SQL SERVER 2005允许自定义聚合函数

不多说了，说明后面是完整的代码，用来将字符串型的字段的各行的值拼成一个大字符串，也就是通常所说的Concat

例如有如下表dict

 

ID	NAME	CATEGORY
1	RED	COLOR
2	BLUE	COLOR
3	APPLE	FRUIT
4	ORANGE	FRUIT

 

执行SQL语句：select category,dbo.concatenate(name) as names from dict group by category.

得到结果表如下

 

category	names
COLOR	REDBLUE
FRUIT	APPLEORANGE

 

如果觉得需要用逗号或分号或其他任何你想要的分隔符分开，可以修改下面的代码来实现。

 

在VS2005中，创建一个连接到目标库的SQL SERVER PROJECT，然后填加一个“聚合”，将下面的代码复制进去，编译后，部署即可，然后在SQL SERVER中的“可编程性”“函数”“聚合函数”中就可以看到该函数了。 

using System;
using System.Data;
using Microsoft.SqlServer.Server;
using System.Data.SqlTypes;
using System.IO;
using System.Text;

[Serializable]
[SqlUserDefinedAggregate(
    Format.UserDefined, //use clr serialization to serialize the intermediate result
    IsInvariantToNulls = true, //optimizer property
    IsInvariantToDuplicates = false, //optimizer property
    IsInvariantToOrder = false, //optimizer property
    MaxByteSize = 8000) //maximum size in bytes of persisted value
]
public class Concatenate : IBinarySerialize
{
    /// <summary>
    /// The variable that holds the intermediate result of the concatenation
    /// </summary>
    private StringBuilder intermediateResult;

    /// <summary>
    /// Initialize the internal data structures
    /// </summary>
    public void Init()
    {
        this.intermediateResult = new StringBuilder();
    }

    /// <summary>
    /// Accumulate the next value, not if the value is null
    /// </summary>
    /// <param name="value"></param>
    public void Accumulate(SqlString value)
    {
        if (value.IsNull)
        {
            return;
        }

        this.intermediateResult.Append(value.Value);
    }

    /// <summary>
    /// Merge the partially computed aggregate with this aggregate.
    /// </summary>
    /// <param name="other"></param>
    public void Merge(Concatenate other)
    {
        this.intermediateResult.Append(other.intermediateResult);
    }

    /// <summary>
    /// Called at the end of aggregation, to return the results of the aggregation.
    /// </summary>
    /// <returns></returns>
    public SqlString Terminate()
    {
        string output = string.Empty;
        //delete the trailing comma, if any
        if (this.intermediateResult != null
            && this.intermediateResult.Length > 0)
        {
            output = this.intermediateResult.ToString(0, this.intermediateResult.Length );
        }

        return new SqlString(output);
    }

    public void Read(BinaryReader r)
    {
        intermediateResult = new StringBuilder(r.ReadString());
    }

    public void Write(BinaryWriter w)
    {
        w.Write(this.intermediateResult.ToString());
    }
}

 

这里有几个比较重要的方法：Terminate，这个方法是聚合最后调用的方法，它返回最后的值。可以是SQL Server的任何标量；Accumulate，聚合每处理一行数据的时候都会调用一次，并将要处理的数据传给方法。可以在函数内部进行比如比较，合并之类的处理。