让我们把你的优秀和最喜欢的扩展方法列一个列表。

要求是必须发布完整的代码,以及如何使用它的示例和解释。

基于对这个主题的高度兴趣,我在Codeplex上建立了一个名为extensionoverflow的开源项目。

请将您的回答标记为接受,以便将代码放入Codeplex项目。

请张贴完整的源代码,而不是一个链接。

Codeplex上新闻:

24.08.2010 Codeplex页面现在在这里:http://extensionoverflow.codeplex.com/

11.11.2008 XmlSerialize / XmlDeserialize现在是实现和单元测试。

11.11.2008仍有发展空间。;-)现在就加入!

11.11.2008第三位贡献者加入了ExtensionOverflow,欢迎加入BKristensen

11.11.2008 FormatWith现在是实现和单元测试。

09.11.2008第二个贡献者加入ExtensionOverflow。欢迎来到chakrit。

我们需要更多的开发人员。: -)

09.11.2008 ThrowIfArgumentIsNull现已在Codeplex上实现和单元测试。


当前回答

有时需要有类的实例,不管是否有效,但不是null

public static T Safe<T>(this T obj) where T : new()
{
    if (obj == null)
    {
        obj = new T();
    }

    return obj;
}

用法如下:

MyClass myClass = Provider.GetSomeResult();
string temp = myClass.Safe().SomeValue;

而不是:

MyClass myClass = Provider.GetSomeResult();
string temp = "some default value";
if (myClass != null)
{
        temp = myClass.SomeValue;
}

如果是口是心非的话,我很抱歉,但我没有找到。

其他回答

public static class EnumerableExtensions
{
    [Pure]
    public static U MapReduce<T, U>(this IEnumerable<T> enumerable, Func<T, U> map, Func<U, U, U> reduce)
    {
        CodeContract.RequiresAlways(enumerable != null);
        CodeContract.RequiresAlways(enumerable.Skip(1).Any());
        CodeContract.RequiresAlways(map != null);
        CodeContract.RequiresAlways(reduce != null);
        return enumerable.AsParallel().Select(map).Aggregate(reduce);
    }
    [Pure]
    public static U MapReduce<T, U>(this IList<T> list, Func<T, U> map, Func<U, U, U> reduce)
    {
        CodeContract.RequiresAlways(list != null);
        CodeContract.RequiresAlways(list.Count >= 2);
        CodeContract.RequiresAlways(map != null);
        CodeContract.RequiresAlways(reduce != null);
        U result = map(list[0]);
        for (int i = 1; i < list.Count; i++)
        {
            result = reduce(result,map(list[i]));
        }
        return result;
    }

    //Parallel version; creates garbage
    [Pure]
    public static U MapReduce<T, U>(this IList<T> list, Func<T, U> map, Func<U, U, U> reduce)
    {
        CodeContract.RequiresAlways(list != null);
        CodeContract.RequiresAlways(list.Skip(1).Any());
        CodeContract.RequiresAlways(map != null);
        CodeContract.RequiresAlways(reduce != null);

        U[] mapped = new U[list.Count];
        Parallel.For(0, mapped.Length, i =>
            {
                mapped[i] = map(list[i]);
            });
        U result = mapped[0];
        for (int i = 1; i < list.Count; i++)
        {
            result = reduce(result, mapped[i]);
        }
        return result;
    }

}

您可能已经知道扩展方法的一个有趣用法是作为一种mixin。一些扩展方法,比如XmlSerializable,几乎污染了所有类;这对大多数人来说没有意义,比如Thread和SqlConnection。

一些功能应该显式地混合到希望拥有它的类中。我对这种类型提出了一种新的表示法,以M为前缀。

XmlSerializable是这样的:

public interface MXmlSerializable { }
public static class XmlSerializable {
  public static string ToXml(this MXmlSerializable self) {
    if (self == null) throw new ArgumentNullException();
    var serializer = new XmlSerializer(self.GetType());
    using (var writer = new StringWriter()) {
      serializer.Serialize(writer, self);
      return writer.GetStringBuilder().ToString();
    }
  }
  public static T FromXml<T>(string xml) where T : MXmlSerializable {
    var serializer = new XmlSerializer(typeof(T));
    return (T)serializer.Deserialize(new StringReader(xml));
  }
}

然后一个类将其混合:

public class Customer : MXmlSerializable {
  public string Name { get; set; }
  public bool Preferred { get; set; }
}

用法很简单:

var customer = new Customer { 
  Name = "Guybrush Threepwood", 
  Preferred = true };
var xml = customer.ToXml();

如果您喜欢这个想法,您可以在项目中为有用的mixin创建一个新的名称空间。你怎么看?

哦,顺便说一下,我认为大多数扩展方法都应该显式地测试null。

// This file contains extension methods for generic List<> class to operate on sorted lists.
// Duplicate values are OK.
// O(ln(n)) is still much faster then the O(n) of LINQ's searches/filters.
static partial class SortedList
{
    // Return the index of the first element with the key greater then provided.
    // If there's no such element within the provided range, it returns iAfterLast.
    public static int sortedFirstGreaterIndex<tElt, tKey>( this IList<tElt> list, Func<tElt, tKey, int> comparer, tKey key, int iFirst, int iAfterLast )
    {
        if( iFirst < 0 || iAfterLast < 0 || iFirst > list.Count || iAfterLast > list.Count )
            throw new IndexOutOfRangeException();
        if( iFirst > iAfterLast )
            throw new ArgumentException();
        if( iFirst == iAfterLast )
            return iAfterLast;

        int low = iFirst, high = iAfterLast;
        // The code below is inspired by the following article:
        // http://en.wikipedia.org/wiki/Binary_search#Single_comparison_per_iteration
        while( low < high )
        {
            int mid = ( high + low ) / 2;
            // 'mid' might be 'iFirst' in case 'iFirst+1 == iAfterLast'.
            // 'mid' will never be 'iAfterLast'.
            if( comparer( list[ mid ], key ) <= 0 ) // "<=" since we gonna find the first "greater" element
                low = mid + 1;
            else
                high = mid;
        }
        return low;
    }

    // Return the index of the first element with the key greater then the provided key.
    // If there's no such element, returns list.Count.
    public static int sortedFirstGreaterIndex<tElt, tKey>( this IList<tElt> list, Func<tElt, tKey, int> comparer, tKey key )
    {
        return list.sortedFirstGreaterIndex( comparer, key, 0, list.Count );
    }

    // Add an element to the sorted array.
    // This could be an expensive operation if frequently adding elements that sort firstly.
    // This is cheap operation when adding elements that sort near the tail of the list.
    public static int sortedAdd<tElt>( this List<tElt> list, Func<tElt, tElt, int> comparer, tElt elt )
    {
        if( list.Count == 0 || comparer( list[ list.Count - 1 ], elt ) <= 0 )
        {
            // either the list is empty, or the item is greater then all elements already in the collection.
            list.Add( elt );
            return list.Count - 1;
        }
        int ind = list.sortedFirstGreaterIndex( comparer, elt );
        list.Insert( ind, elt );
        return ind;
    }

    // Find first exactly equal element, return -1 if not found.
    public static int sortedFindFirstIndex<tElt, tKey>( this List<tElt> list, Func<tElt, tKey, int> comparer, tKey elt )
    {
        int low = 0, high = list.Count - 1;

        while( low < high )
        {
            int mid = ( high + low ) / 2;
            if( comparer( list[ mid ], elt ) < 0 )
                low = mid + 1;
            else
                high = mid; // this includes the case when we've found an element exactly matching the key
        }
        if( high >= 0 && 0 == comparer( list[ high ], elt ) )
            return high;
        return -1;
    }

    // Return the IEnumerable that returns array elements in the reverse order.
    public static IEnumerable<tElt> sortedReverse<tElt>( this List<tElt> list )
    {
        for( int i=list.Count - 1; i >= 0; i-- )
            yield return list[ i ];
    }
}

我经常用这个来表示可为空的数字。我帮助用0,NaN,无穷大…

public static bool IsNullOrDefault<T>(this T? o) 
    where T : struct
{
        return o == null || o.Value.Equals(default(T));
}

在将表单输入放入数据库之前解析表单输入时,这些扩展方法对我来说非常有用

public static int? ToInt(this string input) 
{
    int val;
    if (int.TryParse(input, out val))
        return val;
    return null;
}

public static DateTime? ToDate(this string input)
{
    DateTime val;
    if (DateTime.TryParse(input, out val))
        return val;
    return null;
}

public static decimal? ToDecimal(this string input)
{
    decimal val;
    if (decimal.TryParse(input, out val))
        return val;
    return null;
}