如何从文本文件中读取数百万行并快速插入表中

时间:2016-07-25 07:36:32

标签: c# datatable bulkinsert large-files read-write

我已经浏览了Insert 2 million rows into SQL Server quickly链接,发现我可以使用批量插入来完成此操作。所以我正在尝试创建数据表(代码如下),但由于这是一个巨大的文件(超过300K行),我在代码中得到OutOfMemoryEexception

string line;
DataTable data = new DataTable();
string[] columns = null;    
bool isInserted = false;           

using (TextReader tr = new StreamReader(_fileName, Encoding.Default))
{
   if (columns == null)
   {
      line = tr.ReadLine();
      columns = line.Split(',');
   }

   for (int iColCount = 0; iColCount < columns.Count(); iColCount++)
   {
      data.Columns.Add("Column" + iColCount, typeof(string));
   }                       

   string[] columnVal;

   while ((line = tr.ReadLine()) != null)
   {
        columnVal = line.Split(','); // OutOfMemoryException throwing in this line
        data.Rows.Add(columnVal);
    }
}

经过长时间的工作后,我将代码修改为如下所示,但是在向数据表添加行时我也得到了OutOfMemoryException

 DataTable data = new DataTable();
 string[] columns = null;
 var line = string.Empty;
 using (TextReader tr = new StreamReader(_fileName, Encoding.Default))
 {
     if (columns == null)
     {
         line = tr.ReadLine();
         columns = line.Split(',');
     }

     for (int iColCount = 0; iColCount < columns.Count(); iColCount++)
     {
        data.Columns.Add("Column" + iColCount, typeof(string));
     }
  }

  // Split the rows in 20000 rows in different list

  var _fileList = File.ReadLines(_fileName, Encoding.Default).ToList();
  var splitChunks = new List<List<string>>();
  splitChunks = SplitFile(_fileList, 20000);

 Parallel.ForEach(splitChunks, lstChunks =>
 {
   foreach (var rows in lstChunks)
   {
     string[] lineFields = rows.Split(',');
     DataRow row = datatbl.NewRow();
     for (int iCount = 0; iCount < lineFields.Count(); iCount++)
     {
        row[iCount] = lineFields[iCount] == string.Empty ? "" : lineFields[iCount].ToString();
     }
     datatbl.Rows.Add(row);
   }
 }); 

我可以使用以下代码进行下一级的批量插入:

SqlConnection SqlConnectionObj = GetSQLConnection();
SqlBulkCopy bulkCopy = new SqlBulkCopy(SqlConnectionObj, SqlBulkCopyOptions.TableLock | SqlBulkCopyOptions.FireTriggers | SqlBulkCopyOptions.UseInternalTransaction, null);
bulkCopy.DestinationTableName = "TempTable";
bulkCopy.WriteToServer(data);

文件包含以下类型的数据

4714,1370,AUSRICHTEN MASCHINELL

4870,1370,PLATTE STECKEN

0153,1900,CAULK GUN

0154,1900,NEW TERMINATOR

0360,1470,MU 186 MACCH。 X LAV。 S / A ASTE PS174

9113-H22,1970,MC DRILL BITS

代码需要将其转换为6行3列。

有没有更快的方法来实现上述功能来读取文件并为批量插入创建数据表?所以我不应该从索引异常中获取内存。

提前致谢。

3 个答案:

答案 0 :(得分:3)

  

您获得OutOfMemoryException的原因是因为您   创建内存数据表并尝试插入300K行   进入它

这是要放入内存的大量数据。

您应该做的就是从文本文件中读取每一定数量的行 - 您需要将其插入数据库中。

如何操作取决于您,您可以使用SQL或批量复制 - 但请记住,您无法读取整个文本文件并将其保存在内存中,因此请以块的形式进行操作。

答案 1 :(得分:1)

SqlBulkCopy.WriteToServerIDataReader的解决方案。我正在使用CSV,但我希望很容易修改其他类型。 SqlBulkCopy只使用IDateReader中的3件事情,我们必须实施它们:

  • public int FieldCount {get; }
  • public bool Read()
  • public object GetValue(int i)

所有其他属性和方法都可以不实现。关于SqlBulkCopy的有趣论文。完整代码:https://dotnetfiddle.net/giG3Ai。这是切割版本:

namespace SqlBulkCopy
{
    using System;
    using System.Collections.Generic;
    using System.IO;
    using System.Diagnostics;
    using System.Data;
    using System.Data.SqlClient;

    public class CsvReader : IDataReader
    {
        private readonly char CSV_DELIMITER = ',';

        private readonly StreamReader _sr;
        private readonly Dictionary<string, Func<string, object>> _csv2SqlType;
        private readonly string[] _headers;

        private string _line;
        private string[] _values;

        public int FieldCount { get { return _headers.Length; } }

        public CsvReader(string filePath, Dictionary<string, Func<string, object>> csvColumn2SqlTypeDict)
        {
            if (string.IsNullOrEmpty(filePath))
                throw new ArgumentException("is null or empty", "filePath");
            if (!System.IO.File.Exists(filePath))
                throw new IOException(string.Format("{0} doesn't exist or access denied", filePath));
            if (csvColumn2SqlTypeDict == null)
                throw new ArgumentNullException("csvColumn2SqlTypeDict");

            _sr = new StreamReader(filePath);
            _csv2SqlType = csvColumn2SqlTypeDict;
            _headers = ReadHeaders();
            ValidateHeaders();
        }
        public object GetValue(int i)
        {
            // Get column value
            var colValue = _values[i];
            // Get column name
            var colName = _headers[i];
            // Try to convert to SQL type
            try { return _csv2SqlType[colName](colValue); }
            catch { return null; }
        }
        public bool Read()
        {
            if (_sr.EndOfStream) return false;

            _line = _sr.ReadLine();
            _values = _line.Split(CSV_DELIMITER);
            // If row is invalid, go to next row
            if (_values.Length != _headers.Length)
                return Read();
            return true;
        }
        public void Dispose()
        {
            _sr.Dispose();
        }
        private void ValidateHeaders()
        {
            if (_headers.Length != _csv2SqlType.Keys.Count)
                throw new InvalidOperationException(string.Format("Read {0} columns, but csv2SqlTypeDict contains {1} columns", _headers.Length, _csv2SqlType.Keys));
            foreach (var column in _headers)
            {
                if (!_csv2SqlType.ContainsKey(column))
                    throw new InvalidOperationException(string.Format("There is no convertor for column '{0}'", column));
            }
        }
        private string[] ReadHeaders()
        {
            var headerLine = _sr.ReadLine();
            if (string.IsNullOrEmpty(headerLine))
                throw new InvalidDataException("There is no header in CSV!");
            var headers = headerLine.Split(CSV_DELIMITER);
            if (headers.Length == 0)
                throw new InvalidDataException("There is no header in CSV after Split!");
            return headers;
        }
    }
    public class Program
    {        
        public static void Main(string[] args)
        {
            // Converter from CSV columns to SQL columns
            var csvColumn2SqlTypeDict = new Dictionary<string, Func<string, object>>
            {
                { "int", (s) => Convert.ToInt32(s) },
                { "str", (s) => s },
                { "double", (s) => Convert.ToDouble(s) },
                { "date", (s) => Convert.ToDateTime(s) },
            };
            Stopwatch sw = Stopwatch.StartNew();
            try
            {
                // example.csv
                /***
                   int,str,double,date
                   1,abcd,2.5,15.04.2002
                   2,dab,2.7,15.04.2007
                   3,daqqb,4.7,14.04.2007
                 ***/
                using (var csvReader = new CsvReader("example.csv", csvColumn2SqlTypeDict))
                {
                    // TODO!!! Modify to your Connection string
                    var cs = @"Server=localhost\SQLEXPRESS;initial catalog=TestDb;Integrated Security=true";
                    using (var loader = new SqlBulkCopy(cs, SqlBulkCopyOptions.Default))
                    {
                        // TODO Modify to your Destination table
                        loader.DestinationTableName = "Test";
                        // Write from csvReader to database
                        loader.WriteToServer(csvReader);
                    }
                }
            }
            catch(Exception ex)
            {
                Console.WriteLine("Got an exception: {0}", ex);
                Console.WriteLine("Press 'Enter' to quit");
                Console.ReadLine();
                return;
            }
            finally { sw.Stop(); }
            Console.WriteLine("Data has been written in {0}", sw.Elapsed);
            Console.WriteLine("Press 'Enter' to quit");
            Console.ReadLine();
        }
        private static void ShowCsv(IDataReader dr)
        {
            int i = 0;
            while (dr.Read())
            {
                Console.WriteLine("Row# {0}", i);
                for (int j = 0; j < dr.FieldCount; j++)
                {
                    Console.WriteLine("{0} => {1}", j, dr.GetValue(j));
                }
                i++;
            }
        }
    }
}

答案 2 :(得分:0)

我发现忘记DataTable并且逐行使用普通的旧SQLClient更快。也更简单。这也击败了流式SQL函数,这可能是在SQL Server数据库中获取数据的最快方式。

尝试并测量速度,看看它是否足够快。如果不是,您可以随时尝试重新格式化文件(如有必要),让SQL Server使用Bulk Insert为您完成工作。