All posts in Clone

In A different take on Deep Copy I talked a bit about the different approaches to how you can accomplish performing a Deep Copy of a model in C#. In this article I want to revisit the topic briefly and demonstrate two methods to handle this concern.

Q: Why not use Object.MemberwiseClone?
A: You can, but it only performs a shallow copy of your model. If you have a more complex model, you will need something that can dig deeper and map all your properties.

The first approach is our Deep Copy method done with serialization. If you have been around the Internet a bit, you will recognize it:


/// <summary>
/// Performs a Deep Copy of an object through serialization.
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="oSource">The o source.</param>
/// <returns></returns>
public static T DeepCopy<T>(this T oSource)
{
    using (var ms = new MemoryStream())
    {
        var formatter = new BinaryFormatter();
        formatter.Serialize(ms, oSource);
        ms.Position = 0;
        return (T)formatter.Deserialize(ms);
    }
}

This is the cleanest way to copy a model.

1) A MemoryStream is created.
2) A BinaryFormatter serializes the source into the MemoryStream.
3) The MemoryStream position is reset for reading purposes.
4) A deserialized object is cast to type T and returned.

A simple example:


var myNewObject = myObject.DeepCopy();

 
This works great in many cases but not in a Framework like Silverlight. Given, not a lot of us use that anymore, but you could still run into a scenario where serialization is not possible. In that case, we are going to create a new instance of the model we intend to copy. Then, we will map properties between the 2 models.


/// <summary>
/// Copies all public properties from one class to another.
/// </summary>
/// <param name="source">The source.</param>
/// <param name="destination">The destination.</param>
/// <exception cref="System.Exception">Source and destination cannot be null and must be 
/// of the same type!</exception>
public static void MapProperties<T>(this T source, T destination) where T : class
{
    // Source and destination must exist.
    if ((source == null || destination == null)) return;

    // Get properties
    var propertyInfos = source.GetType().GetProperties();
    if (!propertyInfos.Any()) return;

    // Process only public properties
    foreach (var propInfo in propertyInfos.Where(x => x.CanWrite))
    {
        // Get value from source and assign to destination.
        var value = propInfo.GetValue(source, null);
        if (value == null) continue;

        // Evaluate
        var valType = value.GetType();

        // Collections
        if (valType.InheritsFrom<ICollection>())
        {
            var sourceCollection = value as IList;
            if (sourceCollection == null) continue;

            // Create new instance of collection
            IList destinationCollection = null;
            destinationCollection = (valType.BaseType == typeof(Array))
                ? Array.CreateInstance(valType.GetElementType(), sourceCollection.Count)
                : (IList)Activator.CreateInstance(valType, null);
            if (destinationCollection == null) continue;

            // Map properties
            foreach (var item in sourceCollection)
            {
                // New item instance
                var newItem = HasDefaultConstructor(item.GetType()) 
                    ? Activator.CreateInstance(item.GetType(), null) 
                    : item;

                // Map properties
                item.MapProperties(newItem);

                // Add to destination collection
                if (valType.BaseType == typeof(Array))
                {
                    destinationCollection[sourceCollection.IndexOf(item)] = newItem;
                }
                else
                {
                    destinationCollection.Add(newItem);
                }
            }

            // Add new collection to destination
            propInfo.SetValue(destination, destinationCollection, null);
        }
        else
        {
            propInfo.SetValue(destination, value, null);
        }

        // Check for properties and propagate if they exist.
        var newPropInfos = value.GetType().GetProperties();
        if (!newPropInfos.Any()) continue;

        // Copy properties for each child where necessary.
        var childSource = source.GetType().GetProperty(propInfo.Name);
        var childDestination = destination.GetType().GetProperty(propInfo.Name);
        childSource.MapProperties(childDestination);
    }
}

/// <summary>
/// Determines whether the type has a default contructor.
/// </summary>
/// <param name="type">The type.</param>
/// <returns></returns>
private static bool HasDefaultConstructor(Type type)
{
    return
        type.GetConstructor(Type.EmptyTypes) != null ||
        type.GetConstructors(BindingFlags.Instance | BindingFlags.Public)
            .Any(x => x.GetParameters().All(p => p.IsOptional));
}

Here is what is going on:

1) We get the properties associated with the class type.
2) We evaluate the public properties.
3) Using reflection, we retrieve the value from the source and apply it to the destination.
4) If the value is not null, we drill down further for more public properties recursively.
Note: Collections need a bit of extra work to ensure we don’t simply copy over their object instances.
5) Repeat until the entire model has been traversed.

A simple example:


SomeType myNewObject = new SomeType();

// Assumption: "myObject" is of type "SomeType"
myObject.MapProperties(myNewObject);

In conclusion: if you can’t serialize, map the properties between the 2 models.

That’s all for now.

In Effective Extensions – Extending Enums in C# we covered how to use a few useful methods to extend Enums in C#. This time we are going to look at Extending Models in C#.

UPDATED 8/29/2014: Added a few more GetPropertyName overloads as well as GetPropertyNames.

Our first method is our initial GetPropertyName which retrieves a property name from a simple expression:


/// <summary>
/// Gets the property name from an expression.
/// </summary>
/// <param name="expression">The expression to evaluate.</param>
/// <returns></returns>
public static string GetPropertyName(this Expression expression)
{
    // Get property name
    switch (expression.NodeType)
    {
        case ExpressionType.Call:
            var callExp = expression as MethodCallExpression;
            if (callExp == null || !callExp.Arguments.Any()) return "";

            // Build body on first argument
            var callBody = callExp.Arguments.First() as MemberExpression;

            // Get property name
            return callBody == null ? "" : callBody.Member.Name;
            break;

        case ExpressionType.AndAlso:
        case ExpressionType.OrElse:
            var andOrExp = expression as BinaryExpression;
            if (andOrExp == null) return "";

            // Build body on last argument
            var andOrBody = andOrExp.Right as MemberExpression;

            // Get property name
            return andOrBody == null ? "" : andOrBody.Member.Name;
            break;

        default:
            // Member
            var body = expression as MemberExpression;

            // Body found
            if (body != null) return body.Member.Name;

            // Get body
            var ubody = (UnaryExpression)expression;
            body = ubody.Operand as MemberExpression;
            return (body != null) ? body.Member.Name : "";
            break;
    }
}

The goal of GetPropertyName is to find the relevant MemberExpression and get the property name associated with it. This is not always a simple task and different Expression types have to be evaluated in unique ways.

Let’s break this down into each case:

  1. The first block deals with a MethodCallExpression. You typically see these with expressions like this:
    x => x.MyCollection.Contains(5). We need to mine this down to the first argument to get the MemberExpression.
  2. The second and third blocks deal with complex expressions using && (AndAlso) or || (OrElse) between 2 or more statements. An example might be like this: x => x.IsActive && x.MyValue == 5. The Right value will hold the right-most, singular value as a MemberExpression, where the Left value will hold the expression less the Right. So, to properly take an expression like this apart, you would need to iterate through until Left is just a MemberExpression.
  3. The last block is a simple MemberExpression evaluation. When all else fails, go this route. Sometimes, casting our expression as a MemberExpression won’t work, so we attempt to cast it as a UnaryExpression and cast it’s Operand property as a MemberExpression.

 

Now, let’s look at some GetPropertyName overloads that will use our initial method to evaluate complex expressions:


/// <summary>
/// Gets the property name from an expression.
/// Example: this would be used for expressions in the format: "() => Name"
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="expression">The expression to evaluate.</param>
/// <returns></returns>
public static string GetPropertyName<T>(this Expression<Func<T>> expression)
{
    return expression.Body.GetPropertyName();
}

/// <summary>
/// Gets the property name from an expression.
/// Example: this would be used for expressions in the format: "() => IsActive"
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="expression">The expression.</param>
/// <returns></returns>
public static string GetPropertyName<T>(this Expression<Func<T, bool>> expression)
{
    return expression.Body.GetPropertyName();
}

/// <summary>
/// Gets the property name from an expression.
/// Example: this would be used for expressions in the format: "x => x.Name"
/// </summary>
/// <typeparam name="T"></typeparam>
/// <typeparam name="TProperty">The type of the property.</typeparam>
/// <param name="expression">The exp.</param>
/// <returns></returns>
public static string GetPropertyName<T, TProperty>(this Expression<Func<T, TProperty>> expression)
{
    return expression.Body.GetPropertyName();
}

  • The first method retrieves a property name from a simple expression such as: () => Name.
  • The second method retrieves a property name from a boolean expression such as: () => IsActive.
  • The third method retrieves a property name from a property-based expression such as: x => x.Name.

 

Q: Okay, so how do we get all the property names of a multi-statement expression that uses && or ||?
A: In order to do that, we need to do a little more work and return an array.

Let’s look at our follow-up method GetPropertyNames:


/// <summary>
/// Gets multiple property names from a complex expression.
/// Example: this would be used for expressions in the format: "x => x.Name &amp;&amp; x.IsActive"
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="expression">The expression.</param>
/// <returns></returns>
public static IList<string> GetPropertyNames<T>(this Expression<Func<T, bool>> expression)
{
    var names = new List<string>();

    // Body
    var expBody = expression.Body;

    // Get property name
    if (expBody is BinaryExpression)
    {
        // If a complex expression, we need to disassemble it from Right to Left,
        // Or, we simply need to get the property from the Left.
        if (expBody.IsComplexBinaryExpression())
        {
            names.AddRange(expBody.GetPropertiesFromBinary());
        }
        else
        {
            names.Add(expBody.GetPropertyNameFromSimpleBinaryExpression());
        }
    }

    else
    {
        // Get a singular property name
        names.Add(expression.GetPropertyName());
    }

    // Return unique list
    return names.GroupBy(x => x).Select(y => y.Key).ToList();
}

/// <summary>
/// Gets the properties from binary.
/// </summary>
/// <param name="expression">The expression.</param>
/// <returns></returns>
private static IEnumerable<string> GetPropertiesFromBinary(this Expression expression)
{
    var names = new List<string>();

    var currentExp = expression;
    while (currentExp is BinaryExpression)
    {
        // Cast expression
        var exp = (currentExp as BinaryExpression);

        // Body
        if (exp.IsComplexBinaryExpression())
        {
            // If "Right" is a BinaryExpression, get it's components, or
            // Just get the property name of the "Right" expression.
            var right = exp.Right as BinaryExpression;
            if (right != null)
            {
                names.AddRange(right.GetPropertiesFromBinary());
            }
            else
            {
                names.Add(exp.Right.GetPropertyName());
            }
        }

        else
        {
            // Add to list
            names.Add(exp.GetPropertyNameFromSimpleBinaryExpression());
        }

        // Next expression
        currentExp = exp.Left;
    }

    // Get last expression
    names.Add(currentExp.GetPropertyName());

    // Return
    return names;
}

/// <summary>
/// Determines whether [is complex binary expression] [the specified exp].
/// </summary>
/// <param name="exp">The exp.</param>
/// <returns></returns>
private static bool IsComplexBinaryExpression(this Expression exp)
{
    var expBody = exp as BinaryExpression;
    if (expBody == null) return false;

    // Evaluate
    return (expBody.NodeType == ExpressionType.AndAlso || expBody.NodeType == ExpressionType.OrElse);
}

/// <summary>
/// Gets the property name from simple binary expression.
/// </summary>
/// <param name="exp">The exp.</param>
/// <returns></returns>
private static string GetPropertyNameFromSimpleBinaryExpression(this Expression exp)
{
    var expBody = exp as BinaryExpression;
    if (expBody == null) return "";

    // Body
    var body = expBody.Left as MemberExpression;

    // Get property name
    return (body != null) ? body.Member.Name : "";
}

Here, we are doing what we discussed earlier, which is iterating through our expression, storing the Right value on each pass until the Left is just a MemberExpression. The result is a list of all properties involved in the expression.
The next method we will evaluate is DeepCopy which completely clones a model through serialization:


/// <summary>
/// Deeps the copy.
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="oSource">The o source.</param>
/// <returns></returns>
public static T DeepCopy<T>(this T oSource)
{
    using (var ms = new MemoryStream())
    {
        var formatter = new BinaryFormatter();
        formatter.Serialize(ms, oSource);
        ms.Position = 0;
        return (T)formatter.Deserialize(ms);
    }
}

As you can see, this serializes the source with a BinaryFormatter and injects it into a MemoryStream. This is considered the most effective way to clone an object.

 

Finally, we have an IsNull method that performs a Deep Null check, which means it inspects each element in an object tree to determine if it is valid:


/// <summary>
/// Checks if property or field and all parent instances if they are null.
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="parent">The parent.</param>
/// <param name="lambda">The lambda expression.</param>
/// <returns><c>true</c> if null; otherwise, <c>false</c>.</returns>
/// <remarks>Only properties and/or fields are allowed for all parts of the expression.</remarks>
public static bool IsNull<T>(this object parent, Expression<Func<T>> lambda)
{
    var memberParts = new List<MemberInfo>();
    var expression = lambda.Body;

    // ExpressionType.Constant occurs once we get to the parent
    while (expression.NodeType != ExpressionType.Constant)
    {
        var memberExpression = (MemberExpression)expression;
        memberParts.Add(memberExpression.Member);
        expression = memberExpression.Expression;
    }

    // Parts are extracted in reverse order, so this corrects it
    memberParts.Reverse();

    // 'instance' keeps track of the instance associated with 'member' below
    var instance = parent;
    while (memberParts.Any())
    {
        // Set the current member for evaluation on this loop
        var member = memberParts.First();

        // Break down how this is evaluated by property vs. field
        switch (member.MemberType)
        {
            case MemberTypes.Property:
                var property = (PropertyInfo)member;

                // Gets the value by invoking the property on the instance
                object value = property.GetValue(instance, null);

                // Compares value with null
                if (value == null) return true;

                // If not null, set 'instance' to the value for the next loop
                instance = value;
                break;

            case MemberTypes.Field:
                var field = (FieldInfo)member;

                // Gets the value by v the field on the instance
                value = field.GetValue(instance);

                // Compares value with null
                if (value == null) return true;

                // If not null, set 'instance' to the value for the next loop
                instance = value;
                break;

            default:
                // Method type members, and others, are not supported
                throw new InvalidOperationException("IsNull: MemberType must be Property or Field.");
        }

        // If not null, remove the first element for the next loop
        memberParts.RemoveAt(0);
    }

    // No nulls were found
    return false;
}

As demonstrated, this method inspects each element from greatest parent down until reaching our target element to ensure all elements in the tree are not null.

The usage would be like this:


// Why do this:
var test = (parent == null || parent.MyProperty == null || parent.MyProperty.ChildProperty == null);

// When you can do this:
var test = parent.IsNull(()=> parent.MyProperty.ChildProperty);

// Check
// "test" is true if parent is null.

 

And that wraps this series on extensions.

Happy coding.

A different take on Deep Copy

Categories: .Net, C#, Clone, Silverlight, WPF
Comments: 1

This article has been revised here.

I have seen several examples of different methods to perform a deep copy in C#. The most common one is with a Stream, implemented like this:


public static T DeepClone<T>(T obj)
{
 using (var ms = new MemoryStream())
 {
   var formatter = new BinaryFormatter();
   formatter.Serialize(ms, obj);
   ms.Position = 0;

   return (T) formatter.Deserialize(ms);
 }
}

That is certainly one way to do it but it won’t work in Silverlight.

The suggested Sliverlight approach:


public static T DeepCopy<T>(this T oSource)
{
    T oClone;
    DataContractSerializer dcs = new DataContractSerializer(typeof(T));

    using (MemoryStream ms = new MemoryStream())
    {
        dcs.WriteObject(ms, oSource);
        ms.Position = 0;
        oClone = (T)dcs.ReadObject(ms);
    }

    return oClone;
}

All right, that works in Silverlight but it requires that all objects passed are Serializable. That is actually a bit of a daunting task for custom objects at times, and we should try to use a simpler more universal approach.

So, I suggest using reflection:

/// <summary>
/// Copies all public properties from one object to another.
/// </summary>
/// <param name="source">The source.</param>
/// <param name="destination">The destination.</param>
/// <exception cref="System.Exception">Source and destination cannot be null and must be 
/// of the same type!</exception>
public static void DeepCopy(object source, object destination)
{
    if ((source != null &amp;&amp; destination != null) &amp;&amp; 
        (source.GetType() == destination.GetType()))
    {
        // Get properties
        var propertyInfos = source.GetType().GetProperties();
        if (propertyInfos.Length <= 0) return;

        // Evaluate
        foreach (var propInfo in propertyInfos)
        {
            // Process only public properties
            if (!propInfo.CanWrite) continue;

            // Get value from source and assign to destination
            object value = propInfo.GetValue(source, null);
            propInfo.SetValue(destination, value, null);
            if (value == null) continue;

            // Check for properties and propogate if they exist
            var newPropInfos = value.GetType().GetProperties();
            if (newPropInfos.Length > 0)
            {
                // Copy properties for each child where necessary
                DeepCopy(
                    source.GetType().GetProperty(propInfo.Name),
                    destination.GetType().GetProperty(propInfo.Name));
            }
        }
    }

    else
    {
        throw new Exception(
            "Source and destination cannot be null and must be of " +
            "the same type!");
    }
}

This will work in Silverlight because of the if block for propInfo.CanWrite. This is added because based on Silverlights rigid security constraints we can only copy public properties.

That’s all for now.