IProvider里有一个用于实现批量插入的插件服务接口IBatcherProvider。
/// <summary> /// 提供数据批量处理的方法。 /// </summary> public interface IBatcherProvider : IProviderService { /// <summary> /// 将 <see cref="DataTable"/> 的数据批量插入到数据库中。 /// </summary> /// <param name="dataTable">要批量插入的 <see cref="DataTable"/>。</param> /// <param name="batchSize">每批次写入的数据量。</param> void Insert(DataTable dataTable, int batchSize = 10000); }
SqlServer的批量插入很简单,使用SqlBulkCopy就可以,以下是该类的实现:
/// <summary> /// 为 System.Data.SqlClient 提供的用于批量操作的方法。 /// </summary> public sealed class MsSqlBatcher : IBatcherProvider { /// <summary> /// 获取或设置提供者服务的上下文。 /// </summary> public ServiceContext ServiceContext { get; set; } /// <summary> /// 将 <see cref="DataTable"/> 的数据批量插入到数据库中。 /// </summary> /// <param name="dataTable">要批量插入的 <see cref="DataTable"/>。</param> /// <param name="batchSize">每批次写入的数据量。</param> public void Insert(DataTable dataTable, int batchSize = 10000) { Checker.ArgumentNull(dataTable, "dataTable"); if (dataTable.Rows.Count == 0) { return; } using (var connection = (SqlConnection)ServiceContext.Database.CreateConnection()) { try { connection.TryOpen(); //给表名加上前后导符 var tableName = DbUtility.FormatByQuote(ServiceContext.Database.Provider.GetService<ISyntaxProvider>(), dataTable.TableName); using (var bulk = new SqlBulkCopy(connection, SqlBulkCopyOptions.KeepIdentity, null) { DestinationTableName = tableName, BatchSize = batchSize }) { //循环所有列,为bulk添加映射 dataTable.EachColumn(c => bulk.ColumnMappings.Add(c.ColumnName, c.ColumnName), c => !c.AutoIncrement); bulk.WriteToServer(dataTable); bulk.Close(); } } catch (Exception exp) { throw new BatcherException(exp); } finally { connection.TryClose(); } } } }
二、Oracle数据批量插入
System.Data.OracleClient不支持批量插入,因此只能使用Oracle.DataAccess组件来作为提供者。
/// <summary> /// Oracle.Data.Access 组件提供的用于批量操作的方法。 /// </summary> public sealed class OracleAccessBatcher : IBatcherProvider { /// <summary> /// 获取或设置提供者服务的上下文。 /// </summary> public ServiceContext ServiceContext { get; set; } /// <summary> /// 将 <see cref="DataTable"/> 的数据批量插入到数据库中。 /// </summary> /// <param name="dataTable">要批量插入的 <see cref="DataTable"/>。</param> /// <param name="batchSize">每批次写入的数据量。</param> public void Insert(DataTable dataTable, int batchSize = 10000) { Checker.ArgumentNull(dataTable, "dataTable"); if (dataTable.Rows.Count == 0) { return; } using (var connection = ServiceContext.Database.CreateConnection()) { try { connection.TryOpen(); using (var command = ServiceContext.Database.Provider.DbProviderFactory.CreateCommand()) { if (command == null) { throw new BatcherException(new ArgumentException("command")); } command.Connection = connection; command.CommandText = GenerateInserSql(ServiceContext.Database, command, dataTable); command.ExecuteNonQuery(); } } catch (Exception exp) { throw new BatcherException(exp); } finally { connection.TryClose(); } } } /// <summary> /// 生成插入数据的sql语句。 /// </summary> /// <param name="database"></param> /// <param name="command"></param> /// <param name="table"></param> /// <returns></returns> private string GenerateInserSql(IDatabase database, DbCommand command, DataTable table) { var names = new StringBuilder(); var values = new StringBuilder(); //将一个DataTable的数据转换为数组的数组 var data = table.ToArray(); //设置ArrayBindCount属性 command.GetType().GetProperty("ArrayBindCount").SetValue(command, table.Rows.Count, null); var syntax = database.Provider.GetService<ISyntaxProvider>(); for (var i = 0; i < table.Columns.Count; i++) { var column = table.Columns[i]; var parameter = database.Provider.DbProviderFactory.CreateParameter(); if (parameter == null) { continue; } parameter.ParameterName = column.ColumnName; parameter.Direction = ParameterDirection.Input; parameter.DbType = column.DataType.GetDbType(); parameter.Value = data[i]; if (names.Length > 0) { names.Append(","); values.Append(","); } names.AppendFormat("{0}", DbUtility.FormatByQuote(syntax, column.ColumnName)); values.AppendFormat("{0}{1}", syntax.ParameterPrefix, column.ColumnName); command.Parameters.Add(parameter); } return string.Format("INSERT INTO {0}({1}) VALUES ({2})", DbUtility.FormatByQuote(syntax, table.TableName), names, values); } }
三、SQLite数据批量插入
SQLite的批量插入只需开启事务就可以了,这个具体的原理不得而知。
public sealed class SQLiteBatcher : IBatcherProvider { /// <summary> /// 获取或设置提供者服务的上下文。 /// </summary> public ServiceContext ServiceContext { get; set; } /// <summary> /// 将 <see cref="DataTable"/> 的数据批量插入到数据库中。 /// </summary> /// <param name="dataTable">要批量插入的 <see cref="DataTable"/>。</param> /// <param name="batchSize">每批次写入的数据量。</param> public void Insert(DataTable dataTable, int batchSize = 10000) { Checker.ArgumentNull(dataTable, "dataTable"); if (dataTable.Rows.Count == 0) { return; } using (var connection = ServiceContext.Database.CreateConnection()) { DbTransaction transcation = null; try { connection.TryOpen(); transcation = connection.BeginTransaction(); using (var command = ServiceContext.Database.Provider.DbProviderFactory.CreateCommand()) { if (command == null) { throw new BatcherException(new ArgumentException("command")); } command.Connection = connection; command.CommandText = GenerateInserSql(ServiceContext.Database, dataTable); if (command.CommandText == string.Empty) { return; } var flag = new AssertFlag(); dataTable.EachRow(row => { var first = flag.AssertTrue(); ProcessCommandParameters(dataTable, command, row, first); command.ExecuteNonQuery(); }); } transcation.Commit(); } catch (Exception exp) { if (transcation != null) { transcation.Rollback(); } throw new BatcherException(exp); } finally { connection.TryClose(); } } } private void ProcessCommandParameters(DataTable dataTable, DbCommand command, DataRow row, bool first) { for (var c = 0; c < dataTable.Columns.Count; c++) { DbParameter parameter; //首次创建参数,是为了使用缓存 if (first) { parameter = ServiceContext.Database.Provider.DbProviderFactory.CreateParameter(); parameter.ParameterName = dataTable.Columns[c].ColumnName; command.Parameters.Add(parameter); } else { parameter = command.Parameters[c]; } parameter.Value = row[c]; } } /// <summary> /// 生成插入数据的sql语句。 /// </summary> /// <param name="database"></param> /// <param name="table"></param> /// <returns></returns> private string GenerateInserSql(IDatabase database, DataTable table) { var syntax = database.Provider.GetService<ISyntaxProvider>(); var names = new StringBuilder(); var values = new StringBuilder(); var flag = new AssertFlag(); table.EachColumn(column => { if (!flag.AssertTrue()) { names.Append(","); values.Append(","); } names.Append(DbUtility.FormatByQuote(syntax, column.ColumnName)); values.AppendFormat("{0}{1}", syntax.ParameterPrefix, column.ColumnName); }); return string.Format("INSERT INTO {0}({1}) VALUES ({2})", DbUtility.FormatByQuote(syntax, table.TableName), names, values); } }
四、MySql数据批量插入
/// <summary> /// 为 MySql.Data 组件提供的用于批量操作的方法。 /// </summary> public sealed class MySqlBatcher : IBatcherProvider { /// <summary> /// 获取或设置提供者服务的上下文。 /// </summary> public ServiceContext ServiceContext { get; set; } /// <summary> /// 将 <see cref="DataTable"/> 的数据批量插入到数据库中。 /// </summary> /// <param name="dataTable">要批量插入的 <see cref="DataTable"/>。</param> /// <param name="batchSize">每批次写入的数据量。</param> public void Insert(DataTable dataTable, int batchSize = 10000) { Checker.ArgumentNull(dataTable, "dataTable"); if (dataTable.Rows.Count == 0) { return; } using (var connection = ServiceContext.Database.CreateConnection()) { try { connection.TryOpen(); using (var command = ServiceContext.Database.Provider.DbProviderFactory.CreateCommand()) { if (command == null) { throw new BatcherException(new ArgumentException("command")); } command.Connection = connection; command.CommandText = GenerateInserSql(ServiceContext.Database, command, dataTable); if (command.CommandText == string.Empty) { return; } command.ExecuteNonQuery(); } } catch (Exception exp) { throw new BatcherException(exp); } finally { connection.TryClose(); } } } /// <summary> /// 生成插入数据的sql语句。 /// </summary> /// <param name="database"></param> /// <param name="command"></param> /// <param name="table"></param> /// <returns></returns> private string GenerateInserSql(IDatabase database, DbCommand command, DataTable table) { var names = new StringBuilder(); var values = new StringBuilder(); var types = new List<DbType>(); var count = table.Columns.Count; var syntax = database.Provider.GetService<ISyntaxProvider>(); table.EachColumn(c => { if (names.Length > 0) { names.Append(","); } names.AppendFormat("{0}", DbUtility.FormatByQuote(syntax, c.ColumnName)); types.Add(c.DataType.GetDbType()); }); var i = 0; foreach (DataRow row in table.Rows) { if (i > 0) { values.Append(","); } values.Append("("); for (var j = 0; j < count; j++) { if (j > 0) { values.Append(", "); } var isStrType = IsStringType(types[j]); var parameter = CreateParameter(database.Provider, isStrType, types[j], row[j], syntax.ParameterPrefix, i, j); if (parameter != null) { values.Append(parameter.ParameterName); command.Parameters.Add(parameter); } else if (isStrType) { values.AppendFormat("'{0}'", row[j]); } else { values.Append(row[j]); } } values.Append(")"); i++; } return string.Format("INSERT INTO {0}({1}) VALUES {2}", DbUtility.FormatByQuote(syntax, table.TableName), names, values); } /// <summary> /// 判断是否为字符串类别。 /// </summary> /// <param name="dbType"></param> /// <returns></returns> private bool IsStringType(DbType dbType) { return dbType == DbType.AnsiString || dbType == DbType.AnsiStringFixedLength || dbType == DbType.String || dbType == DbType.StringFixedLength; } /// <summary> /// 创建参数。 /// </summary> /// <param name="provider"></param> /// <param name="isStrType"></param> /// <param name="dbType"></param> /// <param name="value"></param> /// <param name="parPrefix"></param> /// <param name="row"></param> /// <param name="col"></param> /// <returns></returns> private DbParameter CreateParameter(IProvider provider, bool isStrType, DbType dbType, object value, char parPrefix, int row, int col) { //如果生成全部的参数,则速度会很慢,因此,只有数据类型为字符串(包含'号)和日期型时才添加参数 if ((isStrType && value.ToString().IndexOf('\'') != -1) || dbType == DbType.DateTime) { var name = string.Format("{0}p_{1}_{2}", parPrefix, row, col); var parameter = provider.DbProviderFactory.CreateParameter(); parameter.ParameterName = name; parameter.Direction = ParameterDirection.Input; parameter.DbType = dbType; parameter.Value = value; return parameter; } return null; } }
五、测试
接下来写一个测试用例来看一下使用批量插入的效果。
[Test] public void TestBatchInsert() { Console.WriteLine(TimeWatcher.Watch(() => InvokeTest(database => { var table = new DataTable("Batcher"); table.Columns.Add("Id", typeof(int)); table.Columns.Add("Name1", typeof(string)); table.Columns.Add("Name2", typeof(string)); table.Columns.Add("Name3", typeof(string)); table.Columns.Add("Name4", typeof(string)); //构造100000条数据 for (var i = 0; i < 100000; i++) { table.Rows.Add(i, i.ToString(), i.ToString(), i.ToString(), i.ToString()); } //获取 IBatcherProvider var batcher = database.Provider.GetService<IBatcherProvider>(); if (batcher == null) { Console.WriteLine("不支持批量插入。"); } else { batcher.Insert(table); } //输出batcher表的数据量 var sql = new SqlCommand("SELECT COUNT(1) FROM Batcher"); Console.WriteLine("当前共有 {0} 条数据", database.ExecuteScalar(sql)); }))); }
以下表中列出了四种数据库生成10万条数据各耗用的时间
数据库 | 耗用时间 |
MsSql | 00:00:02.9376300 |
Oracle | 00:00:01.5155959 |
SQLite | 00:00:01.6275634 |
MySql | 00:00:05.4166891 |