比较C#中的求和方法

Question

我正在研究一个使用大量求和方法的项目的一部分.这些求和方法应用于Datatable

为了测试最佳方法,我使用以下方法

数据结构

class LogParser
{
     public DataTable PGLStat_Table = new DataTable();
     public LogParser()
     {
         PGLStat_Table.Columns.Add("type", typeof(string)); 
         PGLStat_Table.Columns.Add("desc", typeof(string)); 
         PGLStat_Table.Columns.Add("count", typeof(int));
         PGLStat_Table.Columns.Add("duration", typeof(decimal));
         PGLStat_Table.Columns.Add("cper", typeof(decimal));
         PGLStat_Table.Columns.Add("dper", typeof(decimal));
         PGLStat_Table.Columns.Add("occurancedata", typeof(string));  
     }       
}

以下方法用于填充表格

LogParser pglp = new LogParser();
Random r2 = new Random();
for (int i = 1; i < 1000000; i++)
{
    int c2 = r2.Next(1, 1000);
    pglp.PGLStat_Table.Rows.Add("Type" + i.ToString(), "desc" + i , c2, 0, 0, 0, " ");
}

• Sum应用于count列,其中c2的值被更新

---

以下方法用于计算总和

方法1使用Compute

Stopwatch s2 = new Stopwatch();
s2.Start();
object sumObject;
sumObject = pglp.PGLStat_Table.Compute("Sum(count)", " ");
s2.Stop();
long d1 = s2.ElapsedMilliseconds;

方法2使用Foreach循环

s2.Restart();
int totalcount = 0;
foreach (DataRow dr in pglp.PGLStat_Table.Rows)
{
   int c = Convert.ToInt32(dr["count"].ToString());
   totalcount = totalcount + c;
}
s2.Stop();
long d2 = s2.ElapsedMilliseconds;

方法3使用Linq

s2.Restart();
var sum = pglp.PGLStat_Table.AsEnumerable().Sum(x => x.Field<int>("count"));
MessageBox.Show(sum.ToString());
s2.Stop();
long d3 = s2.ElapsedMilliseconds;

比较结果后

a)foreach是最快的481ms

b)接下来是linq 1016ms

c)然后计算2253ms

---

查询1

我在下面的语句中意外地将"c2改为i"

 pglp.PGLStat_Table.Rows.Add("Type" + i.ToString(), "desc" + i , i, 0, 0, 0, " ");

Linq语句产生错误

算术运算导致溢出.

而Compute和Foreach循环仍然能够完成计算,尽管可能不正确.

这种行为是引起关注还是我错过了指令？(计算的数字也很大)

查询2

我的印象是Linq做得最快,是否有优化的方法或参数使其表现更好.

谢谢你的建议

阿文德

Answer 1

接下来是最快的总和（使用预计算 DataColumn 并直接转换为 int）：

  static int Sum(LogParser pglp)
  {
    var column = pglp.PGLStat_Table.Columns["count"];
    int totalcount = 0;
    foreach (DataRow dr in pglp.PGLStat_Table.Rows)
    {
      totalcount += (int)dr[column];
    }
    return totalcount;
  }

统计：

00:00:00.1442297, for/each, by column, (int)
00:00:00.1595430, for/each, by column, Field<int>
00:00:00.6961964, for/each, by name, Convert.ToInt
00:00:00.1959104, linq, cast<DataRow>, by column, (int)

其他代码：

  static int Sum_ForEach_ByColumn_Field(LogParser pglp)
  {
    var column = pglp.PGLStat_Table.Columns["count"];
    int totalcount = 0;
    foreach (DataRow dr in pglp.PGLStat_Table.Rows)
    {
      totalcount += dr.Field<int>(column);
    }
    return totalcount;
  }
  static int Sum_ForEach_ByName_Convert(LogParser pglp)
  {
    int totalcount = 0;
    foreach (DataRow dr in pglp.PGLStat_Table.Rows)
    {
      int c = Convert.ToInt32(dr["count"].ToString());
      totalcount = totalcount + c;
    }
    return totalcount;
  }
  static int Sum_Linq(LogParser pglp)
  {
    var column = pglp.PGLStat_Table.Columns["count"];
    return pglp.PGLStat_Table.Rows.Cast<DataRow>().Sum(row => (int)row[column]);
  }


    var data = GenerateData();
    Sum(data);
    Sum_Linq2(data);
    var count = 3;
    foreach (var info in new[]
      {
        new {Name = "for/each, by column, (int)", Method = (Func<LogParser, int>)Sum},
        new {Name = "for/each, by column, Field<int>", Method = (Func<LogParser, int>)Sum_ForEach_ByColumn_Field},
        new {Name = "for/each, by name, Convert.ToInt", Method = (Func<LogParser, int>)Sum_ForEach_ByName_Convert},
        new {Name = "linq, cast<DataRow>, by column, (int)", Method = (Func<LogParser, int>)Sum_Linq},
      })
    {
      var watch = new Stopwatch();
      for (var i = 0; i < count; ++i)
      {
        watch.Start();
        var sum = info.Method(data);
        watch.Stop();
      }
      Console.WriteLine("{0}, {1}", TimeSpan.FromTicks(watch.Elapsed.Ticks / count), info.Name);
    }