好吧，以下是我的看法:

完全懒惰:工作在无限枚举上 没有中间复制/缓冲 O(n)执行时间 当内部序列仅被部分消耗时也适用

public static IEnumerable<IEnumerable<T>> Chunks<T>(this IEnumerable<T> enumerable, int chunkSize) { if (chunkSize < 1) throw new ArgumentException("chunkSize must be positive"); using (var e = enumerable.GetEnumerator()) while (e.MoveNext()) { var remaining = chunkSize; // elements remaining in the current chunk var innerMoveNext = new Func<bool>(() => --remaining > 0 && e.MoveNext()); yield return e.GetChunk(innerMoveNext); while (innerMoveNext()) {/* discard elements skipped by inner iterator */} } } private static IEnumerable<T> GetChunk<T>(this IEnumerator<T> e, Func<bool> innerMoveNext) { do yield return e.Current; while (innerMoveNext()); } Example Usage var src = new [] {1, 2, 3, 4, 5, 6}; var c3 = src.Chunks(3); // {{1, 2, 3}, {4, 5, 6}}; var c4 = src.Chunks(4); // {{1, 2, 3, 4}, {5, 6}}; var sum = c3.Select(c => c.Sum()); // {6, 15} var count = c3.Count(); // 2 var take2 = c3.Select(c => c.Take(2)); // {{1, 2}, {4, 5}} Explanations The code works by nesting two yield based iterators. The outer iterator must keep track of how many elements have been effectively consumed by the inner (chunk) iterator. This is done by closing over remaining with innerMoveNext(). Unconsumed elements of a chunk are discarded before the next chunk is yielded by the outer iterator. This is necessary because otherwise you get inconsistent results, when the inner enumerables are not (completely) consumed (e.g. c3.Count() would return 6). Note: The answer has been updated to address the shortcomings pointed out by @aolszowka.

如果源集合实现了IList < T >(按索引随机访问)，我们可以使用下面的方法。它只在真正访问元素时获取元素，因此这对于延迟求值的集合特别有用。类似于unbounded IEnumerable< T >，但是对于IList< T >。

    public static IEnumerable<IEnumerable<T>> Chunkify<T>(this IList<T> src, int chunkSize)
    {
        if (src == null) throw new ArgumentNullException(nameof(src));
        if (chunkSize < 1) throw new ArgumentOutOfRangeException(nameof(chunkSize), $"must be > 0, got {chunkSize}");

        for(var ci = 0; ci <= src.Count/chunkSize; ci++){
            yield return Window(src, ci*chunkSize, Math.Min((ci+1)*chunkSize, src.Count)-1);
        }
    }

    private static IEnumerable<T> Window<T>(IList<T> src, int startIdx, int endIdx)
    {
        Console.WriteLine($"window {startIdx} - {endIdx}");
        while(startIdx <= endIdx){
            yield return src[startIdx++];
        }
    }

这是一个古老的解决方案，但我有一个不同的方法。我使用Skip来移动到所需的偏移量，并使用Take来提取所需的元素数量:

public static IEnumerable<IEnumerable<T>> Chunk<T>(this IEnumerable<T> source, 
                                                   int chunkSize)
{
    if (chunkSize <= 0)
        throw new ArgumentOutOfRangeException($"{nameof(chunkSize)} should be > 0");

    var nbChunks = (int)Math.Ceiling((double)source.Count()/chunkSize);

    return Enumerable.Range(0, nbChunks)
                     .Select(chunkNb => source.Skip(chunkNb*chunkSize)
                     .Take(chunkSize));
}

我认为下面的建议是最快的。为了能够使用数组，我牺牲了源Enumerable的惰性。复制和提前知道每个子列表的长度。

public static IEnumerable<T[]> Chunk<T>(this IEnumerable<T> items, int size)
{
    T[] array = items as T[] ?? items.ToArray();
    for (int i = 0; i < array.Length; i+=size)
    {
        T[] chunk = new T[Math.Min(size, array.Length - i)];
        Array.Copy(array, i, chunk, 0, chunk.Length);
        yield return chunk;
    }
}

只是说说我的意见。如果你想要“桶”列表(从左到右可视化)，你可以做以下事情:

 public static List<List<T>> Buckets<T>(this List<T> source, int numberOfBuckets)
    {
        List<List<T>> result = new List<List<T>>();
        for (int i = 0; i < numberOfBuckets; i++)
        {
            result.Add(new List<T>());
        }

        int count = 0;
        while (count < source.Count())
        {
            var mod = count % numberOfBuckets;
            result[mod].Add(source[count]);
            count++;
        }
        return result;
    }

一般来说，CaseyB建议的方法很好，事实上，如果你传入一个List<T>，很难对它进行错误处理，也许我会将其更改为:

public static IEnumerable<IEnumerable<T>> ChunkTrivialBetter<T>(this IEnumerable<T> source, int chunksize)
{
   var pos = 0; 
   while (source.Skip(pos).Any())
   {
      yield return source.Skip(pos).Take(chunksize);
      pos += chunksize;
   }
}

这将避免大量的呼叫链。然而，这种方法有一个普遍的缺陷。它实现了每个块的两个枚举，以突出问题，尝试运行:

foreach (var item in Enumerable.Range(1, int.MaxValue).Chunk(8).Skip(100000).First())
{
   Console.WriteLine(item);
}
// wait forever 

为了克服这个问题，我们可以尝试Cameron的方法，该方法出色地通过了上述测试，因为它只遍历了一次枚举。

问题是，它有一个不同的缺陷，它将每个块中的每个项具体化，这种方法的问题是，你会占用大量内存。

为了说明尝试运行:

foreach (var item in Enumerable.Range(1, int.MaxValue)
               .Select(x => x + new string('x', 100000))
               .Clump(10000).Skip(100).First())
{
   Console.Write('.');
}
// OutOfMemoryException

最后，任何实现都应该能够处理块的乱序迭代，例如:

Enumerable.Range(1,3).Chunk(2).Reverse().ToArray()
// should return [3],[1,2]

许多高度最优的解决方案，比如我对这个答案的第一次修正，都失败了。在casperOne的优化答案中也可以看到同样的问题。

要解决所有这些问题，您可以使用以下方法:

namespace ChunkedEnumerator
{
    public static class Extensions 
    {
        class ChunkedEnumerable<T> : IEnumerable<T>
        {
            class ChildEnumerator : IEnumerator<T>
            {
                ChunkedEnumerable<T> parent;
                int position;
                bool done = false;
                T current;


                public ChildEnumerator(ChunkedEnumerable<T> parent)
                {
                    this.parent = parent;
                    position = -1;
                    parent.wrapper.AddRef();
                }

                public T Current
                {
                    get
                    {
                        if (position == -1 || done)
                        {
                            throw new InvalidOperationException();
                        }
                        return current;

                    }
                }

                public void Dispose()
                {
                    if (!done)
                    {
                        done = true;
                        parent.wrapper.RemoveRef();
                    }
                }

                object System.Collections.IEnumerator.Current
                {
                    get { return Current; }
                }

                public bool MoveNext()
                {
                    position++;

                    if (position + 1 > parent.chunkSize)
                    {
                        done = true;
                    }

                    if (!done)
                    {
                        done = !parent.wrapper.Get(position + parent.start, out current);
                    }

                    return !done;

                }

                public void Reset()
                {
                    // per http://msdn.microsoft.com/en-us/library/system.collections.ienumerator.reset.aspx
                    throw new NotSupportedException();
                }
            }

            EnumeratorWrapper<T> wrapper;
            int chunkSize;
            int start;

            public ChunkedEnumerable(EnumeratorWrapper<T> wrapper, int chunkSize, int start)
            {
                this.wrapper = wrapper;
                this.chunkSize = chunkSize;
                this.start = start;
            }

            public IEnumerator<T> GetEnumerator()
            {
                return new ChildEnumerator(this);
            }

            System.Collections.IEnumerator System.Collections.IEnumerable.GetEnumerator()
            {
                return GetEnumerator();
            }

        }

        class EnumeratorWrapper<T>
        {
            public EnumeratorWrapper (IEnumerable<T> source)
            {
                SourceEumerable = source;
            }
            IEnumerable<T> SourceEumerable {get; set;}

            Enumeration currentEnumeration;

            class Enumeration
            {
                public IEnumerator<T> Source { get; set; }
                public int Position { get; set; }
                public bool AtEnd { get; set; }
            }

            public bool Get(int pos, out T item) 
            {

                if (currentEnumeration != null && currentEnumeration.Position > pos)
                {
                    currentEnumeration.Source.Dispose();
                    currentEnumeration = null;
                }

                if (currentEnumeration == null)
                {
                    currentEnumeration = new Enumeration { Position = -1, Source = SourceEumerable.GetEnumerator(), AtEnd = false };
                }

                item = default(T);
                if (currentEnumeration.AtEnd)
                {
                    return false;
                }

                while(currentEnumeration.Position < pos) 
                {
                    currentEnumeration.AtEnd = !currentEnumeration.Source.MoveNext();
                    currentEnumeration.Position++;

                    if (currentEnumeration.AtEnd) 
                    {
                        return false;
                    }

                }

                item = currentEnumeration.Source.Current;

                return true;
            }

            int refs = 0;

            // needed for dispose semantics 
            public void AddRef()
            {
                refs++;
            }

            public void RemoveRef()
            {
                refs--;
                if (refs == 0 && currentEnumeration != null)
                {
                    var copy = currentEnumeration;
                    currentEnumeration = null;
                    copy.Source.Dispose();
                }
            }
        }

        public static IEnumerable<IEnumerable<T>> Chunk<T>(this IEnumerable<T> source, int chunksize)
        {
            if (chunksize < 1) throw new InvalidOperationException();

            var wrapper =  new EnumeratorWrapper<T>(source);

            int currentPos = 0;
            T ignore;
            try
            {
                wrapper.AddRef();
                while (wrapper.Get(currentPos, out ignore))
                {
                    yield return new ChunkedEnumerable<T>(wrapper, chunksize, currentPos);
                    currentPos += chunksize;
                }
            }
            finally
            {
                wrapper.RemoveRef();
            }
        }
    }

    class Program
    {
        static void Main(string[] args)
        {
            int i = 10;
            foreach (var group in Enumerable.Range(1, int.MaxValue).Skip(10000000).Chunk(3))
            {
                foreach (var n in group)
                {
                    Console.Write(n);
                    Console.Write(" ");
                }
                Console.WriteLine();
                if (i-- == 0) break;
            }


            var stuffs = Enumerable.Range(1, 10).Chunk(2).ToArray();

            foreach (var idx in new [] {3,2,1})
            {
                Console.Write("idx " + idx + " ");
                foreach (var n in stuffs[idx])
                {
                    Console.Write(n);
                    Console.Write(" ");
                }
                Console.WriteLine();
            }

            /*

10000001 10000002 10000003
10000004 10000005 10000006
10000007 10000008 10000009
10000010 10000011 10000012
10000013 10000014 10000015
10000016 10000017 10000018
10000019 10000020 10000021
10000022 10000023 10000024
10000025 10000026 10000027
10000028 10000029 10000030
10000031 10000032 10000033
idx 3 7 8
idx 2 5 6
idx 1 3 4
             */

            Console.ReadKey();


        }

    }
}

对于块的无序迭代，您还可以引入一轮优化，这超出了本文的范围。

至于你应该选择哪种方法呢?这完全取决于你要解决的问题。如果你不关心第一个缺陷，那么简单的答案是非常有吸引力的。

注意与大多数方法一样，这对于多线程来说是不安全的，如果你想让它线程安全，你需要修改EnumeratorWrapper。