从文本文件导入到SQL Server数据库,是ADO.NET太慢了吗?

Vim*_*987 2 c# sql-server ado.net

我的程序现在仍在运行,以将数据从日志文件导入远程SQL Server数据库.日志文件大小约为80MB,包含大约470000行,大约25000行数据.我的程序只能导入300行/秒,这真的很糟糕.:(

public static int ImportData(string strPath)
{
    //NameValueCollection collection = ConfigurationManager.AppSettings;

    using (TextReader sr = new StreamReader(strPath))
    {
        sr.ReadLine(); //ignore three first lines of log file
        sr.ReadLine();
        sr.ReadLine();
        string strLine;
        var cn = new SqlConnection(ConnectionString);
        cn.Open();

        while ((strLine = sr.ReadLine()) != null)
        {
            {
                if (strLine.Trim() != "") //if not a blank line, then import into database
                {
                    InsertData(strLine, cn);
                    _count++;
                }
            }
        }
        cn.Close();
        sr.Close();

        return _count;
    }
}
Run Code Online (Sandbox Code Playgroud)

InsertData只是使用ADO.NET的普通插入方法.它使用解析方法:

public Data(string strLine)
{
    string[] list = strLine.Split(new[] {'\t'});
    try
    {
        Senttime = DateTime.Parse(list[0] + " " + list[1]);
    }
    catch (Exception)
    {
    }

    Clientip = list[2];
    Clienthostname = list[3];

    Partnername = list[4];
    Serverhostname = list[5];
    Serverip = list[6];

    Recipientaddress = list[7];
    Eventid = Convert.ToInt16(list[8]);
    Msgid = list[9];
    Priority = Convert.ToInt16(list[10]);
    Recipientreportstatus = Convert.ToByte(list[11]);
    Totalbytes = Convert.ToInt32(list[12]);
    Numberrecipient = Convert.ToInt16(list[13]);
    DateTime temp;
    if (DateTime.TryParse(list[14], out temp))
    {
        OriginationTime = temp;
    }
    else
    {
        OriginationTime = null;
    }
    Encryption = list[15];
    ServiceVersion = list[16];
    LinkedMsgid = list[17];
    MessageSubject = list[18];
    SenderAddress = list[19];
}
Run Code Online (Sandbox Code Playgroud)

InsertData方法:

private static void InsertData(string strLine, SqlConnection cn)
{
    var dt = new Data(strLine); //parse the log line into proper fields 
    const string cnnStr =
        "INSERT INTO LOGDATA ([SentTime]," + "[client-ip]," +
        "[Client-hostname]," + "[Partner-Name]," + "[Server-hostname]," +
        "[server-IP]," + "[Recipient-Address]," + "[Event-ID]," + "[MSGID]," +
        "[Priority]," + "[Recipient-Report-Status]," + "[total-bytes]," +
        "[Number-Recipients]," + "[Origination-Time]," + "[Encryption]," +
        "[service-Version]," + "[Linked-MSGID]," + "[Message-Subject]," +
        "[Sender-Address]) " + " VALUES (     " + "@Senttime," + "@Clientip," +
        "@Clienthostname," + "@Partnername," + "@Serverhostname," + "@Serverip," +
        "@Recipientaddress," + "@Eventid," + "@Msgid," + "@Priority," +
        "@Recipientreportstatus," + "@Totalbytes," + "@Numberrecipient," +
        "@OriginationTime," + "@Encryption," + "@ServiceVersion," +
        "@LinkedMsgid," + "@MessageSubject," + "@SenderAddress)";


    var cmd = new SqlCommand(cnnStr, cn) {CommandType = CommandType.Text};

    cmd.Parameters.AddWithValue("@Senttime", dt.Senttime);
    cmd.Parameters.AddWithValue("@Clientip", dt.Clientip);
    cmd.Parameters.AddWithValue("@Clienthostname", dt.Clienthostname);
    cmd.Parameters.AddWithValue("@Partnername", dt.Partnername);
    cmd.Parameters.AddWithValue("@Serverhostname", dt.Serverhostname);
    cmd.Parameters.AddWithValue("@Serverip", dt.Serverip);
    cmd.Parameters.AddWithValue("@Recipientaddress", dt.Recipientaddress);
    cmd.Parameters.AddWithValue("@Eventid", dt.Eventid);
    cmd.Parameters.AddWithValue("@Msgid", dt.Msgid);
    cmd.Parameters.AddWithValue("@Priority", dt.Priority);
    cmd.Parameters.AddWithValue("@Recipientreportstatus", dt.Recipientreportstatus);
    cmd.Parameters.AddWithValue("@Totalbytes", dt.Totalbytes);
    cmd.Parameters.AddWithValue("@Numberrecipient", dt.Numberrecipient);
    if (dt.OriginationTime != null)
        cmd.Parameters.AddWithValue("@OriginationTime", dt.OriginationTime);
    else
        cmd.Parameters.AddWithValue("@OriginationTime", DBNull.Value);
            //if OriginationTime was null, then insert with null value to this column
    cmd.Parameters.AddWithValue("@Encryption", dt.Encryption);
    cmd.Parameters.AddWithValue("@ServiceVersion", dt.ServiceVersion);
    cmd.Parameters.AddWithValue("@LinkedMsgid", dt.LinkedMsgid);
    cmd.Parameters.AddWithValue("@MessageSubject", dt.MessageSubject);
    cmd.Parameters.AddWithValue("@SenderAddress", dt.SenderAddress);
    cmd.ExecuteNonQuery();
}
Run Code Online (Sandbox Code Playgroud)

我的程序如何运行得更快?非常感谢!

Mat*_*lls 13

使用SqlBulkCopy.

编辑:我创建了一个最小的实现IDataReader并创建了一个Batch类型,以便我可以使用插入任意的内存数据SqlBulkCopy.这是重要的一点:

IDataReader dr = batch.GetDataReader();
using (SqlTransaction tx = _connection.BeginTransaction())
{
    try
    {
        using (SqlBulkCopy sqlBulkCopy =
            new SqlBulkCopy(_connection, SqlBulkCopyOptions.Default, tx))
        {
            sqlBulkCopy.DestinationTableName = TableName;
            SetColumnMappings(sqlBulkCopy.ColumnMappings);
            sqlBulkCopy.WriteToServer(dr);
            tx.Commit();
        }
    }
    catch
    {
        tx.Rollback();
        throw;
    }
}
Run Code Online (Sandbox Code Playgroud)

其余的实现留给读者:)

提示:IDataReader您需要实现的唯一部分是Read,GetValueFieldCount.

  • SQLBulkCopy是要走的路.我曾经在SQL 6.5/7.0天中使用bcp来从CSV导入数据并发现它非常快.SqlBulkCopy本质上是与托管代码相同的功能. (2认同)