Compiling C

Running your C# code through the C# compiler produces two important pieces of information: code and metadata. The following sections describe these two items and then finish up by examining the binary building block of .NET code: the assembly.

Microsoft Intermediate Language (MSIL)

The code that is output by the C# compiler is written in a language called Microsoft Intermediate Language, or MSIL. MSIL is made up of a specific set of instructions that specify how your code should be executed. It contains instructions for operations such as variable initialization, calling object methods, and error handling, just to name a few. C# is not the only language in which source code changes into MSIL during the compilation process. All .NET-compatible languages, including Visual Basic .NET and Managed C++, produce MSIL when their source code is compiled. Because all of the .NET languages compile to the same MSIL instruction set, and because all of the .NET languages use the same runtime, code from different languages and different compilers can work together easily.

MSIL is not a specific instruction set for a physical CPU. It knows nothing about the CPU in your machine, and your machine knows nothing about MSIL. How, then, does your .NET code run at all, if your CPU can't read MSIL? The answer is that the MSIL code is turned into CPU-specific code when the code is run for the first time. This process is called "just-in-time" compilation, or JIT. The job of a JIT compiler is to translate your generic MSIL code into machine code that can be executed by your CPU.

You may be wondering about what seems like an extra step in the process. Why generate MSIL when a compiler could generate CPU-specific code directly? After all, compilers have always done this in the past. There are a couple of reasons for this. First, MSIL enables your compiled code to be easily moved to different hardware. Suppose you've written some C# code and you'd like it to run on both your desktop and a handheld device. It's very likely that those two devices have different types of CPUs. If you only had a C# compiler that targeted a specific CPU, then you'd need two C# compilers: one that targeted your desktop CPU and another that targeted your handheld CPU. You'd have to compile your code twice, ensuring that you put the right code on the right device. With MSIL, you compile once. Installing the .NET Framework on your desktop machine includes a JIT compiler that translates your MSIL into CPU-specific code for your desktop. Installing the .NET Framework on your handheld includes a JIT compiler that translates that same MSIL into CPU-specific code for your handheld. You now have a single MSIL code base that can run on any device that has a .NET JIT compiler. The JIT compiler on that device takes care of making your code run on the device.

Another reason for the compiler's use of MSIL is that the instruction set can be easily read by a verification process. Part of the job of the JIT compiler is to verify your code to ensure that it is as clean as possible. The verification process ensures that your code is accessing memory properly and that it is using the correct variable types when calling methods that expect a specific type. These checks ensure that your code doesn't execute any instructions that could make the code crash. The MSIL instruction set was designed to make this verification process relatively straightforward. CPU-specific instruction sets are optimized for quick execution of the code, but they produce code that can be hard to read and, therefore, hard to verify. Having a C# compiler that directly outputs CPU-specific code can make code verification difficult or even impossible. Allowing the .NET Framework JIT compiler to verify your code ensures that your code accesses memory in a bug-free way and that variable types are properly used.


The compilation process also outputs metadata, which is an important piece of the .NET codesharing story. Whether you use C# to build an end-user application or you use C# to build a class library to be used by someone else's application, you're going to want to make use of some already-compiled .NET code. That code may be supplied by Microsoft as a part of the .NET Framework, or it may be supplied by a user over the Internet. The key to using this external code is letting the C# compiler know what classes and variables are in the other code base so that it can match up the source code you write with the code found in the precompiled code base that you're working with.

Think of metadata as a "table of contents" for your compiled code. The C# compiler places metadata in the compiled code along with the generated MSIL. This metadata accurately describes all the classes you wrote and how they are structured. All of the classes' methods and variable information is fully described in the metadata, ready to be read by other applications. Visual Basic .NET, for example, may read the metadata for a .NET library to provide the IntelliSense capability of listing all of the methods available for a particular class.

If you've ever worked with COM (Component Object Model), you may be familiar with type libraries. Type libraries aimed to provide similar "table of contents" functionality for COM objects. However, type libraries suffered from some limitations, not the least of which was the fact that not all of the data relevant to the object was put into the type library. Metadata in .NET does not have this shortcoming. All of the information needed to describe a class in code is placed into the metadata. You can think of metadata as having all of the benefits of COM type libraries without the limitations.


Sometimes, you will use C# to build an end-user application. These applications are packaged as executable files with an extension of .EXE. Windows has always worked with .EXE files as application programs, and C# fully supports building .EXE files.

However, there may be times when you don't want to build an entire application. Instead, you may want to build a code library that can be used by others. You may also want to build some utility classes in C#, for example, and then hand the code off to a Visual Basic .NET developer, who will use your classes in a Visual Basic .NET application. In cases like this, you won't be building an application. Instead, you'll be building an assembly.

An assembly is a package of code and metadata. When you deploy a set of classes in an assembly, you are deploying the classes as a unit; and those classes share the same level of version control, security information, and activation requirements. Think of an assembly as a "logical DLL." If you're familiar with Microsoft Transaction Server or COM+, you can think of an assembly as the .NET equivalent of a package.

There are two types of assemblies: private assemblies and global assemblies. When you build your assembly, you don't need to specify whether you want to build a private or a global assembly. The difference is apparent when you deploy your assembly. With a private assembly, you make your code available to a single application. Your assembly is packaged as a DLL, and is installed into the same directory as the application using it. With a deployment of a private assembly, the only application that can use your code is the executable that lives in the same directory as your assembly.

If you want to share your code among many applications, you might want to consider deploying your code as a global assembly. Global assemblies can be used by any .NET application on the system, regardless of the directory in which it is installed. Microsoft ships assemblies as a part of the .NET Framework, and each of the Microsoft assemblies is installed as a global assembly. The .NET Framework contains a list of global assemblies in a facility called the global assembly cache, and the .NET Microsoft Framework SDK includes utilities to both install and remove assemblies from the global assembly cache.

0 0

Post a comment