概念作过多的介绍。言归正传,看代码:
- public static void TableValuedToDB(DataTable dt)
- {
- SqlConnection sqlConn = new SqlConnection(
- ConfigurationManager.ConnectionStrings["ConnStr"].ConnectionString);
- const string TSqlStatement =
- "insert into BulkTestTable (Id,UserName,Pwd)" +
- " SELECT nc.Id, nc.UserName,nc.Pwd" +
- " FROM @NewBulkTestTvp AS nc";
- SqlCommand cmd = new SqlCommand(TSqlStatement, sqlConn);
- SqlParameter catParam = cmd.Parameters.AddWithValue("@NewBulkTestTvp", dt);
- catParam.SqlDbType = SqlDbType.Structured;
-
- catParam.TypeName = "dbo.BulkUdt";
- try
- {
- sqlConn.Open();
- if (dt != null && dt.Rows.Count != 0)
- {
- cmd.ExecuteNonQuery();
- }
- }
- catch (Exception ex)
- {
- throw ex;
- }
- finally
- {
- sqlConn.Close();
- }
- }
-
- public static DataTable GetTableSchema()
- {
- DataTable dt = new DataTable();
- dt.Columns.AddRange(new DataColumn[]{
- new DataColumn("Id",typeof(int)),
- new DataColumn("UserName",typeof(string)),
- new DataColumn("Pwd",typeof(string))});
-
- return dt;
- }
-
- static void Main(string[] args)
- {
- Stopwatch sw = new Stopwatch();
- for (int multiply = 0; multiply < 10; multiply++)
- {
- DataTable dt = TableValued.GetTableSchema();
- for (int count = multiply * 100000; count < (multiply + 1) * 100000; count++)
- {
- DataRow r = dt.NewRow();
- r[0] = count;
- r[1] = string.Format("User-{0}", count * multiply);
- r[2] = string.Format("Pwd-{0}", count * multiply);
- dt.Rows.Add(r);
- }
- sw.Start();
- TableValued.TableValuedToDB(dt);
- sw.Stop();
- Console.WriteLine(string.Format("Elapsed Time is {0} Milliseconds", sw.ElapsedMilliseconds));
- }
-
- Console.ReadLine();
- }
耗时图以下:

比Bulk还快5秒。
如需转载,请注明此文原创自CSDN TJVictor专栏:http://blog.csdn.net/tjvictor/archive/2009/07/18/4360030.aspx