Skip to content
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


This is my xaml compiler, being written as part of Google's summer of code

Xaml is a language for concisely describing how to connect up a bunch of 
objects. As an example, in the avalon world:
<Window Title="Hi!">

Would produce a child of Window where the constructor does this:
this.Title = "Hi!";
Button button = new Button();

Except of course it can handle trees of arbitrary depth, so descriptions of
complicated GUIs are not a problem and potentially much more readable than 
the code version.

To use this thing:
cd ../../class/WindowsBase
make install
cd ../PresentationFramework
make install
cd ../../tools/xamlc
make install

That will get everything compiled. There's a demo in the mcs/tools/xamlc/demo
directory; to run it, cd into that directory and do "make run". Stuff should 
happen, along the lines of a program test.exe being generated and run to 
produce this output:

Hello World
Hello World
YYYI'm surrounded!ZZZ
What should I say?

I say.


In the same directory, "make run2" should demonstrate the runtime creation of
objects from xaml files.

How it works
The compiler's main job is to identify the types to which your xaml file refers
and use that to ensure that correct types appear in the output xaml file. This 
is also an issue when performing creation of objects from the xaml file at 
runtime because of the need to perform conversions between types.

The following examples are given in terms of objects being instantiated directly
by the parser at runtime; if this is not your chosen approach, then things work
somewhat differently. The former approach, of calling Parser.LoadXml(), will
create the object at runtime, while the latter approach of invoking xamlc will
create a source code file representing the object that would otherwise have
been created at runtime. The process by which this functions is that the class
is defined as inheriting the class being represented by the top-level element of
the xml tree, and every other object mentioned is either a variable local to the
function or a field of the class.

A xaml file is an Xml file where meaning is derived from processing 
instructions, text, elements, and attributes. Elements typically indicate class
instances or properties, attributes refer mainly to properties, and processing
instructions are used to allow types to be resolved correctly.

In order to facilitate the correct resolution of types, every element must have
a specified XML namespace. The XML namespace is an arbitrary string that has
been defined as the identifier for an Assembly-ClrNamespace pair. The 
definition of these pairs is achieved with processing instructions (PIs). A line
<?Mapping ClrNamespace="Org.Project" Assembly="myassembly.dll" XmlNamespace="cc" ?>

indicates that every element in the cc namespace corresponds to a type defined 
in myassembly.dll in the Org.Project namespace. There is substantial similarity
of purpose between the Mapping PI and C#'s using statement, but the Mapping PI
is substantially more precise.

Elements representing class instances in XAML files are required to have the
name of the class to be instantiated as the element name. The top-level element
must therefore be named according to what kind of object you wish to be produced
by the compilation process. For example, if "XX" is defined (in the manner
described above) as corresponding to an assembly where the relevant namespace
defines a Window class, the xaml document <x:Window xmlns:x="XX" /> will have a
similar effect to the C# code "new Window()".

If Xaml is to succeed in its goal of simplifying code in the style of that which
is commonly used to initialize GUIs, it must have a way of adding objects as
children to other objects. This maps to child elements in Xml - the standard
case is for a child element of an object element to represent an object which 
is a child of the aforementioned object. The notion of a child object is defined
by the System.Windows.Serialization.IAddChild interface, which provides an
AddChild method for making objects children of the implementor of the interface.

We can therefore see that given a namespace defined similarly to that in the
previous example containing Window and Button objects that the following

<x:Window xmlns:x="XX">
	<Button />
	<Button />

Is equivalent to the following C# code:
Window w = new Window();
((System.Windows.IAddChild)w).AddChild(new Button());
((System.Windows.IAddChild)w).AddChild(new Button());

The same mechanism also allows for the addition of text. The document:
<x:Button xmlns:x="XX">I Am Button.</x:Button>

Is equivalent to:
Button b = new Button();
((System.Windows.IAddChild)b).AddText("I Am Button.");

So far, we've seen mechanisms for type resolution, making an element a child of
another one, and adding text for an element. We still need mechanisms for 
setting properties and event handlers.

The basic mechanism by which properties can be set is the xml attribute. The 
name of the attribute corresponds to the name of the property, and the value is
converted from a string with .net's standard TypeConverter framework and 
assigned to the property. If our hypothetical Window class has a property 
Title of type String, we could do something like the following:
<x:Window Title="The title of a window" xmlns:x="XX" />

To produce code equivalent to
Window w = new Window();
w.Title = "The title of a window";

While if Window defined a SecondsToDestruction property of type Int32, code like
<x:Window SecondsToDestruction="5" xmlns:x="XX" />

To produce code equivalent to:
Window w = new Window();
w.SecondsToDestruction = (new Int32Converter()).ConvertFromString("5");

This system means that any property, the type of which has a converter allowing
instances to be created from strings, can be assigned to in your Xaml code.

While this is sufficiently general for many cases, it is impractical for cases
where you are unable for whatever reason to express the property value as a
string. For cases like these the complex property syntax is provided. In order
to use this syntax, you define an element as a child of the element representing
the object the property of which you want to set. The elements name should be 
the name of the class with the property as member, followed by a ., followed by
the name of the property. Continuing on with our Window example, if the Window
had a Thumbnail property we might do something like this:
<x:Window xmlns:x="XX">
		<Button />

In order to produce code equivalent to:
Window w = new Window();
w.Thumbnail = new Button();

This syntax allows arbitrarily complex objects to be used as property values.

The syntax for events follows the syntax for properties. In order to add an
event handler, all that is neccesary is to add an attribute to the object's
element with the name of the event as its name and the name of the event handler
as its value. For example, if the Window has an OnClose event:
<x:Window OnClose="close_handler" xmlns:x="XX" />

Would correspond to code like:
Window w = new Window();
w.OnClose += close_handler;
Something went wrong with that request. Please try again.