Je sais que c'est une vieille question, mais il y a eu un certain nombre de commentaires sur le fait que toutes les solutions ont jusqu'à présent la gestion du temps vérifie pour assurer le type de données est un ENUM. Voici une solution complète (avec quelques exemples) d'une solution avec des contrôles de temps de compilation (ainsi que des commentaires et des discussions de mes collègues développeurs)
//There is no good way to constrain a generic class parameter to an Enum. The hack below does work at compile time,
// though it is convoluted. For examples of how to use the two classes EnumIndexedArray and ObjEnumIndexedArray,
// see AssetClassArray below. Or, e.g.
// EConstraint.EnumIndexedArray<int, YourEnum> x = new EConstraint.EnumIndexedArray<int, YourEnum>();
// See this post
// http://stackoverflow.com/questions/79126/create-generic-method-constraining-t-to-an-enum/29581813#29581813
// and the answer/comments by Julien Lebosquain
public class EConstraint : HackForCompileTimeConstraintOfTEnumToAnEnum<System.Enum> { }//THIS MUST BE THE ONLY IMPLEMENTATION OF THE ABSTRACT HackForCompileTimeConstraintOfTEnumToAnEnum
public abstract class HackForCompileTimeConstraintOfTEnumToAnEnum<SystemEnum> where SystemEnum : class
{
//For object types T, users should use EnumIndexedObjectArray below.
public class EnumIndexedArray<T, TEnum>
where TEnum : struct, SystemEnum
{
//Needs to be public so that we can easily do things like intIndexedArray.data.sum()
// - just not worth writing up all the equivalent methods, and we can't inherit from T[] and guarantee proper initialization.
//Also, note that we cannot use Length here for initialization, even if Length were defined the same as GetNumEnums up to
// static qualification, because we cannot use a non-static for initialization here.
// Since we want Length to be non-static, in keeping with other definitions of the Length property, we define the separate static
// GetNumEnums, and then define the non-static Length in terms of the actual size of the data array, just for clarity,
// safety and certainty (in case someone does something stupid like resizing data).
public T[] data = new T[GetNumEnums()];
//First, a couple of statics allowing easy use of the enums themselves.
public static TEnum[] GetEnums()
{
return (TEnum[])Enum.GetValues(typeof(TEnum));
}
public TEnum[] getEnums()
{
return GetEnums();
}
//Provide a static method of getting the number of enums. The Length property also returns this, but it is not static and cannot be use in many circumstances.
public static int GetNumEnums()
{
return GetEnums().Length;
}
//This should always return the same as GetNumEnums, but is not static and does it in a way that guarantees consistency with the member array.
public int Length { get { return data.Length; } }
//public int Count { get { return data.Length; } }
public EnumIndexedArray() { }
// [WDS 2015-04-17] Remove. This can be dangerous. Just force people to use EnumIndexedArray(T[] inputArray).
// [DIM 2015-04-18] Actually, if you think about it, EnumIndexedArray(T[] inputArray) is just as dangerous:
// For value types, both are fine. For object types, the latter causes each object in the input array to be referenced twice,
// while the former causes the single object t to be multiply referenced. Two references to each of many is no less dangerous
// than 3 or more references to one. So all of these are dangerous for object types.
// We could remove all these ctors from this base class, and create a separate
// EnumIndexedValueArray<T, TEnum> : EnumIndexedArray<T, TEnum> where T: struct ...
// but then specializing to TEnum = AssetClass would have to be done twice below, once for value types and once
// for object types, with a repetition of all the property definitions. Violating the DRY principle that much
// just to protect against stupid usage, clearly documented as dangerous, is not worth it IMHO.
public EnumIndexedArray(T t)
{
int i = Length;
while (--i >= 0)
{
this[i] = t;
}
}
public EnumIndexedArray(T[] inputArray)
{
if (inputArray.Length > Length)
{
throw new Exception(string.Format("Length of enum-indexed array ({0}) to big. Can't be more than {1}.", inputArray.Length, Length));
}
Array.Copy(inputArray, data, inputArray.Length);
}
public EnumIndexedArray(EnumIndexedArray<T, TEnum> inputArray)
{
Array.Copy(inputArray.data, data, data.Length);
}
//Clean data access
public T this[int ac] { get { return data[ac]; } set { data[ac] = value; } }
public T this[TEnum ac] { get { return data[Convert.ToInt32(ac)]; } set { data[Convert.ToInt32(ac)] = value; } }
}
public class EnumIndexedObjectArray<T, TEnum> : EnumIndexedArray<T, TEnum>
where TEnum : struct, SystemEnum
where T : new()
{
public EnumIndexedObjectArray(bool doInitializeWithNewObjects = true)
{
if (doInitializeWithNewObjects)
{
for (int i = Length; i > 0; this[--i] = new T()) ;
}
}
// The other ctor's are dangerous for object arrays
}
public class EnumIndexedArrayComparator<T, TEnum> : EqualityComparer<EnumIndexedArray<T, TEnum>>
where TEnum : struct, SystemEnum
{
private readonly EqualityComparer<T> elementComparer = EqualityComparer<T>.Default;
public override bool Equals(EnumIndexedArray<T, TEnum> lhs, EnumIndexedArray<T, TEnum> rhs)
{
if (lhs == rhs)
return true;
if (lhs == null || rhs == null)
return false;
//These cases should not be possible because of the way these classes are constructed.
// HOWEVER, the data member is public, so somebody _could_ do something stupid and make
// data=null, or make lhs.data == rhs.data, even though lhs!=rhs (above check)
//On the other hand, these are just optimizations, so it won't be an issue if we reomve them anyway,
// Unless someone does something really dumb like setting .data to null or resizing to an incorrect size,
// in which case things will crash, but any developer who does this deserves to have it crash painfully...
//if (lhs.data == rhs.data)
// return true;
//if (lhs.data == null || rhs.data == null)
// return false;
int i = lhs.Length;
//if (rhs.Length != i)
// return false;
while (--i >= 0)
{
if (!elementComparer.Equals(lhs[i], rhs[i]))
return false;
}
return true;
}
public override int GetHashCode(EnumIndexedArray<T, TEnum> enumIndexedArray)
{
//This doesn't work: for two arrays ar1 and ar2, ar1.GetHashCode() != ar2.GetHashCode() even when ar1[i]==ar2[i] for all i (unless of course they are the exact same array object)
//return engineArray.GetHashCode();
//Code taken from comment by Jon Skeet - of course - in http://stackoverflow.com/questions/7244699/gethashcode-on-byte-array
//31 and 17 are used commonly elsewhere, but maybe because everyone is using Skeet's post.
//On the other hand, this is really not very critical.
unchecked
{
int hash = 17;
int i = enumIndexedArray.Length;
while (--i >= 0)
{
hash = hash * 31 + elementComparer.GetHashCode(enumIndexedArray[i]);
}
return hash;
}
}
}
}
//Because of the above hack, this fails at compile time - as it should. It would, otherwise, only fail at run time.
//public class ThisShouldNotCompile : EConstraint.EnumIndexedArray<int, bool>
//{
//}
//An example
public enum AssetClass { Ir, FxFwd, Cm, Eq, FxOpt, Cr };
public class AssetClassArrayComparator<T> : EConstraint.EnumIndexedArrayComparator<T, AssetClass> { }
public class AssetClassIndexedArray<T> : EConstraint.EnumIndexedArray<T, AssetClass>
{
public AssetClassIndexedArray()
{
}
public AssetClassIndexedArray(T t) : base(t)
{
}
public AssetClassIndexedArray(T[] inputArray) : base(inputArray)
{
}
public AssetClassIndexedArray(EConstraint.EnumIndexedArray<T, AssetClass> inputArray) : base(inputArray)
{
}
public T Cm { get { return this[AssetClass.Cm ]; } set { this[AssetClass.Cm ] = value; } }
public T FxFwd { get { return this[AssetClass.FxFwd]; } set { this[AssetClass.FxFwd] = value; } }
public T Ir { get { return this[AssetClass.Ir ]; } set { this[AssetClass.Ir ] = value; } }
public T Eq { get { return this[AssetClass.Eq ]; } set { this[AssetClass.Eq ] = value; } }
public T FxOpt { get { return this[AssetClass.FxOpt]; } set { this[AssetClass.FxOpt] = value; } }
public T Cr { get { return this[AssetClass.Cr ]; } set { this[AssetClass.Cr ] = value; } }
}
//Inherit from AssetClassArray<T>, not EnumIndexedObjectArray<T, AssetClass>, so we get the benefit of the public access getters and setters above
public class AssetClassIndexedObjectArray<T> : AssetClassIndexedArray<T> where T : new()
{
public AssetClassIndexedObjectArray(bool bInitializeWithNewObjects = true)
{
if (bInitializeWithNewObjects)
{
for (int i = Length; i > 0; this[--i] = new T()) ;
}
}
}
commentaire minuscule au sujet de votre nom « DaysOfTheWeek »: La norme C# dit que les énumérations de type non-flags devraient avoir des noms singuliers et les énumérations de style flags devraient avoir plusieurs noms, donc "DayOfTheWeek" serait meilleur. http://msdn.microsoft.com/en-us/library/ms229040.aspx – RenniePet