views:

442

answers:

6

Hi,

I'm needing to know the best way to compare two objects and to find out if there equal. I'm overriding both GethashCode and Equals. So a basic class looks like:

public class Test
{
    public int Value { get; set; }
    public string String1 { get; set; }
    public string String2 { get; set; }

    public override int GetHashCode()
    {
        return Value ^ String1.GetHashCode() ^ String2.GetHashCode();
    }

    public override bool Equals( object obj )
    {
        return GetHashCode() == obj.GetHashCode();
    }
}

So for testing purposes I created two objects:

Test t = new Test()
{
    Value = 1,
    String1 ="One",
    String2 = "One"
};

Test t2 = new Test()
{
    Value = 1,
    String1 = "Two",
    String2 = "Two"
};

bool areEqual = t.Equals( t2 );

In testing this areEqual returns true event though both objects are different. I realise this is because String1 and String2 are the same value in each object and thus cancels each other out when hashing.

Is there a better way off hashing object that the method I have that will resolve my issue?

+17  A: 

Your current equality method is broken - there are more values than possible hash codes. It's entirely reasonable (and expected) that you will occasionally have values which are unequal but give the same hash. Equals should check the actual values:

public override bool Equals(object obj)
{
    Test test = obj as Test;
    if (obj == null)
    {
        return false;
    }
    return Value == test.Value &&
        String1 == test.String1 &&
        String2 == test.String2;
}

A few things to note:

  • Your way of generating the hashcode will give the same value for any fixed Value if String1 and String2 are the same; it will also blow up if String1 or String2 is null. This is an unfortunate aspect of using XOR for hashing. I prefer something like this:

    // Put this extension method in a utility class somewhere
    public static int SafeGetHashCode<T>(this T value) where T : class
    {
        return t == null ? 0 : t.GetHashCode();
    }
    
    
    // and this in your actual class
    public override int GetHashCode()
    {
        int hash = 19;
        hash = hash * 31 + Value;
        hash = hash * 31 + String1.SafeGetHashCode();
        hash = hash * 31 + String2.SafeGetHashCode();
        return hash;
    }
    
  • Generally speaking, equality becomes tricky when inheritance gets involved. You may want to consider sealing your class.

  • You may also want to implement IEquatable<Test>

Jon Skeet
Re `SafeGetHashCode` - just to mention that `EqualityComparer<T>.GetHashCode` will do something very similar - the extension method is quite tidy, though ;p
Marc Gravell
@Jon Your answer is really good, whereas the first word is wrong: "No" is the wrong answer to his question:"Is there a better way ..."Moreover Simons Test class does not have a base class nor does it implement a interface. so the override keyword is really wrong
Oops
@Oops: Will fix the "No"... but the Test class *does* have a base class - the implicit base class of System.Object. These methods *are* overriding the Object implementations.
Jon Skeet
+14  A: 

Your Equals is incorrect - that should define what it means for two things to be equal - and having the same hash-code does not mean equality (however; a different hash-code does mean non-equality). If "equality" means "both strings are pairwise equal", then test that.

Re a better hash; xor is notorious for this, since it is trivial to get 0 by xor a value with itself. A better approach may be something like:

int i = 0x65407627;
i = (i * -1521134295) + Value.GetHashCode();
i = (i * -1521134295) + (String1 == null ? 0 : String1.GetHashCode());
i = (i * -1521134295) + (String2 == null ? 0 : String2.GetHashCode());
return i;
Marc Gravell
(updated, borrowing the seed and multiplier that the C# compiler uses for anonymous types)
Marc Gravell
@Marc: Why did you choose those values for generating the hash code?
Carlos Loth
@Carlos: See Marc's previous comment :)
Jon Skeet
@Jon: I've seen if before, but I haven't understood. Thanks for trying to help.
Carlos Loth
@Carlos - I honestly don't know why MS chose those ones *specifically* - the seed+factor approach is common enough, though. The values can be pretty much anything that gives a decent scatter (and the -ve may help with that)
Marc Gravell
A: 

For any two objects, object equality implies hash code equality, however, hash code equality does not imply object equality. From Object.GetHashCode on MSDN:

A hash function must have the following properties:

If two objects compare as equal, the GetHashCode method for each object must return the same value. However, if two objects do not compare as equal, the GetHashCode methods for the two object do not have to return different values.

In other words, your Equals is written wrong. It should be something like:

public override bool Equals(object obj)
{
    Test other = obj as Test;
    if (other == null)
        return false;

    return (Value == other.Value)
        && (String1 == other.String1)
        && (String2 == other.String2);
}

GetHashCode is good for collections (like Dictionary<K, V>) to quickly determine approximate equality. Equals is for comparing if two objects really are the same.

Chris Schmich
+2  A: 

In addition to all the great answers I would recommend that you read this article by Alois Kraus

cyberzed
A: 

wouldn't a function Equals always test only against the same type, shouldn't it be:

//override
    public bool Equals(Test other)//(object obj) 
    {
        //return GetHashCode() == obj.GetHashCode();
        return (Value == other.Value) &&
               (String1 == other.String1) &&
               (String2 == other.String2);
    }

regards Oops

Oops
You *should* override the method from Object. You could *also* implement `IEquatable<Test>`, which is when you'd specify this signature.
Jon Skeet
hmm don't know, is it an issue of the VS2010 beta? If I use override here, VS says: 'Test.Equals(Test)': no suitable method found to override
Oops
+1  A: 

Well!!!!! (MSDN)

lavi
Really wicked formatting to be honest...
cyberzed
Copying multiple pages of MSDN isn't generally useful; it is better to link and highlight any bits of particular significance. But not the whole page.
Marc Gravell