C# Sortable collection which allows duplicate keys

问题:

I am writing a program to set a sequence in which various objects will appear in report. The sequence is the Y position (cell) on Excel spreadsheet.

A demo part of code is below. What I want to accomplish is to have a collection, which will allow me to add multiple objects and I can get a sorted collection based on the sequence

SortedList list = new SortedList();

Header h = new Header();
h.XPos = 1;
h.name = "Header_1";
list.Add(h.XPos, h);

h = new Header();
h.XPos = 1;
h.name = "Header_2";
list.Add(h.XPos, h);

I know that the SortedList will not allow this and I have been searching for alternate. I don't want to eliminate the duplicates and already tried List<KeyValuePair<int, object>>.

Thanks.

回答1:

Use your own IComparer!

Like already stated in some other answers, you should use your own comparer class. For this sake I use a generic IComparer class, that works with anything that implements IComparable:

/// <summary>
/// Comparer for comparing two keys, handling equality as beeing greater
/// Use this Comparer e.g. with SortedLists or SortedDictionaries, that don't allow duplicate keys
/// </summary>
/// <typeparam name="TKey"></typeparam>
public class DuplicateKeyComparer<TKey>
                :
             IComparer<TKey> where TKey : IComparable
{
    #region IComparer<TKey> Members

    public int Compare(TKey x, TKey y)
    {
        int result = x.CompareTo(y);

        if (result == 0)
            return 1;   // Handle equality as beeing greater
        else
            return result;
    }

    #endregion
}

You will use it when instancing a new SortedList, SortedDictionary etc:

SortedList<int, MyValueClass> slist = new SortedList<int, MyValueClass>(new DuplicateKeyComparer<int>());

Here int is the key that can be duplicate.

回答2:

You can safely use List<> . The List has a Sort method , an overload of which accepts IComparer. You can create your own sorter class as . Here's an example :

private List<Curve> Curves;
this.Curves.Sort(new CurveSorter());

public class CurveSorter : IComparer<Curve>
{
    public int Compare(Curve c1, Curve c2)
    {
        return c2.CreationTime.CompareTo(c1.CreationTime);
    }
}

回答3:

Simplest solution (compared to all of the above): use SortedSet<T>, it accepts an IComparer<SortableKey> class, then implement the Compare method this way:

public int Compare(SomeClass x, SomeClass y)
{
    var compared = x.SomeSortableKeyTypeField.CompareTo(y.SomeSortableKeyTypeField);
    if (compared != 0)
        return compared;

    // to allow duplicates
    var hashCodeCompare = x.GetHashCode().CompareTo(y.GetHashCode());
    if (hashCodeCompare != 0)
        return hashCodeCompare;

    if (Object.ReferenceEquals(x, y))
        return 0;

    // for weird duplicate hashcode cases, throw as below or implement your last chance comparer
    throw new ComparisonFailureException();

}

回答4:

I use the following:

public class TupleList<T1, T2> : List<Tuple<T1, T2>> where T1 : IComparable
{
    public void Add(T1 item, T2 item2)
    {
        Add(new Tuple<T1, T2>(item, item2));
    }

    public new void Sort()
    {
        Comparison<Tuple<T1, T2>> c = (a, b) => a.Item1.CompareTo(b.Item1);
        base.Sort(c);
    }

}

My test case:

[TestMethod()]
    public void SortTest()
    {
        TupleList<int, string> list = new TupleList<int, string>();
        list.Add(1, "cat");
        list.Add(1, "car");
        list.Add(2, "dog");
        list.Add(2, "door");
        list.Add(3, "elephant");
        list.Add(1, "coconut");
        list.Add(1, "cab");
        list.Sort();
        foreach(Tuple<int, string> tuple in list)
        {
            Console.WriteLine(string.Format("{0}:{1}", tuple.Item1,tuple.Item2));
        }
        int expected_first = 1;
        int expected_last = 3;
        int first = list.First().Item1;  //requires using System.Linq
        int last = list.Last().Item1;    //requires using System.Linq
        Assert.AreEqual(expected_first, first);
        Assert.AreEqual(expected_last, last);
    }

The output:

1:cab
1:coconut
1:car
1:cat
2:door
2:dog
3:elephant

回答5:

The problem is that the data structure design doesn't match the requirements: It is necessary to store several Headers for the same XPos. Therefore, SortedList<XPos, value> should not have a value of Header, but a value of List<Header>. It's a simple and small change, but it solves all problems and avoids creating new problems like other suggested solutions (see explanation below):

using System;
using System.Collections.Generic;

namespace TrySortedList {
  class Program {

    class Header {
      public int XPos;
      public string Name;
    }

    static void Main(string[] args) {
      SortedList<int, List<Header>> sortedHeaders = new SortedList<int,List<Header>>();
      add(sortedHeaders, 1, "Header_1");
      add(sortedHeaders, 1, "Header_2");
      add(sortedHeaders, 2, "Header_3");
      foreach (var headersKvp in sortedHeaders) {
        foreach (Header header in headersKvp.Value) {
          Console.WriteLine(header.XPos + ": " + header.Name);
        }
      }
    }

    private static void add(SortedList<int, List<Header>> sortedHeaders, int xPos, string name) {
      List<Header> headers;
      if (!sortedHeaders.TryGetValue(xPos, out headers)){
        headers = new List<Header>();
        sortedHeaders[xPos] = headers;
      }
      headers.Add(new Header { XPos = xPos, Name = name });
    }
  }
}

Output:
1: Header_1
1: Header_2
2: Header_3

Please note that adding a "funny" key, like adding a random number or pretending that 2 XPos with the same value are different lead to many other problems. For example it becomes difficult or even impossible to remove a particular Header.

Also note that the sorting performance is much better if only few List<Header> have to be sorted than every Header. Example: If there are 100 XPos and each has 100 headers, 10000 Header need to be sorted as opposed to 100 List<Header>.

Of course, also this solution has a disadvantage: If there are many XPos with only 1 Header, as many Lists need to be created, which is some overhead.

回答6:

Thanks a lot for your help. While searching more, I found this solution. (Available in Stackoverflow.com in other question)

First, I created a class which would encapsulate my objects for classes (Headers,Footer etc)

public class MyPosition
{
    public int Position { get; set; }
    public object MyObjects{ get; set; }
}

So this class is supposed to hold on the objects, and PosX of each object goes as int Position

List<MyPosition> Sequence= new List<MyPosition>();
Sequence.Add(new MyPosition() { Position = 1, Headerobject });
Sequence.Add(new MyPosition() { Position = 2, Headerobject1 });
Sequence.Add(new MyPosition() { Position = 1, Footer });

League.Sort((PosA, PosB) => PosA.Position.CompareTo(PosB.Position));

What eventually I get is the sorted "Sequence" list.

回答7:

This collection class will maintain duplicates and insert sort order for the duplicate. The trick is to tag the items with a unique value as they are inserted to maintain a stable sort order. Then we wrap it all up in an ICollection interface.

public class SuperSortedSet<TValue> : ICollection<TValue>
{
    private readonly SortedSet<Indexed<TValue>> _Container;
    private int _Index = 0;
    private IComparer<TValue> _Comparer;

    public SuperSortedSet(IComparer<TValue> comparer)
    {
        _Comparer = comparer;
        var c2 = new System.Linq.Comparer<Indexed<TValue>>((p0, p1) =>
        {
            var r = _Comparer.Compare(p0.Value, p1.Value);
            if (r == 0)
            {
                if (p0.Index == -1
                    || p1.Index == -1)
                    return 0;

                return p0.Index.CompareTo(p1.Index);

            }
            else return r;
        });
        _Container = new SortedSet<Indexed<TValue>>(c2);
    } 

    public IEnumerator<TValue> GetEnumerator() { return _Container.Select(p => p.Value).GetEnumerator(); }

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

    public void Add(TValue item) { _Container.Add(Indexed.Create(_Index++, item)); }

    public void Clear() { _Container.Clear();}

    public bool Contains(TValue item) { return _Container.Contains(Indexed.Create(-1,item)); }

    public void CopyTo(TValue[] array, int arrayIndex)
    {
        foreach (var value in this)
        {
            if (arrayIndex >= array.Length)
            {
                throw new ArgumentException("Not enough space in array");
            }
            array[arrayIndex] = value;
            arrayIndex++;
        }
    }

    public bool Remove(TValue item) { return _Container.Remove(Indexed.Create(-1, item)); }

    public int Count {
        get { return _Container.Count; }
    }
    public bool IsReadOnly {
        get { return false; }
    }
}

a test class

[Fact]
public void ShouldWorkWithSuperSortedSet()
{
    // Sort points according to X
    var set = new SuperSortedSet<Point2D>
        (new System.Linq.Comparer<Point2D>((p0, p1) => p0.X.CompareTo(p1.X)));

    set.Add(new Point2D(9,10));
    set.Add(new Point2D(1,25));
    set.Add(new Point2D(11,-10));
    set.Add(new Point2D(2,99));
    set.Add(new Point2D(5,55));
    set.Add(new Point2D(5,23));
    set.Add(new Point2D(11,11));
    set.Add(new Point2D(21,12));
    set.Add(new Point2D(-1,76));
    set.Add(new Point2D(16,21));

    var xs = set.Select(p=>p.X).ToList();
    xs.Should().BeInAscendingOrder();
    xs.Count.Should()
       .Be(10);
    xs.ShouldBeEquivalentTo(new[]{-1,1,2,5,5,9,11,11,16,21});

    set.Remove(new Point2D(5,55));
    xs = set.Select(p=>p.X).ToList();
    xs.Count.Should()
       .Be(9);
    xs.ShouldBeEquivalentTo(new[]{-1,1,2,5,9,11,11,16,21});

    set.Remove(new Point2D(5,23));
    xs = set.Select(p=>p.X).ToList();
    xs.Count.Should()
       .Be(8);
    xs.ShouldBeEquivalentTo(new[]{-1,1,2,9,11,11,16,21});

    set.Contains(new Point2D(11, 11))
       .Should()
       .BeTrue();

    set.Contains(new Point2D(-1, 76))
        .Should().BeTrue();

    // Note that the custom compartor function ignores the Y value
    set.Contains(new Point2D(-1, 66))
        .Should().BeTrue();

    set.Contains(new Point2D(27, 66))
        .Should().BeFalse();

}

The tagging struct

public struct Indexed<T>
{
    public int Index { get; private set; }
    public T Value { get; private set; }
    public Indexed(int index, T value) : this()
    {
        Index = index;
        Value = value;
    }

    public override string ToString()
    {
        return "(Indexed: " + Index + ", " + Value.ToString () + " )";
    }
}

public class Indexed
{
    public static Indexed<T> Create<T>(int indexed, T value)
    {
        return new Indexed<T>(indexed, value);
    }
}

The lambda comparer helper

public class Comparer<T> : IComparer<T>
{
    private readonly Func<T, T, int> _comparer;

    public Comparer(Func<T, T, int> comparer)
    {
        if (comparer == null)
            throw new ArgumentNullException("comparer");
        _comparer = comparer;
    }

    public int Compare(T x, T y)
    {
        return _comparer(x, y);
    }
}

回答8:

Did you try Lookup<TKey, TElement> that will allow duplicate keys http://msdn.microsoft.com/en-us/library/bb460184.aspx

回答9:

The problem is that you use something as key that isn't a key (cause it occurs multiple times).

So if you have real coordinates you should maybe take the Point as the key for your SortedList.

Or you create a List<List<Header>> where your first list index defines the x-position and the inner list index the y-position (or vice versa if you like).

回答10:

Create a class and query the list:

Public Class SortingAlgorithm
{
    public int ID {get; set;}
    public string name {get; set;}
    public string address1 {get; set;}
    public string city {get; set;}
    public string state {get; set;}
    public int age {get; set;}
}

//declare a sorting algorithm list
List<SortingAlgorithm> sortAlg = new List<SortingAlgorithm>();

//Add multiple values to the list
sortAlg.Add( new SortingAlgorithm() {ID = ID, name = name, address1 = address1, city = city, state = state, age = age});
sortAlg.Add( new SortingAlgorithm() {ID = ID, name = name, address1 = address1, city = city, state = state, age = age});
sortAlg.Add( new SortingAlgorithm() {ID = ID, name = name, address1 = address1, city = city, state = state, age = age});

//query and order by the list
  var sortedlist = (from s in sortAlg
                    select new { s.ID, s.name, s.address1, s.city, s.state, s.age })
                                                     .OrderBy(r => r.ID)
                                                     .ThenBy(r=> r.name)
                                                     .ThenBy(r=> r.city)
                                                     .ThenBy(r=>r.state)
                                                     .ThenBy(r=>r.age);

回答11:

Linq.Lookup is cool and all, but if your target is to simply loop over the "keys" while allowing them to be duplicated you can use this structure:

List<KeyValuePair<String, String>> FieldPatterns = new List<KeyValuePair<string, string>>() {
   new KeyValuePair<String,String>("Address","CommonString"),
   new KeyValuePair<String,String>("Username","UsernamePattern"),
   new KeyValuePair<String,String>("Username","CommonString"),
};

Then you can write:

foreach (KeyValuePair<String,String> item in FieldPatterns)
{
   //use item.Key and item.Value
}

HTH

回答12:

The key (pun intended) to this is to create an IComparable-based class that maintains equality and hashing, but never compares to 0 if not equal. This can be done, and can be created with a couple bonuses - stable sorting (that is, values added to the sorted list first will maintain their position), and ToString() can simply return the actual key string value.

Here's a struct key that should do the trick:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Threading;

namespace System
{
    /// <summary>
    /// Defined in Totlsoft.Util.
    /// A key that will always be unique but compares
    /// primarily on the Key property, which is not required
    /// to be unique.
    /// </summary>
    public struct StableKey : IComparable<StableKey>, IComparable
    {
        private static long s_Next;
        private long m_Sequence;
        private IComparable m_Key;

        /// <summary>
        /// Defined in Totlsoft.Util.
        /// Constructs a StableKey with the given IComparable key.
        /// </summary>
        /// <param name="key"></param>
        public StableKey( IComparable key )
        {
            if( null == key )
                throw new ArgumentNullException( "key" );

            m_Sequence = Interlocked.Increment( ref s_Next );
            m_Key = key;
        }

        /// <summary>
        /// Overridden. True only if internal sequence and the
        /// Key are equal.
        /// </summary>
        /// <param name="obj"></param>
        /// <returns></returns>
        public override bool Equals( object obj )
        {
            if( !( obj is StableKey ) )
                return false;

            var dk = (StableKey)obj;

            return m_Sequence.Equals( dk.m_Sequence ) &&
                Key.Equals( dk.Key );
        }

        /// <summary>
        /// Overridden. Gets the hash code of the internal
        /// sequence and the Key.
        /// </summary>
        /// <returns></returns>
        public override int GetHashCode()
        {
            return m_Sequence.GetHashCode() ^ Key.GetHashCode();
        }

        /// <summary>
        /// Overridden. Returns Key.ToString().
        /// </summary>
        /// <returns></returns>
        public override string ToString()
        {
            return Key.ToString();
        }

        /// <summary>
        /// The key that will be compared on.
        /// </summary>
        public IComparable Key
        {
            get
            {
                if( null == m_Key )
                    return 0;

                return m_Key;
            }
        }

        #region IComparable<StableKey> Members

        /// <summary>
        /// Compares this Key property to another. If they
        /// are the same, compares the incremented value.
        /// </summary>
        /// <param name="other"></param>
        /// <returns></returns>
        public int CompareTo( StableKey other )
        {
            var cmp = Key.CompareTo( other.Key );
            if( cmp == 0 )
                cmp = m_Sequence.CompareTo( other.m_Sequence );

            return cmp;
        }

        #endregion

        #region IComparable Members

        int IComparable.CompareTo( object obj )
        {
            return CompareTo( (StableKey)obj );
        }

        #endregion
    }
}

回答13:

You can use the SortedList, use your value for the TKey, and int (count) for the TValue.

Here's a sample: A function that sorts the letters of a word.

    private string sortLetters(string word)
    {
        var input = new System.Collections.Generic.SortedList<char, int>();

        foreach (var c in word.ToCharArray())
        {
            if (input.ContainsKey(c))
                input[c]++;
            else
                input.Add(c, 1);
        }

        var output = new StringBuilder();

        foreach (var kvp in input)
        {
            output.Append(kvp.Key, kvp.Value);
        }

        string s;

        return output.ToString();

    }

回答14:

The trick is to augment your object with a unique key. See the following test which passes. I want to keep my points sorted by their X value. Just using a naked Point2D in my comparison function will cause points with the same X value to be eliminated. So I wrap the Point2D in a tagging class called Indexed.

[Fact]
public void ShouldBeAbleToUseCustomComparatorWithSortedSet()
{
    // Create comparer that compares on X value but when X
    // X values are uses the index
    var comparer = new 
        System.Linq.Comparer<Indexed<Point2D>>(( p0, p1 ) =>
        {
            var r = p0.Value.X.CompareTo(p1.Value.X);
            return r == 0 ? p0.Index.CompareTo(p1.Index) : r;
        });

    // Sort points according to X
    var set = new SortedSet<Indexed<Point2D>>(comparer);

    int i=0;

    // Create a helper function to wrap each point in a unique index
    Action<Point2D> index = p =>
    {
        var ip = Indexed.Create(i++, p);
        set.Add(ip);
    };

    index(new Point2D(9,10));
    index(new Point2D(1,25));
    index(new Point2D(11,-10));
    index(new Point2D(2,99));
    index(new Point2D(5,55));
    index(new Point2D(5,23));
    index(new Point2D(11,11));
    index(new Point2D(21,12));
    index(new Point2D(-1,76));
    index(new Point2D(16,21));
    set.Count.Should()
       .Be(10);
    var xs = set.Select(p=>p.Value.X).ToList();
    xs.Should()
      .BeInAscendingOrder();
    xs.ShouldBeEquivalentTo(new[]{-1,1,2,5,5,9,11,11,16,21});

}

Utilities to make this work are

A comparer that takes a lambda

public class Comparer<T> : IComparer<T>
{
    private readonly Func<T, T, int> _comparer;

    public Comparer(Func<T, T, int> comparer)
    {
        if (comparer == null)
            throw new ArgumentNullException("comparer");
        _comparer = comparer;
    }

    public int Compare(T x, T y)
    {
        return _comparer(x, y);
    }
}

A tagging struct

public struct Indexed<T>
{
    public int Index { get; private set; }
    public T Value { get; private set; }
    public Indexed(int index, T value) : this()
    {
        Index = index;
        Value = value;
    }

    public override string ToString()
    {
        return "(Indexed: " + Index + ", " + Value.ToString () + " )";
    }
}

public class Indexed
{
    public static Indexed<T> Create<T>(int indexed, T value)
    {
        return new Indexed<T>(indexed, value);
    }
}