Download the demo code here

Introduction
Demo Code
- Change Tracker
  - Demonstration Using Tests
    - Simple Test Cases
    - List Test Cases
  - So How Does The Tracker Work
- Convert
- Hash Compute
- If-Then-Else
- Method Call
- Property Get/Set
  - Simple Example
  - Complex Example
- Dynamic Where
- That's It

Introduction

This article is a strange one in a lot of ways, as there is no real end product with it as such, it is more of a how do I do X / recipe type article. So what are the recipes, and what will it show you, the reader how to do?

Well I am going to start with a wee story first (don't worry the point is one its way). A while back I got this email out the blue from this guy in the states who was creating a MVVM framework for Windows 8, and I have kept in contact with this dude (and he really is a dude, if I can convince him to upload an image, you will see what I mean) and we got talking about his IOC container within his MVVM framework, which is like MEF for Windows 8.

Looking through Ian's code, it was immediately obvious to me, than Ian really (and I do mean really) knew how to use the Expression API within .NET. This namespace has always intrigued me, so I talked to Ian about the merits of writing a joint kind of article, where we would effectively come up with some scenarios to solve using the Expression API.

Ian said he would be more than happy to write the code to any scenarios I could come up with, if I was happy to do the article writing side of things. This seemed a very fair way to do it, so we have done just that.

Now you may be asking what is so cool about the Expression API, well one thing I quite like is that you can literally write entire programs in it, and another thing that you see time and time again, is creating compiled lambda expressions that have been built up on the fly, which compile down to a delegate, so provide uber fast performance when compared to reflection. That is why the Expresson API can be useful (at least we feel that way).

That is essentially what this article is all about, we have a bunch of scenarios that I came up with (which I hope are good ones) which Ian has coded up. The examples range from simple property get/set through to some rather complex examples where we show you how to do things like how to create If-Then-Else Expressions and compute a HashCode for an object based of its property values.

Here is a list of the scenarios we will be covering

Change Tracker	This example will show you how to use the `Expression` APIs to monitor an arbitrary object(s) for change notifications, or arbitrary collection(s) for change notifications. Where the object(s) may be deeply nested within a tree like hierarchy
Convert	This example will show you how to use the `Expression.Convert`and `Expression.TypeAs` APIs to carry out convert and casting operations
HashCompute	This example will show you how to create an expression that will be able to generate a dynamic hash code based on examining a source input objects properties.
IfThenElse	This example will show you how to create conditional Expression tree logic using the `Expression.IfThenElse` expression APIs
MethodCall	This example will show you how to call an obects method where there are no parameters required, and shall also show a more complex example where the method being called requires parameters
PropertyGetSet	This example will show you how to get/set an obects property, and shall also show a more complex example where we show you how to get/set a property value on a nested property deep in an objects hierarchy of properties
WhereClause	This example will show you how to build a dynamic where clause that can be applied to a `IEnumerable<T>` using the `Where(...)` extension method

We hope these examples will help to show how to do some of the more common programming tasks that you may encounter, using the Expression API.

Demo Code

In this section you will find all the examples. There is pretty much a 1-1 mapping between these sections and the actual demo code projects.

Change Tracker

This is by far the most complicated of all of the examples in what it actually provides, but conversely the actual Expression API usage is not as complicated as some of the other examples you will find here. The reason we have chosen to start with this one, is that we honestly think it's a useful piece of code (that is if you take the "Law Of Demeter" with a pinch of salt). It's a lot to take in, but please bear with us it's a journey (we hope) that is worth taking.

We think the best way to start with this one, is to explain in bullet points what we wanted to create, so here is that original list that both Ian and I agreed on before we started this article:

The tracker should allow the tracking of many objects at once
The tracker should either yield a IObservable<bool> for the IsDirty property (we were originally going to use Rx), or raise an overall IsDirty event that users of the tracker could use. We opted for a simple event in the end, as we did not want to burden non Rx users with an extra Dll, and extra learning.
The tracker should be deemed "Not Dirty" if a tracked value is set back to its original value. This would only obviously work for simple properties, such as Int, double, string etc etc
Would have some sort of method to mark the overall tracker as "Not Dirty" to allow the change tracking to effectively be reset back to its original state
There should be some sort of change tracker interface that allows a new object to be tracked for changes by providing a Expression that could be used to pull out the property. Where the object being requested to track would have point to a object who supported change tracking via the use of the INotifyPropertyChanged interface
There should be some sort of change tracker interface that allows a List object(s) to be tracked for add/removing by providing a Expression that could be used to pull out the property. Where the object being requested to track would have point to a object who supported change tracking via the use of the INotifyCollectionChangedinterface
There should be some sort of change tracker interface that allows a List object(s) to be tracked for changes by providing a Expression that could NOT only be used to pull out the property, but could also supply a Predicate<T> that must be satisfied before the tracked list items would be considered to contribute to the overall dirty state of the tracker.

So that was the brief for the change tracker, based on that Ian and I came up with the following interface designs for the change tracker.

public class IsDiryChangedArgs : EventArgs
{
	/// <summary>
	/// Default constructor
	/// </summary>
	/// <param name="isDirty">the new value for is dirty</param>
	public IsDiryChangedArgs(bool isDirty)
	{
	IsDirty = isDirty;
	}

	/// <summary>
	/// The new is dirty value
	/// </summary>
	public bool IsDirty { get; private set; }
}




public interface ITracker : IDisposable
{
	/// <summary>
	/// Track all properties on the specified object
	/// </summary>
	/// <typeparam name="T"></typeparam>
	/// <param name="objectToTrack"></param>
	void TrackObject<T>(T objectToTrack);

	/// <summary>
	/// Track a specified property on an object or a child object
	/// </summary>
	/// <typeparam name="T"></typeparam>
	/// <typeparam name="TProp"></typeparam>
	/// <param name="objectToTrack"></param>
	/// <param name="trackingExpression"></param>
	void TrackObject<T, TProp>(T objectToTrack, Expression<Func<T, TProp>> trackingExpression);

	/// <summary>
	/// True if any of the objects being tracked is dirty
	/// </summary>
	bool IsDirty { get; }

	/// <summary>
	/// Mark all components in the tracker as clean
	/// </summary>
	void MarkAsClean();

	/// <summary>
	/// List of component trackers associated with this tracker
	/// </summary>
	/// <returns></returns>
	IEnumerable<IComponentTracker> ComponentTrackers();

	/// <summary>
	/// Event is raised when the IsDirty flag is changed
	/// </summary>
	event EventHandler<IsDiryChangedArgs> IsDirtyChanged;
}

We feel the best place to learn how to use this would be by looking at the tests, so let's see an example or 2 from the tests that we have provided with it:

If we consider the following tests files and there relationships:

CLICK IMAGE FOR A LARGER VERSION

Demonstration Using Tests

Then examine some of the test cases, it should start to make a bit more sense.

Simple Test Cases

These test cases cover how you would add a tracker for non list type properties that you wished to track. From these tests you should be able to see how you can track a object, or even a object which is deep down the property tree. (Remember through that the object(s) being tracked must implement the INotifyPropertyChanged).

[TestClass]
public class SimpleTrackingTests
{
	[TestMethod]
	public void TrackFirstLayer()
	{
		bool eventCalled = false;
		Child1 child1 = new Child1{ TestInt = 10};
		Tracker tracker = new Tracker();

		Assert.IsFalse(tracker.IsDirty);
		tracker.IsDirtyChanged += (sender, args) => eventCalled = true;

		tracker.TrackObject(child1, x => x.TestInt);
		Assert.IsFalse(tracker.IsDirty);

		child1.TestInt = 5;
		Assert.IsTrue(tracker.IsDirty);
		Assert.IsTrue(eventCalled);

		eventCalled = false;
		child1.TestInt = 10;
		Assert.IsFalse(tracker.IsDirty); 
		Assert.IsTrue(eventCalled);
	}

	[TestMethod]
	public void TrackSecondLayer()
	{
		bool eventCalled = false;
		ChildA childA = new ChildA
			{
	            Child1 = new Child1
	              { 
	                  TestInt = 10
	              }
            };


		Tracker tracker = new Tracker();
		tracker.IsDirtyChanged += (sender, args) => eventCalled = true;

		Assert.IsFalse(tracker.IsDirty);
		tracker.TrackObject(childA, x => x.Child1.TestInt);
		Assert.IsFalse(tracker.IsDirty);
		childA.Child1.TestInt = 5;
		Assert.IsTrue(tracker.IsDirty);
		Assert.IsTrue(eventCalled);

		eventCalled = false;
		childA.Child1.TestInt = 10;

		Assert.IsFalse(tracker.IsDirty); 
		Assert.IsTrue(eventCalled);
	}




	[TestMethod]
	public void TrackThirdLayer()
	{
		bool eventCalled = false;
		RootObject rootObject = 
			new RootObject
		    {
			   ChildA = new ChildA
			   {
				    Child1 = new Child1
				    {
				       TestInt = 5
				    }
			    }
		    };


		Tracker tracker = new Tracker();
		tracker.IsDirtyChanged += (sender, args) => eventCalled = true;
		Assert.IsFalse(tracker.IsDirty);

		tracker.TrackObject(rootObject,x => x.ChildA.Child1.TestInt);
		Assert.IsFalse(tracker.IsDirty);

		rootObject.ChildA.Child1.TestInt = 10;
		Assert.IsTrue(tracker.IsDirty);
		Assert.IsTrue(eventCalled);


		eventCalled = false;
		rootObject.ChildA.Child1.TestInt = 5;
		Assert.IsFalse(tracker.IsDirty);
		Assert.IsTrue(eventCalled);
	}

	[TestMethod]
	public void MarkAsClean()
	{
		bool eventCalled = false;
		ChildA childA = new ChildA
		{
			Child1 = new Child1 { TestInt = 10 }
		};

		Tracker tracker = new Tracker();
		tracker.IsDirtyChanged += (sender, args) => eventCalled = true;
		Assert.IsFalse(tracker.IsDirty);
		tracker.TrackObject(childA, x => x.Child1.TestInt);
		Assert.IsFalse(tracker.IsDirty);

		childA.Child1.TestInt = 5;
		Assert.IsTrue(tracker.IsDirty);
		Assert.IsTrue(eventCalled);

		tracker.MarkAsClean();
		Assert.IsFalse(tracker.IsDirty);
		eventCalled = false;
		childA.Child1.TestInt = 10;

		Assert.IsTrue(tracker.IsDirty);
		Assert.IsTrue(eventCalled);

		childA.Child1.TestInt = 5;

		Assert.IsFalse(tracker.IsDirty);
		Assert.IsTrue(eventCalled);
	}
}

List Test Cases

These test cases cover how you would add a tracker for list type properties that you wished to track. From these tests you should be able to see how you can track a list for the following scenarios

Simple Add/Remove (item being tracked must implement INotifyCollectionChanged)
Track a List where a certain Predicate<T> holds true
Track a nested List

[TestClass]
public class ListTrackingTests
{
	[TestMethod]
	public void TrackAllTest()
	{
		ObservableCollection<Child1> newList = new ObservableCollection<Child1>
			{
				new Child1 { TestInt = 5 },
				new Child1 { TestInt = 10 }
			};

		Tracker tracker = new Tracker();
		tracker.TrackObject(newList);

		Assert.IsFalse(tracker.IsDirty);
		newList[0].TestInt = 10;

		Assert.IsTrue(tracker.IsDirty);
		newList[0].TestInt = 5;
		Assert.IsFalse(tracker.IsDirty);
	}




	[TestMethod]
	public void FilterListTest()
	{
		ObservableCollection<Child1> testList = new ObservableCollection<Child1>();
		ListObject listContainer = new ListObject
		{
			Child1List = testList
		};

		Tracker tracker = new Tracker();

		tracker.TrackObject(listContainer, 
			c => c.Child1List.TrackList(x => x.TestInt > 0));

		testList.Add(new Child1 { TestInt = 0, TestString = "Start" });
		Assert.IsFalse(tracker.IsDirty);

		testList.Add(new Child1 { TestInt = 1, TestString = "Start" });

		Assert.IsTrue(tracker.IsDirty);
	}




	[TestMethod]
	public void NestedFilterListTest()
	{
		ObservableCollection<Child1> testList = new ObservableCollection<Child1>();
		ListObject listContainer = new ListObject
			{
				Child1List = testList
			};

		Tracker tracker = new Tracker();
		tracker.TrackObject(listContainer, 
			c => c.Child1List.TrackList(x => x.TestInt > 0).TestString);

		testList.Add(new Child1 { TestInt = 0, TestString = "Start" });
		Assert.IsFalse(tracker.IsDirty);
		testList.Add(new Child1 { TestInt = 1, TestString = "Start" });
		Assert.IsTrue(tracker.IsDirty);
	}




	[TestMethod]
	public void NestedListTest()
	{
		ObservableCollection<Child1> testList = new ObservableCollection<Child1>();
		ListObject listContainer = new ListObject
		{
			Child1List = testList
		};

		testList.Add(new Child1 { TestString = "Start" });
		Tracker tracker = new Tracker();

		tracker.TrackObject(listContainer, 
			c => c.Child1List.TrackList().TestString); 
			
		Assert.IsFalse(tracker.IsDirty);
		testList[0].TestString = "Hello";

		Assert.IsTrue(tracker.IsDirty);

		testList[0].TestString = "Start";
		Assert.IsFalse(tracker.IsDirty);
	}
}

So How Does The Tracker Work

One of the things that became obvious fairly quickly was that we would need to have different type of trackers. So we came up with 3 of them which can be seen in the following diagram

Where the IComponentTracker interface looks like this:

/// <summary>
/// IComponentTracker is the interface all component trackers implements.
/// </summary>
public interface IComponentTracker : IDisposable
{
	/// <summary>
	/// The component being tracker
	/// </summary>
	object TrackedComponent { get; }

	/// <summary>
	/// True if the component or one of it's children is dirty
	/// </summary>
	bool IsDirty { get; }

	/// <summary>
	/// mark the component and all it's children as clean
	/// </summary>
	void MarkAsClean();

	/// <summary>
	/// A list of child trackers
	/// </summary>
	/// <returns></returns>
	IEnumerable<IComponentTracker> ChildTrackers();

	/// <summary>
	/// Event that is raised when the is dirty flag is changed
	/// </summary>
	event EventHandler<IsDiryChangedArgs> IsDirtyChanged;
}

We will try not to outline every aspect of these classes but rather focus on the most salient points of each. However before we get into each of the 3 types of component tracker, let us just look at a common bit of code that all 3 of the component trackers make use of in order to build Expression(s) to obtain getters for the properties to track.

Here is the relevant code that essentially builds up lookup delegates against a property name for an object:

/// <summary>
/// ComponentTrackerHelper is used create property accessor and new component trackers. 
/// </summary>
public class ComponentTrackerHelper
{
	private readonly Dictionary<string, Func<object, object>> accessMethods = new Dictionary<string, Func<object, object>>();

	/// <summary>
	/// Access a named property on an object
	/// </summary>
	/// <param name="target"></param>
	/// <param name="propertyName"></param>
	/// <returns></returns>
	public object AccessProperty(object target, string propertyName)
	{
		Func<object, object> accessDelegate = GetPropertyAccessor(target, propertyName);

		if (accessDelegate != null)
		{
			return accessDelegate(target);
		}
		throw new Exception(string.Format("Could not access property {0} on {1}", propertyName, target.GetType().FullName));
	}

	/// <summary>
	/// Creates a new property accessor function
	/// </summary>
	/// <param name="target"></param>
	/// <param name="propertyName"></param>
	/// <returns></returns>
	public Func<object, object> GetPropertyAccessor(object target, string propertyName)
	{
		string fullPropertyName = target.GetType().FullName + "|" + propertyName;
		Func<object, object> accessDelegate;

		if (!accessMethods.TryGetValue(fullPropertyName, out accessDelegate))
		{
			PropertyInfo propertyInfo = target.GetType().GetProperty(propertyName);
			// create the parameter for the instance parameter of type object
			ParameterExpression inParameter = Expression.Parameter(typeof(object), "objectParam");
			// cast the instance to it's real type
			Expression castExpression = Expression.Convert(inParameter, target.GetType());
			// create the property access expression
			Expression propertyAccessExpression =
				Expression.Property(castExpression, propertyInfo);
			// cast the property access expression to an object
			Expression returnCastExpression = Expression.Convert(propertyAccessExpression, typeof(object));
			accessDelegate = Expression.Lambda<Func<object, object>>(returnCastExpression, inParameter).Compile();
			accessMethods.Add(fullPropertyName, accessDelegate);
		}
		return accessDelegate;
	}

	/// <summary>
	/// Creates a new component tracker based on the provided TrackerInfo
	/// </summary>
	/// <param name="target"></param>
	/// <param name="path"></param>
	/// <returns></returns>
	public IComponentTracker CreateTracker(object target, TrackerInfo path)
	{
		switch (path.TrackerType)
		{
			case ComponentTrackerType.Node:
				return new NodeComponentTracker(this, target, path);
			case ComponentTrackerType.Leaf:
				return new LeafComponentTracker(this, target, path);
			case ComponentTrackerType.List:
				return new ListComponentTracker(this, target as IEnumerable, path);
		}
		throw new Exception("Unknown tracker type: " + path.TrackerType);
	}
}

Now on with the different types of component trackers:

LeafComponentTracker

Tracks every change made on an object.
The object MUST be a INotifyPropertyChanged implementing object
We will track old and new values such that we know if it is truly dirty by comparing an "Original" value with a current value. This is handled by the use of a WeakReference, to make sure the object is still capable of being garbage collected.
Should also support a "Not Dirty" state if returned to the the original value

Here is the the most relevant code for this type of tracker

/// <summary>
/// Initialize the object, setting up property trackers for all relavent properties
/// </summary>
/// <param name="helper"></param>
/// <param name="objectToTrack"></param>
/// <param name="trackerInfo"></param>
private void Initialize(ComponentTrackerHelper helper, object objectToTrack, TrackerInfo trackerInfo)
{
	propertyTrackers = new Dictionary<string, PropertyTracker>();
	this.helper = helper;
	this.objectToTrack = objectToTrack;
	this.trackerInfo = trackerInfo;
	// listen to property change events
	if (objectToTrack is INotifyPropertyChanged)
	{
		((INotifyPropertyChanged)objectToTrack).PropertyChanged += OnPropertyChanged;
	}
	// loop through all public properties
	foreach (PropertyInfo propertyInfo in objectToTrack.GetType().GetProperties(BindingFlags.Public | BindingFlags.Instance))
	{
		// only monitor properties that are write able
		if (propertyInfo.CanWrite)
		{
			bool trackWeakly = !(propertyInfo.PropertyType.IsValueType || propertyInfo.PropertyType == typeof(string));
			// create a new property tracking object and a property accessor delegate
			PropertyTracker propertyTracker = new PropertyTracker(trackWeakly)
	
			// use the new property access delegate to retrieve the property from the object
			object origValue = propertyTracker.PropertyAccess(objectToTrack);

			// store the original value 
			propertyTracker.SetOriginalValue(origValue);

			// save the new property tracker into the dictionary by property name
			propertyTrackers[propertyInfo.Name] = propertyTracker;
		}
	}
}


/// <summary>
/// This is the property changed handler for the object being tracker
/// </summary>
/// <param name="sender"></param>
/// <param name="propertyChangedEventArgs"></param>
private void OnPropertyChanged(object sender, PropertyChangedEventArgs propertyChangedEventArgs)
{
	PropertyTracker propertyTrackingInfo;
	// Get the property tracker associated with the property that changed
	if (propertyTrackers.TryGetValue(propertyChangedEventArgs.PropertyName, out propertyTrackingInfo))
	{
		bool dirtyChanged = false;
		bool hasOriginalValue;
		object originalValue = propertyTrackingInfo.GetOriginalValue(out hasOriginalValue);
		object newValue = propertyTrackingInfo.PropertyAccess(sender);

		if (newValue != null)
		{
			// Property is reverting back to original value so setting is dirty to false
			if (newValue.Equals(originalValue))
			{
				propertyTrackingInfo.IsDirty = false;
				dirtyChanged = true;
			}
			// else if the property is clean we need to dirty it up
			else if (!propertyTrackingInfo.IsDirty)
			{
				propertyTrackingInfo.IsDirty = true;
				dirtyChanged = true;
			}
		}
		else if (!hasOriginalValue)
		{
			// the original value was null and the new value is null so the property is now clean
			if (propertyTrackingInfo.IsDirty)
			{
				propertyTrackingInfo.IsDirty = false;
				dirtyChanged = true;
			}
		}
		else // the new value is null and we have an original value
		{
			// only set to true if the property is clean
			if (!propertyTrackingInfo.IsDirty)
			{
				propertyTrackingInfo.IsDirty = true;
				dirtyChanged = true;
			}
		}


		// only check for dirty properties if the property dirty flag changed
		if (dirtyChanged)
		{
			IsDirty = propertyTrackers.Values.Any(x => x.IsDirty);
		}
	}
}

ListComponentTracker

Tracks changes to a ObservableCollection (or any other collection that implenents INotifyCollectionChanged), such as Add/Remove/Replace/Reset
The object MUST be a INotifyCollectionChanged implementing object

Here is the most relevant code:

/// <summary>
/// Initialize the tracker
/// </summary>
/// <param name="helper"></param>
/// <param name="objectToTrack"></param>
/// <param name="trackerInfo"></param>
private void Initialize(ComponentTrackerHelper helper, IEnumerable objectToTrack, TrackerInfo trackerInfo)
{
	componentTrackers = new Dictionary<object, IComponentTracker>();
	this.helper = helper;
	this.objectToTrack = objectToTrack;
	this.trackerInfo = trackerInfo;

	// add all children to be tracked
	AddChildren(objectToTrack);

	// if this list implements the collection changed event we need to listen to it
	if (objectToTrack is INotifyCollectionChanged)
	{
		((INotifyCollectionChanged)objectToTrack).CollectionChanged += OnCollectionChanged;
	}
}

/// <summary>
/// Handler for collection changed. It adds and removes trackers as the list changes
/// </summary>
/// <param name="sender"></param>
/// <param name="notifyCollectionChangedEventArgs"></param>
private void OnCollectionChanged(object sender, NotifyCollectionChangedEventArgs notifyCollectionChangedEventArgs)
{
	bool fireChanged = false;
	bool changed = false;

	if (!IsDirty)
	{
		fireChanged = true;
	}

	switch (notifyCollectionChangedEventArgs.Action)
	{
		case NotifyCollectionChangedAction.Add:
			changed = AddChildren(notifyCollectionChangedEventArgs.NewItems);
			break;
		case NotifyCollectionChangedAction.Remove:
			changed = RemoveChildren(notifyCollectionChangedEventArgs.OldItems);
			break;
		case NotifyCollectionChangedAction.Replace:
			changed = RemoveChildren(notifyCollectionChangedEventArgs.OldItems);
			changed |= AddChildren(notifyCollectionChangedEventArgs.NewItems);
			break;
		case NotifyCollectionChangedAction.Reset:
			changed = RemoveChildren(componentTrackers.Values.ToArray());
			break;
	}

	if (changed)
	{
		isDirty = true;
	}

	if (isDirty && fireChanged && IsDirtyChanged != null)
	{
		IsDirtyChanged(this, new IsDiryChangedArgs(isDirty));
	}
}




/// <summary>
/// Handler for child tracker is dirty changing
/// </summary>
/// <param name="sender"></param>
/// <param name="eventArgs"></param>
private void ChildTrackerIsDirtyChanged(object sender, IsDiryChangedArgs eventArgs)
{
	// short circuit if we are already dirty
	if (isDirty)
	{
		return;
	}

	IComponentTracker childTracker = (IComponentTracker)sender;
	bool newValue = childTracker.IsDirty;
	bool fireChange = true;

	foreach (KeyValuePair<object, IComponentTracker> kvp in componentTrackers)
	{
		if (kvp.Key == sender)
		{
			continue;
		}

		if (kvp.Value.IsDirty != newValue)
		{
			fireChange = false;
			break;
		}
	}

	if (fireChange && IsDirtyChanged != null)
	{
		IsDirtyChanged(this, eventArgs);
	}
}


/// <summary>
/// Add children to be tracked
/// </summary>
/// <param name="childObjects"></param>
/// <returns></returns>
private bool AddChildren(IEnumerable childObjects)
{
	bool returnValue = false;
	IListComponentFilter listFilter = trackerInfo.ListFilter;

	// if we have child tracker info create a new tracker for the object
	if (trackerInfo.ChildTrackerInfo != null)
	{
		foreach (object child in childObjects)
		{
			// filter the object if a filter exists
			if (listFilter != null && !listFilter.FilterComponent(child))
			{
				continue;
			}

			IComponentTracker childTracker =
				helper.CreateTracker(child, trackerInfo.ChildTrackerInfo);
			childTracker.IsDirtyChanged += ChildTrackerIsDirtyChanged;
			componentTrackers.Add(child, childTracker);
			returnValue = true;
		}
	}
	else 
	{
		// Create a leaf tracker for each child in the list
		foreach (object child in childObjects)
		{
			if (listFilter != null && !listFilter.FilterComponent(child))
			{
				continue;
			}

			IComponentTracker childTracker =
				helper.CreateTracker(child, new TrackerInfo { TrackerType = ComponentTrackerType.Leaf });

			childTracker.IsDirtyChanged += ChildTrackerIsDirtyChanged;
			componentTrackers.Add(child, childTracker);
			returnValue = true;
		}
	}
	return returnValue;
}




/// <summary>
/// Remove child object that need to stop being tracked
/// </summary>
/// <param name="childObjects"></param>
/// <returns></returns>
private bool RemoveChildren(IEnumerable childObjects)
{
	bool returnValue = false;
	// foreach child object try and find a tracker and dispose it
	foreach (object child in childObjects)
	{
		IComponentTracker oldTracker;
		if (componentTrackers.TryGetValue(child, out oldTracker))
		{
			componentTrackers.Remove(child);
			oldTracker.IsDirtyChanged -= ChildTrackerIsDirtyChanged;
			oldTracker.Dispose();

			returnValue = true;
		}
	}




	return returnValue;
}

NodeComponentTracker

Tracks a specific property on an object
- The object MUST be a INotifyCollectionChanged implementing object

Here are the most relevant methods:

/// <summary>
/// Initialize the new tracker
/// </summary>
/// <param name="helper"></param>
/// <param name="objectToTrack"></param>
/// <param name="trackerInfo"></param>
private void Initialize(ComponentTrackerHelper helper, object objectToTrack, TrackerInfo trackerInfo)
{
	this.helper = helper;
	this.objectToTrack = objectToTrack;
	this.trackerInfo = trackerInfo;
	// calculate if we should track the property weakly
	bool trackWeakly = !(trackerInfo.PropertyType.IsValueType || trackerInfo.PropertyType == typeof(string));
	object currentValue = helper.AccessProperty(objectToTrack, trackerInfo.PropertyName);

	propertyTracker = new PropertyTracker(trackWeakly);
	propertyTracker.SetOriginalValue(currentValue);
	// if we have a current value and child tracker info start tracking
	if (currentValue != null && trackerInfo.ChildTrackerInfo != null)
	{
		childComponentTracker = helper.CreateTracker(currentValue, trackerInfo.ChildTrackerInfo);
		childComponentTracker.IsDirtyChanged += ChildComponentTrackerOnIsDirtyChanged;
	}

	INotifyPropertyChanged propertyChangeObject = objectToTrack as INotifyPropertyChanged;
	if (propertyChangeObject != null)
	{
		propertyChangeObject.PropertyChanged += PropertyChangedOnTrackedObject;
	}
}




/// <summary>
/// Handler for child component tracker changing is dirty
/// </summary>
/// <param name="sender"></param>
/// <param name="eventArgs"></param>
private void ChildComponentTrackerOnIsDirtyChanged(object sender, IsDiryChangedArgs eventArgs)
{
	IComponentTracker componentTracker = sender as IComponentTracker;
	if (!propertyTracker.IsDirty && IsDirtyChanged != null)
	{
		IsDirtyChanged(this, eventArgs);
	}
}

/// <summary>
/// The property change handler for the object being tracked
/// </summary>
/// <param name="sender"></param>
/// <param name="propertyChangedEventArgs"></param>
private void PropertyChangedOnTrackedObject(object sender, PropertyChangedEventArgs propertyChangedEventArgs)
{
	if (childComponentTracker != null)
	{
		childComponentTracker.IsDirtyChanged -= ChildComponentTrackerOnIsDirtyChanged;
		childComponentTracker.Dispose();
		childComponentTracker = null;
	}

	bool hasOriginalValue;
	object originalValue = propertyTracker.GetOriginalValue(out hasOriginalValue);
	object newValue = helper.AccessProperty(objectToTrack, trackerInfo.PropertyName);

	if (newValue != null)
	{
		if (originalValue != null &&& newValue.Equals(originalValue))
		{
			IsDirty = false;
		}
		else
		{
			IsDirty = true;
		}
	}
	else
	{
		if (hasOriginalValue)
		{
			IsDirty = true;
		}
		else
		{
			IsDirty = false;
		}
	}

	// if there is a new value and we have a child tracker info we need to create a new tracker and listen to it
	if (newValue != null &&&& trackerInfo.ChildTrackerInfo != null)
	{
		childComponentTracker = helper.CreateTracker(newValue, trackerInfo.ChildTrackerInfo);
		childComponentTracker.IsDirtyChanged += ChildComponentTrackerOnIsDirtyChanged;
	}
}

Now we know there is a lot of code here, and it may be hard to take in, but have a play with the unit tests, the core API is actually very simple, and we feel this one is actually very very useful.

Convert

Expression.Convert

This simple example shows you how to use Cast (achieved using Expression.Convert) which takes an object and casts to type of T

Here is the relevant code:

/// <summary>
/// Creates a new function that cast an object to a type T.
/// it will throw an exception if the cast is invalid
/// </summary>
/// <typeparam name="T"></typeparam>
/// <returns></returns>
public static Func<object, T> CreateStaticCastStatement<T>()
{
	ParameterExpression inObject = Expression.Parameter(typeof(object));

	// static cast expression (T)inObject
	Expression convertExpression = Expression.Convert(inObject, typeof(T));
			
	return Expression.Lambda<Func<object,T>>(convertExpression,inObject).Compile();
}

Which we could use like this. If the cast call fails an Exceptionwill be thrown.

Func<object,int> intStaticCastMethod = ConvertStatementCreater.CreateStaticCastStatement<int>();
int testValue = intStaticCastMethod(8);

Expression.TypeAs

We can also use Expression.TypeAs to return null, if the cast fails instead of throwing an Exception which the previous example will do for an invalid cast.

/// <summary>
/// Creates a new function that casts an object it will return null if the cast fails
/// </summary>
/// <typeparam name="T"></typeparam>
/// <returns></returns>
public static Func<object, T> CreateAsCastStatement<T>()
{
	ParameterExpression inObject = Expression.Parameter(typeof(object));
			
	// cast using as statement   inObject as T
	Expression convertExpression = Expression.TypeAs(inObject, typeof(T));

	return Expression.Lambda<Func<object, T>>(convertExpression, inObject).Compile();
}

Which we could use like this

Func<object, string> stringAsCastMethod = ConvertStatementCreater.CreateAsCastStatement<string>();

if (stringAsCastMethod("Hello") != "Hello")
{
	throw new Exception("Cast did not work");	
}

if (stringAsCastMethod(789) != null)
{
	throw new Exception("Should have been null");
}

Hash Compute

In this example we kind of had a bit of fun, where we show you how you could create a generic GetHashCode function for any object, which is achieved by examining all the public properties of the T object type that is passed to the GenericHashCalculator<T>. For each property value a new Expression.Property is setup and then an Expression.Call is done to call the GetHashCode() method for the property object instance. The results of which are appended to an ongoing StringBuilder string concatenation (where the int value of the individual property GetHashCode() is passed to the StringBuilder.Append() method). At the end of iterating all the public properties the final StringBuilder string, GetHashCode()method is called which is effectievly the overall hash value for the object.

So if we use this demo class

public class Person
{
	public string FirstName { get; set; }
	public string LastName { get; set; }
	public string EmailAddress { get; set; }
	public int Age { get; set; }
	public int NumberOfSiblings { get; set; }
	public int NumberOfChildren { get; set; }
}

We can build up a single properties hash value that is appended to the ongoing StringBuilder string value (as an int) which will be used as the final hash value

private Expression CreatePropertyHashExpression(
    PropertyInfo propertyInfo,
	ParameterExpression stringBuilderParameter,
	ParameterExpression objectToHashParameter)
{
	Expression returnExpression = null;
	Expression propertyExpression = Expression.Property(objectToHashParameter, propertyInfo);
	Expression propertyHashCodeExpression = Expression.Call(propertyExpression, getHashCodeMethod);
	Expression stringBuildAppendExpression = Expression.Call(
        stringBuilderParameter,
	stringBuilderAppendMethod,
	propertyHashCodeExpression);
			
    if (propertyInfo.PropertyType.IsValueType)
	{
	returnExpression = stringBuildAppendExpression;
	}
	else
	{
	Expression notEqualToNullExpress = Expression.NotEqual(propertyExpression, Expression.Constant(null));
	returnExpression = Expression.IfThen(notEqualToNullExpress, stringBuildAppendExpression);
	}
	return returnExpression;
}

And this is the part that creates the final hash by calling the accumulated string value (from the StringBuilder that was used to create the indvidual property part hashcode values) GetHashCode() method

private void CreateHashDelegate()
{
	ParameterExpression objectToHashParameter = Expression.Parameter(typeof(T), "objectToHash");
	ParameterExpression stringBuilderParameter = Expression.Variable(typeof(StringBuilder), "stringBuilder");
	Expression assignAndCreateExpression = Expression.Assign(stringBuilderParameter,
		Expression.New(stringBuilderConstructor,
		Expression.Constant(typeof(T).FullName)));

	List<Expression> computeHashStatements = new List<Expression>();
	computeHashStatements.Add(assignAndCreateExpression);

	foreach (PropertyInfo propertyInfo in typeof(T).GetProperties(BindingFlags.Instance | BindingFlags.Public))
	{
		computeHashStatements.Add(
			CreatePropertyHashExpression(propertyInfo, 
                stringBuilderParameter, objectToHashParameter));
	}

	ParameterExpression tempStringVariable = Expression.Variable(typeof(string), "tempString");
	Expression toStringForBuilder = Expression.Call(stringBuilderParameter, stringBuildToStringMethod);
	Expression assignToStringToTempString = Expression.Assign(tempStringVariable, toStringForBuilder);
	Expression finalGetHashCodeFromString = Expression.Call(tempStringVariable, getHashCodeMethod);
	computeHashStatements.Add(assignToStringToTempString);
	computeHashStatements.Add(finalGetHashCodeFromString);
	BlockExpression finalBlock = BlockExpression.Block(typeof(int),
		new[] { stringBuilderParameter, tempStringVariable },
		computeHashStatements);

	hashDelegate = Expression.Lambda<Func<T, int>>(finalBlock, objectToHashParameter).Compile();
}

This is how you might use this GenericHashCalculator<T> class

GenericHashCalculator<Person> hasher = new GenericHashCalculator<Person>();
Person testPerson1 = new Person();
Person testPerson2 = new Person();
int hashValue1 = hasher.Compute(testPerson1);
int hashValue2 = hasher.Compute(testPerson2);
testPerson1.FirstName = "Ian";
testPerson2.FirstName = "Ian";
int hashValue3 = hasher.Compute(testPerson1);
int hashValue4 = hasher.Compute(testPerson2);
if (hashValue1 != hashValue2)
{
	throw new Exception("Hashes need to match");
}
if (hashValue1 == hashValue3)
{
	throw new Exception("Hashes must not match");
}
if (hashValue3 != hashValue4)
{
	throw new Exception("Hashes need to match");
}
testPerson1.LastName = "Johnson";
testPerson2.LastName = "IsDementedExpressionMonster";
hashValue1 = hasher.Compute(testPerson1);
hashValue2 = hasher.Compute(testPerson2);
if (hashValue1 == hashValue2)
{
	throw new Exception("Hashes should not match");
}

If-Then-Else

In this example we will look at how we can construct If-Then-Else flow control using Expression trees.

In pseudo code this is what we are trying to do

.Block() {
  .If ($var1 > 0) {
    .Return returnLabel { "Positive" }
  } .Else {
     .If ($var1 < 0) {
         .Return returnLabel { "Negative" }
     } .Else {
         .Return returnLabel { "Zero" }
     }
  };
  .Label
      ""
  .LabelTarget returnLabel:

It turns our that there is actually an Expression.IfThenElse API so translating this pseudo code into to an Expression API isn't that bad actually, here is the relevant code:

public static Func<int, string> CreatePositiveNegative()
{
	// Create the int parameter that is passed into the function
	ParameterExpression intParameter = Expression.Parameter(typeof(int));
	LabelTarget returnTarget = Expression.Label(typeof(string),"returnLabel");
	// create expression that tests the int parameter is > 0
	Expression testForPositive = Expression.GreaterThan(intParameter, Expression.Constant(0));
	// create expression that tests the int parameter is < 0
	Expression testForNegative = Expression.LessThan(intParameter, Expression.Constant(0));
	// Creates a statement that returns the string "Negative" if the int
	// parameter is less than 0 else it returns the string "Zero"
	Expression ifNegativeElseZeroExpression =
		Expression.IfThenElse(testForNegative, 
			Expression.Return(returnTarget, Expression.Constant(NegativeString)),
			Expression.Return(returnTarget, Expression.Constant(ZeroString)));
	// creates a statement that returns the string Positive if the parameter is 
	// greater than 0 else run the negative if/then/else expression
	Expression ifPositiveElseStatement =
		Expression.IfThenElse(testForPositive,
			Expression.Return(returnTarget, Expression.Constant(PositiveString)),
			ifNegativeElseZeroExpression);
	// creates a block that contains the if statement and a return label (GoTo statement)
	Expression expressionBody = Expression.Block(ifPositiveElseStatement,Expression.Label(returnTarget,Expression.Constant("")));

	return Expression.Lambda<Func<int, string>>(expressionBody, intParameter).Compile();
}

Here is an example of how we would use this

Func<int, string> intTestFunc = IfThenElseStatementCreater.CreatePositiveNegative();

if (intTestFunc(1) != IfThenElseStatementCreater.PositiveString)
{
	throw new Exception("Wrong value, should have been positive");
}

if (intTestFunc(-1) != IfThenElseStatementCreater.NegativeString)
{
	throw new Exception("Wrong value, should have been negative");
}

if (intTestFunc(0) != IfThenElseStatementCreater.ZeroString)
{
	throw new Exception("Wrong value, should have been zero");
}

We could of course replace the constant string values shown below with some more meaningful data structures

public const string PositiveString = "Positive";
public const string NegativeString = "Negative";
public const string ZeroString = "Zero";

Method Call

Methods are something that every programmer will have to call in their every day programming existence, so it stands to reason that you may occassionally have to create a delegate to call a method that you want to call a lot. Recall at the beginning of this article I stated that you could do a lot of this with Reflection, however using Reflection we take the cost of looking up a method each time, a far better option is to create some delegate that we can use when we need it, Ok, it still takes time to create the comiled Expression tree - to delegate (Func<TR>) the 1st time but after that it should be nearly as fast as calling a delegate that was known at compile time.

We have created 2 examples of how to construct method call delegate Expression trees, both of these will make use of this object to call the methods on

public class SomeBasicClass
{
    public string SomeMethod()
    {
       return "Hello World";
    }

    public string MoreComplexMethod(int param1, double param2, string param3)
    {
	return "Hello World " + param1 + " " + param2 + " " + param3;
    }
}

Simple Example

The simple case is fairly straight forward as there are no parameters for the example method (SomeMethod) to deal with, we just need to obtain a MethodInfo an create an Expression.Lambda to call the correct method. Here is the relevant code for this:

protected void Initialize(MethodInfo methodInfo)
{
	Type declaringType = methodInfo.DeclaringType;

	// the input parameter to the function
	ParameterExpression inputObject = Expression.Parameter(typeof(object));

	// loval variable of type declaringType
	ParameterExpression tVariable = Expression.Variable(declaringType);

	// cast the input object to be declaring type
	Expression castExpression = Expression.Convert(inputObject, declaringType);

	// assign the cast value to the tVaraible variable
	Expression assignmentExpression = Expression.Assign(tVariable, castExpression);
			
	// call the specified method on tVariable
	Expression callExpression = Expression.Call(tVariable, methodInfo);

	// create a block statement that contains the body of the linq statement
	BlockExpression body = Expression.Block(new[] { tVariable }, assignmentExpression, callExpression);

	// create a new delegate that takes in an object and returns an object
	invokeMethod = Expression.Lambda<Func<object, object>>(body, inputObject).Compile();
}

Compex Example

The complex case is a little more tricky as we need to deal with creating a bunch of parameters for the method that will be called (MoreComplexMethod), as such there is a bit more code, the guts of it is pretty similar stuff though, we are just buildng up an array of parameters that will be supplied to the final method call (Expression.Call)

protected void Initialize(MethodInfo methodInfo)
{
	// set the number of parameters so we can do a sanity check when we call it.
	parameterCount = methodInfo.GetParameters().Count();

	Type declaringType = methodInfo.DeclaringType;

	// the parameter to call the method on
	ParameterExpression inputObject = Expression.Parameter(typeof(object), "inputObject");
	// the input parameters for the method
	ParameterExpression parameterArray = Expression.Parameter(typeof(object[]), "parameters");
	// loval variable of type declaringType
	ParameterExpression tVariable = Expression.Variable(declaringType);

	// keep a list of the variable we declare for use when we define the body
	List<ParameterExpression> variableList = new List<ParameterExpression> { tVariable };
	// cast the input object to be declaring type
	Expression castExpression = Expression.Convert(inputObject, declaringType);

	// assign the cast value to the tVaraible variable
	Expression assignmentExpression = Expression.Assign(tVariable, castExpression);

	List<Expression> bodyExpressions = new List<Expression> { assignmentExpression };

	Expression callExpression = null;
	if (parameterCount == 0)
	{
		// call the specified method on tVariable
		callExpression = Expression.Call(tVariable, methodInfo);
	}
	else // we need to build up the parameters
	{
		int parameterNum = 0;
		List<ParameterExpression> callArguements = new List<ParameterExpression>();
		// we need to convert each parameter in the array to match the method signature
		foreach (ParameterInfo parameterInfo in methodInfo.GetParameters())
		{
			// create new local variable of the same type as the paramete
			ParameterExpression newVariable = Expression.Variable(parameterInfo.ParameterType, "param" + parameterNum);
			// add it to the list of local variable we have
			callArguements.Add(newVariable);

			// create an expression to access the parameter array
			// using a constant indexer [parameterNum]
			Expression arrayAccess = Expression.ArrayAccess(parameterArray, Expression.Constant(parameterNum));
			// cast the array access expression to be the parameter type
			Expression castArrayValue = Expression.Convert(arrayAccess, parameterInfo.ParameterType);

			// assign the casted value to the local variable
			Expression variableAssign = Expression.Assign(newVariable, castArrayValue);
			// add the variable assignment expression to the list of expression that is the body
			bodyExpressions.Add(variableAssign);
			// move to the next parameter in the array
			parameterNum++;
		}
		// add all the new local variables to the variable list
		variableList.AddRange(callArguements);

		// call the method on tVaraible with the specified callArguement list
		callExpression = Expression.Call(tVariable, methodInfo, callArguements);
	}
	// add the call expression to the body
	bodyExpressions.Add(callExpression);

	// create a block statement that contains the body of the linq statement
	BlockExpression body = Expression.Block(variableList, bodyExpressions);




	// create a new delegate that takes in an object and object array and returns an object
	invokeMethod = Expression.Lambda<Func<object, object[], object>>(body, inputObject,parameterArray).Compile();
}

Property Get/Set

A very common requirement that seems to happen all the time is to get/set property values on an object. To illustrate this, we will work through a simple case, and then move on to a more complex case.

Simple-Example

This is a bulk standard property get/set where we will be using this source object

public class BusinessObject
{
public ChildObjectA A { get; set; }
public int IntProp { get; set; }
public string StringProp { get; set; }
}

Getter

So how do we code up a getter expression?

/// <summary>
/// Creates a new Func that takes an instance object and returns the property value
/// 
/// (System.Object)((PropertyGetSetExample.BusinessObject)$objectParam).IntProp
/// </summary>
/// <param name="propertyInfo"></param>
/// <returns></returns>
private Func<object,object> CreateGetMethod(PropertyInfo propertyInfo)
{
	// create the parameter for the instance parameter of type object
	ParameterExpression inParameter = Expression.Parameter(typeof(object), "objectParam");


	// cast the instance to it's real type
	Expression castExpression = Expression.Convert(inParameter, objectType);


	// create the property access expression
	Expression propertyAccessExpression =
		Expression.Property(castExpression, propertyInfo);

	// cast the property access expression to an object
	Expression returnCastExpression = Expression.Convert(propertyAccessExpression, typeof(object));

	return Expression.Lambda<Func<object, object>>(returnCastExpression, inParameter).Compile();
}

Where we would use this as shown below. This would be something like this expressed using Reflection var x = (object)typeof(BusinessObject).GetProperty("IntProp").GetValue(bo, null); the difference being with the Expression API example is that it would be a compiled Func<object,object> so would be very quick when used again, whilst doing the same thing via Reflection would be slow each time.

BusinessObject bo = new BusinessObject { IntProp = 80 };
PropertyInfo propertyInfo = objectType.GetProperty("IntProp",BindingFlags.Public | BindingFlags.Instance);
var getMethod = CreateGetMethod(propertyInfo);
if ((int)getMethod(bo) != 80)
{
	throw new Exception("Value should have been 80");
}

Setter

So how about a setter, lets have a look at that now shall we.

/// <summary>
/// Creates a new Action that takes the instance object and the new value for the property and sets it
/// 
/// ((PropertyGetSetExample.BusinessObject)$objectParam).IntProp = (System.Int32)$newValue
/// </summary>
/// <param name="propertyInfo"></param>
/// <returns></returns>
private Action<object,object> CreateSetMethod(PropertyInfo propertyInfo)
{
	// define the first parameter to the action the object parameter
	ParameterExpression objectParameter = Expression.Parameter(typeof(object),"objectParam");




	// define the new value parameter for the action
	ParameterExpression newValueParameter = Expression.Parameter(typeof(object),"newValue");

	// cast the instance to it's proper type
	Expression castExpression = Expression.Convert(objectParameter, objectType);

	// cast the new value to be the correct property type
	Expression newValueCastExpression = Expression.Convert(newValueParameter, propertyInfo.PropertyType);

	// create a new property access expression and assign the cast new value to it
	Expression assignExpression = Expression.Assign(Expression.Property(castExpression, propertyInfo),
			                                        newValueCastExpression);

	return Expression.Lambda<Action<object, object>>(assignExpression, objectParameter, newValueParameter).Compile();
}

Which you would use like this

BusinessObject bo = new BusinessObject { IntProp = 80 };
PropertyInfo propertyInfo = objectType.GetProperty("IntProp",BindingFlags.Public | BindingFlags.Instance);
var setMethod = CreateSetMethod(propertyInfo);
setMethod(bo, 160);
if ((int)getMethod(bo) != 160)
{
	throw new Exception("Value should have been 160");
}

Complex-Example

There is also a more complex example, which uses these demo objects, and allows us to drill deep into an objects nested properties

public class BusinessObject
{
public ChildObjectA A { get; set; }
public int IntProp { get; set; }
public string StringProp { get; set; }
}

public class ChildObjectA
{
public ChildObjectB B { get; set; }
}

public class ChildObjectB
{
public ChildObjectC C { get; set; }
}

public class ChildObjectC
{
public int IntProp { get; set; }
public string StringPro { get; set; }
}

There is no doube that this example is a lot more complicated, as such I will just concentrate on the getter, for nested properties. The bulk of the work is finding and Expression that could turn something like "A.B.C.IntProp" into an actual Expression delegate. As you can imagine we need to recursively examine each of these values that are seperated by a "." and create a Expression portion that represents that part of the property tree expression that we are trying to construct. This part is done using this code:

// <summary>
/// Creates the linq expression the goes down the dotted property chain (A.B.C)
/// It will recursively work it's way down the chain
/// </summary>
/// <param name="valueParameter">the current object in the chain</param>
/// <param name="throwParameter">if thrown parameter</param>
/// <param name="objectType">object type</param>
/// <param name="propertyName">property name (a.b.c)</param>
/// <param name="createAction">function that will be called once the final property is reached. 
/// This the part that is different between get and set</param>
/// <returns></returns>
private static Expression CreateAccessPathCode(
        ParameterExpression valueParameter,
		ParameterExpression throwParameter,
		Type objectType,
		string propertyName,
		Func<Type, string, Expression, Expression> createAction)
{
	Expression returnExpression = null;
	int firstPeriod = propertyName.IndexOf('.');

	// there are more properties in the chain
	if (firstPeriod > 0)
	{
		string currentPropertyName = propertyName.Substring(0, firstPeriod);
		string theRest = propertyName.Substring(firstPeriod + 1);
		Type propertyOrFieldType = GetPropertyOrFieldType(objectType, currentPropertyName);
		// create local variable to hold the property we are about to inspect
		ParameterExpression newValue = Expression.Variable(propertyOrFieldType);

		// assign the property to the local variable
		Expression assignExpression = Expression.Assign(newValue,
			Expression.PropertyOrField(valueParameter, currentPropertyName));
	
		// recurse back up the chain passing the local variable in as the starting point
		// and theRest of the string for the property name
		Expression recurse =
			CreateAccessPathCode(newValue,throwParameter,
				propertyOrFieldType,
				theRest,
				createAction);

		// if the property type is not a value type we need to test it for null
		if (!propertyOrFieldType.GetTypeInfo().IsValueType)
		{
			// expression to create a new Exception object with the message provided
			Expression newExceptionExpression = Expression.New(
				exceptionConstructor,
				Expression.Constant(string.Format("Could not find property {1} on type {0}",
										objectType.FullName,
										currentPropertyName)));




			// create a statement that will throw a new exception if the throwParameter is true
			Expression throwIfMissingTrueExpression =
				Expression.IfThen(Expression.IsTrue(throwParameter),
					Expression.Throw(newExceptionExpression));

			// create if statement that tests the local variable for null
			Expression ifExpression =
				Expression.IfThenElse(
					Expression.NotEqual(newValue, Expression.Constant(null)),
						recurse, throwIfMissingTrueExpression);

			// create a new block containing the new local variable and if expression
			returnExpression =
				Expression.Block(new[] { newValue }, new[] { assignExpression, ifExpression });
		}
		else
		{
			// create a new block containing the recurse statement and local variable
			returnExpression =
				Expression.Block(new[] { newValue }, new[] { assignExpression, recurse });
		}
	}
	else
	{
		// execute create action and return it.
		returnExpression = createAction(objectType, propertyName, valueParameter);
	}
	return returnExpression;
}

The next step is the actual getter. Which looks like this.

public static GetPropertyDelegate CreateGetPropertyDelegate(Type instanceType, string propertyName, Type indexType)
{
	// create input parameters for the method.
	ParameterExpression valueObjectParameter = Expression.Parameter(typeof(object), "valueObject");
	ParameterExpression indexParameter = Expression.Parameter(typeof(object), "index");
	ParameterExpression throwParameter = Expression.Parameter(typeof(bool), "throwIfMissing");

	// define a return variable at the base of 
	ParameterExpression returnValueExpression = Expression.Variable(typeof(object), "returnValue");

	// local variable that is the proper type
	ParameterExpression castValue = Expression.Variable(instanceType, "castValue");
	// assign the value object to the local castValue
	Expression castExpression =
		Expression.Assign(castValue, Expression.Convert(valueObjectParameter, instanceType));
	Expression accessBlock
		= CreateAccessPathCode(castValue,
			throwParameter,
			instanceType,
			propertyName,
			(type, name, expression) =>
			{
				Expression returnValue = null;
				if (indexType != null)
				{
					PropertyInfo propertyInfo = type.GetProperty(propertyName);
					if (propertyInfo != null)
					{
						// assign the property to the returnValue using the index
						returnValue = Expression.Assign(
							returnValueExpression,
							Expression.Convert(
								Expression.Property(expression,
								propertyInfo,
								Expression.Convert(indexParameter, indexType)),
								typeof(object)));
					}
					else
					{
						throw new Exception(
							string.Format("Could not find property {0} on type {1}",
								type.FullName,
								name));
					}
				}
				else
				{
				// assign the final property to the returnValue
					returnValue = Expression.Assign(
						returnValueExpression,
						Expression.Convert(
							Expression.PropertyOrField(expression, name), typeof(object)));
				}

				return returnValue;
			});

	// create body of method
	BlockExpression returnBlock =
		Expression.Block(new[] { castValue, returnValueExpression },
			castExpression,
			accessBlock,
			returnValueExpression);

	return Expression.Lambda<GetPropertyDelegate>(
		returnBlock, valueObjectParameter, indexParameter, throwParameter).
									Compile();
}

Granted this code is quite hairy, but when you think about what it is doing it is quite cool. It is created a delegate that allows you to create a property getter delegate against an objects nested properties, like this:

BusinessObject bo = new BusinessObject
{
	A = new ChildObjectA
	{
		B = new ChildObjectB
		{
			C = new ChildObjectC { IntProp = 90 }
		}
	}
};var getValue =
	ComplexPropertyAccessor.CreateGetPropertyDelegate(typeof(BusinessObject), "A.B.C.IntProp", null);

//returnValue will be 90
var returnValue = getValue(bo, null, true);

It can be seen that this code, actually calls the recursive method CreateAccessPathCodewhere it passes in a callback action that is called when the recursion is over, where the callback will be called with the final object Type, the final property name and the final value parameter. The callback is then used as part of a Expression.Block that is in turn used as the final output that is used to create a LambdaExpression that is finally retuned to the user to use as the property getter delegeate.

Dynamic Where

The idea behind this one is a simple one, which is a very common requirement. We have a IEnumerable<T> where T is some object and we wish to filter it. However we want to build up the filter dynamically.

Say the IEnumerable<T> was made up of Person objetct, so we have IEnumerable<T> where Person looked like this:

public class Person
{
    public string FirstName { get; set; }
    public string LastName { get; set; }
    public int Age { get; set; }
}

If were not buiding up a filter dynamically we could simple do something like this

var x = people.Where(x => x.FirstName="sam" && LastName.Contains("Bree"));

If however we want to build up the filter at runtime based on some input criteria, such as letting the user pick which fields / methods to use in the filter, we would need to come up with a where builder. What we thought was that the where part could be considering its own class, and when we come to build up the entire dynamic where clause filter we would be combining a number of where clause part objects.

The WhereClausePart is shown below:

public enum CompareMethod
{
	Equal,
	GreaterThan,
	GreaterThanOrEqual,
	LessThan,
	LessThanOrEqual,
	Contains,
	StartsWith,
	EndsWith
}


/// <summary>
/// This class represents one part of a complex where statement
/// </summary>
public class WhereClausePart
{
	public WhereClausePart()
	{
			
	}

	public WhereClausePart(string propertyName, CompareMethod compareMethod, object compareValue)
	{
		PropertyName = propertyName;
		CompareMethod = compareMethod;
		CompareValue = compareValue;
	}

	public string PropertyName { get; set; }
	public object CompareValue { get; set; }
	public CompareMethod CompareMethod { get; set; }
}

Which we could then use to construct a complex dynamic where filter using 2 maiin helper methods, namely:

CreateAndWhereClause : Which effectively ANDs all the provided WhereClausePart objects together using the ExpressionAPI Expression.AndAlso
CreateOrWereClause : Which effectively ORs all the provided WhereClausePart objects together using the ExpressionAPI Expression.OrElse

That is what the bulk of the following code is doing. There are also helper methods to do the actual comparison based on the comparison type that was provided as part of the WhereClausePart object(s)

public static class WhereClauseCreator
{
	private static readonly MethodInfo containsMethod;
	private static readonly MethodInfo endsWithMethod;
	private static readonly MethodInfo startsWithMethod;
		
	static WhereClauseCreator()
	{
		containsMethod = typeof(WhereClauseCreator).GetMethod("Contains",BindingFlags.NonPublic | BindingFlags.Static);
		endsWithMethod = typeof(WhereClauseCreator).GetMethod("EndsWith", BindingFlags.NonPublic | BindingFlags.Static);
		startsWithMethod = typeof(WhereClauseCreator).GetMethod("StartsWith", BindingFlags.NonPublic | BindingFlags.Static);
	}

	/// <summary>
	/// Creates a Func(T,bool) that can be used as part of a Where statement in Linq
	/// The comparission statements will be linked using an or operator
	/// </summary>
	/// <typeparam name="T"></typeparam>
	/// <param name="whereClause"></param>
	/// <returns></returns>
	public static Func<T, bool> CreateOrWhereClause<T>(IEnumerable<WhereClausePart> whereClause)
	{
		ParameterExpression inParameter = Expression.Parameter(typeof(T));
		Expression whereStatement = null;

		// foreach part of the where clause create the compare expression then or it with the current where statement
		foreach (WhereClausePart part in whereClause)
		{
			Expression propertyEqualStatement = CompareExpression<T>(inParameter, part);
			if (whereStatement == null)
			{
				whereStatement = propertyEqualStatement;
			}
			else
			{
				whereStatement = Expression.OrElse(whereStatement, propertyEqualStatement);
			}
		}

		if (whereStatement == null)
		{
			whereStatement = Expression.Constant(true);
		}
		return Expression.Lambda<Func<T, bool>>(whereStatement, inParameter).Compile();
	}

	public static Func<T, bool> CreateAndWhereClause<T>(IEnumerable<WhereClausePart> whereClause)
	{
		ParameterExpression inParameter = Expression.Parameter(typeof(T));
		Expression whereStatement = null;




		foreach (WhereClausePart part in whereClause)
		{
			Expression propertyEqualStatement = CompareExpression<T>(inParameter, part);
			if (whereStatement == null)
			{
				whereStatement = propertyEqualStatement;
			}
			else
			{
				whereStatement = Expression.AndAlso(whereStatement, propertyEqualStatement);
			}
		}
		if (whereStatement == null)
		{
			whereStatement = Expression.Constant(true);
		}
		return Expression.Lambda<Func<T, bool>>(whereStatement, inParameter).Compile();
	}

	/// <summary>
	/// Creates an expression where it compares a property to a particular value
	/// </summary>
	/// <typeparam name="T"></typeparam>
	/// <param name="inParameter"></param>
	/// <param name="part"></param>
	/// <returns></returns>
	private static Expression CompareExpression<T>(ParameterExpression inParameter, WhereClausePart part)
	{
		PropertyInfo propertyInfo = typeof(T).GetProperty(part.PropertyName);
		// based on the comparission method we change the Expression we return back
		switch (part.CompareMethod)
		{
			case CompareMethod.Equal:
				return Expression.Equal(Expression.Property(inParameter, propertyInfo), 
					Expression.Constant(part.CompareValue));
			case CompareMethod.GreaterThan:
				return Expression.GreaterThan(Expression.Property(inParameter, propertyInfo), 
					Expression.Constant(part.CompareValue));
			case CompareMethod.GreaterThanOrEqual: 
				return Expression.GreaterThanOrEqual(Expression.Property(inParameter, propertyInfo), 
					Expression.Constant(part.CompareValue));
			case CompareMethod.LessThan:
				return Expression.LessThan(Expression.Property(inParameter, propertyInfo), 
					Expression.Constant(part.CompareValue));
			case CompareMethod.LessThanOrEqual:
				return Expression.LessThanOrEqual(Expression.Property(inParameter, propertyInfo), 
					Expression.Constant(part.CompareValue));
			case CompareMethod.Contains:
				return Expression.Call(containsMethod,
					Expression.Property(inParameter, propertyInfo),
					Expression.Constant(part.CompareValue));
			case CompareMethod.StartsWith:
				return Expression.Call(startsWithMethod,
				Expression.Property(inParameter, propertyInfo),
					Expression.Constant(part.CompareValue));
			case CompareMethod.EndsWith:
				return Expression.Call(endsWithMethod,
					Expression.Property(inParameter, propertyInfo),
					Expression.Constant(part.CompareValue));

		}
		return null;
	}

	/// <summary>
	/// I've created a wrapper around Contains so that I don't need to check for null in the linq expression 
	/// </summary>
	/// <param name="testString"></param>
	/// <param name="containsValue"></param>
	/// <returns></returns>
	private static bool Contains(string testString, string containsValue)
	{
		if (testString != null)
		{
			return testString.Contains(containsValue);
		}
		return false;
	}


	/// <summary>
	/// I've created a wrapper around EndsWith so that I don't need to check for null in the linq expression 
	/// </summary>
	/// <param name="testString"></param>
	/// <param name="endsWithValue"></param>
	/// <returns></returns>
	private static bool EndsWith(string testString, string endsWithValue)
	{
		if (testString != null)
		{
			return testString.EndsWith(endsWithValue);
		}
		return false;
	}

	/// <summary>
	/// I've created a wrapper around StartsWith so that I don't need to check for null in the linq expression 
	/// </summary>
	/// <param name="testString"></param>
	/// <param name="startsWithValue"></param>
	/// <returns></returns>
	private static bool StartsWith(string testString, string startsWithValue)
	{
		if (testString != null)
		{
			return testString.StartsWith(startsWithValue);
		}

		return false;
	}
}

The actual usage of this may look something like this:

List<Person> persons = CreatePersonList();
Func<Person, bool> clause =
	WhereClauseCreator.CreateAndWhereClause<Person>(new[]
		{
			new WhereClausePart("LastName",CompareMethod.Equal, "Doe"),
 			new WhereClausePart("Age",CompareMethod.GreaterThan, 20) 
		});

if (persons.Where(clause).Count() != 4)
{
	throw new Exception("There should be only 4.");
}

Func<Person, bool> orClause =
	WhereClauseCreator.CreateOrWhereClause<Person>(new[]
		{
			new WhereClausePart("LastName",CompareMethod.Equal, "Johnson"),
 			new WhereClausePart("FirstName",CompareMethod.StartsWith, "J") 
		});

if (persons.Where(orClause).Count() != 8)
{
	throw new Exception("Result should be 8.");
}

That's It

Anayway hope you enjoyed this one. We both feel that this sort of Expression API cookbook approach can be a pretty useful way of learning more about the Expression API. I personally salute Ian for his Expression skills, which far outway my own, and we both hope that this article may prove useful to some of you. I know I have learned quite a bit just from working through Ians examples.

As always if you enjoyed the article, feel free to leave a comment/vote.

Expression API Cookbook