在c#中合并2个或更多字典(Dictionary<TKey, TValue>)的最佳方法是什么? (像LINQ这样的3.0特性就可以了)。

我正在考虑一个方法签名,如下所示:

public static Dictionary<TKey,TValue>
                 Merge<TKey,TValue>(Dictionary<TKey,TValue>[] dictionaries);

or

public static Dictionary<TKey,TValue>
                 Merge<TKey,TValue>(IEnumerable<Dictionary<TKey,TValue>> dictionaries);

关于重复键的处理:在发生冲突的情况下,保存到字典中的值并不重要,只要它是一致的。


当前回答

对于c#新手来说,我害怕看到复杂的答案。

这里有一些简单的答案。 合并d1 d2,等等。字典和处理任何重叠键(“b”在下面的例子中):

示例1

{
    // 2 dictionaries,  "b" key is common with different values

    var d1 = new Dictionary<string, int>() { { "a", 10 }, { "b", 21 } };
    var d2 = new Dictionary<string, int>() { { "c", 30 }, { "b", 22 } };

    var result1 = d1.Concat(d2).GroupBy(ele => ele.Key).ToDictionary(ele => ele.Key, ele => ele.First().Value);
    // result1 is  a=10, b=21, c=30    That is, took the "b" value of the first dictionary

    var result2 = d1.Concat(d2).GroupBy(ele => ele.Key).ToDictionary(ele => ele.Key, ele => ele.Last().Value);
    // result2 is  a=10, b=22, c=30    That is, took the "b" value of the last dictionary
}

示例2

{
    // 3 dictionaries,  "b" key is common with different values

    var d1 = new Dictionary<string, int>() { { "a", 10 }, { "b", 21 } };
    var d2 = new Dictionary<string, int>() { { "c", 30 }, { "b", 22 } };
    var d3 = new Dictionary<string, int>() { { "d", 40 }, { "b", 23 } };

    var result1 = d1.Concat(d2).Concat(d3).GroupBy(ele => ele.Key).ToDictionary(ele => ele.Key, ele => ele.First().Value);
    // result1 is  a=10, b=21, c=30, d=40    That is, took the "b" value of the first dictionary

    var result2 = d1.Concat(d2).Concat(d3).GroupBy(ele => ele.Key).ToDictionary(ele => ele.Key, ele => ele.Last().Value);
    // result2 is  a=10, b=23, c=30, d=40    That is, took the "b" value of the last dictionary
}

对于更复杂的场景,请参见其他答案。 希望这有帮助。

其他回答

@Tim:应该是注释,但是注释不允许代码编辑。

Dictionary<string, string> t1 = new Dictionary<string, string>();
t1.Add("a", "aaa");
Dictionary<string, string> t2 = new Dictionary<string, string>();
t2.Add("b", "bee");
Dictionary<string, string> t3 = new Dictionary<string, string>();
t3.Add("c", "cee");
t3.Add("d", "dee");
t3.Add("b", "bee");
Dictionary<string, string> merged = t1.MergeLeft(t2, t2, t3);

注意:我应用了@ANeves对@Andrew Orsich的解决方案的修改,所以mergleft现在看起来像这样:

public static Dictionary<K, V> MergeLeft<K, V>(this Dictionary<K, V> me, params IDictionary<K, V>[] others)
    {
        var newMap = new Dictionary<K, V>(me, me.Comparer);
        foreach (IDictionary<K, V> src in
            (new List<IDictionary<K, V>> { me }).Concat(others))
        {
            // ^-- echk. Not quite there type-system.
            foreach (KeyValuePair<K, V> p in src)
            {
                newMap[p.Key] = p.Value;
            }
        }
        return newMap;
    }

基于上面的答案,但添加一个Func-parameter,让调用者处理重复:

public static Dictionary<TKey, TValue> Merge<TKey, TValue>(this IEnumerable<Dictionary<TKey, TValue>> dicts, 
                                                           Func<IGrouping<TKey, TValue>, TValue> resolveDuplicates)
{
    if (resolveDuplicates == null)
        resolveDuplicates = new Func<IGrouping<TKey, TValue>, TValue>(group => group.First());

    return dicts.SelectMany<Dictionary<TKey, TValue>, KeyValuePair<TKey, TValue>>(dict => dict)
                .ToLookup(pair => pair.Key, pair => pair.Value)
                .ToDictionary(group => group.Key, group => resolveDuplicates(group));
}

下面是我使用的一个helper函数:

using System.Collections.Generic;
namespace HelperMethods
{
    public static class MergeDictionaries
    {
        public static void Merge<TKey, TValue>(this IDictionary<TKey, TValue> first, IDictionary<TKey, TValue> second)
        {
            if (second == null || first == null) return;
            foreach (var item in second) 
                if (!first.ContainsKey(item.Key)) 
                    first.Add(item.Key, item.Value);
        }
    }
}

我会这样做:

dictionaryFrom.ToList().ForEach(x => dictionaryTo.Add(x.Key, x.Value));

简单易行。根据这篇博客文章,它甚至比大多数循环更快,因为它的底层实现通过索引而不是枚举来访问元素(参见这个答案)。

如果存在重复,它当然会抛出异常,因此您必须在合并之前进行检查。

考虑到字典键查找和删除的性能,因为它们是哈希操作,并且考虑到问题的措辞是最好的方式,我认为下面是一个完全有效的方法,而其他方法有点过于复杂,恕我冒昧。

    public static void MergeOverwrite<T1, T2>(this IDictionary<T1, T2> dictionary, IDictionary<T1, T2> newElements)
    {
        if (newElements == null) return;

        foreach (var e in newElements)
        {
            dictionary.Remove(e.Key); //or if you don't want to overwrite do (if !.Contains()
            dictionary.Add(e);
        }
    }

或者如果你在多线程应用程序中工作,你的字典无论如何都需要线程安全,你应该这样做:

    public static void MergeOverwrite<T1, T2>(this ConcurrentDictionary<T1, T2> dictionary, IDictionary<T1, T2> newElements)
    {
        if (newElements == null || newElements.Count == 0) return;

        foreach (var ne in newElements)
        {
            dictionary.AddOrUpdate(ne.Key, ne.Value, (key, value) => value);
        }
    }

然后可以对其进行包装,使其处理字典的枚举。无论如何,您看到的是~O(3n)(所有条件都是完美的),因为. add()将在幕后执行额外的、不必要的但实际上是免费的Contains()。我觉得没有比这更好的了。

如果希望限制大型集合上的额外操作,则应该将将要合并的每个字典的Count相加,并将目标字典的容量设置为该值,这样可以避免以后调整大小的成本。最终产品是这样的…

    public static IDictionary<T1, T2> MergeAllOverwrite<T1, T2>(IList<IDictionary<T1, T2>> allDictionaries)
    {
        var initSize = allDictionaries.Sum(d => d.Count);
        var resultDictionary = new Dictionary<T1, T2>(initSize);
        allDictionaries.ForEach(resultDictionary.MergeOverwrite);
        return resultDictionary;
    }

注意,我在这个方法中引入了一个IList<T>…主要是因为如果你接受一个IEnumerable<T>,你已经向同一个集合的多个枚举开放了,如果你从一个延迟的LINQ语句中获得字典集合,这可能是非常昂贵的。