What's the framework mechanism behind dependency properties?

Question

I have been reading about dependency properties in several books but all have one thing in common, they just tell us how they are implemented( using static readonly DependencyProperty etc.) but does not tell the exact way they work from inside.

I mean they are implemented as static but still applies to all objects.
Second point of confusion is attached properties.

Is there any tutorial available that can explain all these concepts in an easy way?

Answer 1

My mental model of how dependency properties work:

Any DependencyObject class implements two special properties. One, a static property of the class, is a dictionary of DependencyProperty objects. Every instance of the class can look inside that dictionary to find metainformation about each DependencyProperty - the property's name, its type, any callbacks that have to be called when it's get and set, how it participates in property inheritance, and so on. When you register a dependency property, you're adding an entry to this dictionary.

The other property is an instance property: it's a dictionary, keyed by DependencyProperty, that contains the local value of each DependencyProperty, if it has been set.

The SetValue and GetValue methods that you implement in the setter and getter of the CLR property are basically lazy evaluation on steroids. Instead of storing and retrieving the value of the property in a backing field, they store and retrieve the value of the property in the value dictionary.

The magic of dependency properties is in what GetValue and SetValue actually do.

GetValue looks up the value for the property in the object's value dictionary. If it doesn't find it, it calls GetValue on the parent element, to get what the parent element thinks the value is. For instance, when you create a TextBox in a Window, anything that looks at the TextBox's FontFamily is actually calling GetValue. Unless you've explicitly set the font, there's no entry in its dictionary for that property. So GetValue asks the parent element for the value. The parent element may or may not have FontFamily set; if not, its call to GetValue to returns the value from its parent. And so on, until the Window object is reached and the actual FontFamily value is found.

If you set FontFamily on the TextBox, SetValue stores the value in the value dictionary. The next time anything needs to get the value of the FontFamily for that TextBox, GetValue finds the value in the dictionary and returns it, so it doesn't need to ask the parent element.

If you set FontFamily on the Window, SetValue not only updates the value in Window's value dictionary, it fires off a property-change event that everything dependent on the property hears. (That's why they're called dependency properties, remember.) And if the thing depending on the property is itself a dependency property, it fires off its own property-change events. This is how it is that changing the FontFamily on the Window changes the font for every control in the window and also prompts WPF to re-render the controls that have changed.

Attached properties work using the same kind of approach. Any object that can have attached properties has a dictionary that the attached properties' values are stored in. When you set Grid.Column on a CheckBox in XAML, you're just adding an entry to that CheckBox's dictionary. When the Grid needs to know what column the CheckBox is in, it looks the value up from that dictionary. When you set Grid.IsSharedSizeScope to True on an object, that object's dictionary will contain a new property - a dictionary that contains widths/heights for each SharedSizeKey.

I should emphasize that this is my mental model. I haven't sat down with Reflector and looked at the actual implementation of Register, GetValue, and SetValue to figure out how they actually work. I may be wrong about the details. But it's a model that accurately predicts how this stuff behaves, so it's good enough.

The concept of storing property values in dictionaries is pretty weird to C# programmers. It's old hat to Python programmers, though. In Python, all class properties - all objects, in fact - are stored in dictionaries, and so you can get to their value either through property accessors or just by looking them up. Dependency properties and attached properties are just another way in which .NET, having stolen everything Java had that was worth stealing, is now plundering Python. (Or from wherever Python plundered them from.) Learning Python has made me a much better C# programmer; I recommend it to any C# developer who hasn't done it yet.

Answer 2

Here is a tutorial on dependency properties http://www.wpftutorial.net/DependencyProperties.html that explains a little bit about how they work.

The short explanation of why the DependencyProperty object is in a static field is that it represents the description of the property, not the value of the property. Each DependencyObject has a mapping from DependencyProperty objects to their values.

This is also how attached properties work. Because each DependencyObject is storing a mapping from any DependencyProperty to a value, any type can create a new DependencyProperty and set it on any existing DependencyObject.