1.17. Exception Handling

That pretty much wraps up the WordCount program—at least in terms of its functionality. The primary remaining issue is that of error detection. For example, what if the user has specified a file that doesn't exist? Or what if the file is in a format we don't currently support? Checking that is simple. To find out if a file exists, we can invoke the Exists() function of the File class, passing to it the string supplied to us by the user:

using System.IO;
if ( ! File. Exists( file_name ))
      // oops ...

The harder part is choosing how to handle and/or report the problem. In the .NET environment, the convention is to report all program anomalies through exception handling. And that is what we will do.

Exception handling consists of two primary components: (1) the recognition and raising of an exception through a throw expression and (2) the handling of the exception within a catch clause. Here is a throw expression:

public StreamReader openFile( string file_name )
{
   if ( file_name == null )
        throw new ArgumentNullException();

   // reach here only if no ArgumentNullException thrown
   if ( ! File.Exists( file_name ))
   {
          string msg = "Invalid file name: " + file_name;
          throw new ArgumentException( msg );
   }

   // reach here if file_name not null and file exists
   if ( ! file_name.EndsWith( ".txt" ))
   {
          string msg = "Sorry. ";
          string ext = Path.GetExtension( file_name );

          if ( ext != String.Empty )
                msg += "We currenly do not support " +
                             ext + " files."

          msg = "
Currenly we only support .txt files.";
          throw new Exception( msg );
   }

   // OK: here only if no exceptions thrown
   return File.OpenText( file_name );
}

The object of a throw expression is always an instance of the Exception class hierarchy. (We look at inheritance and class hierarchies in our discussion of object-oriented programming in Chapter 3.) The Exception class is defined in the System namespace. It is initialized with a string message identifying the nature of the exception. The ArgumentException class represents a subtype of the Exception class. It more precisely identifies the category of exception. The ArgumentNullException class is in turn a subtype of ArgumentException. It is the most specific of these three exception objects.

Once an exception has been thrown, normal program execution is suspended. The exception-handling facility searches the method call chain in reverse order for a catch clause capable of handling the exception.

We handle an exception by matching the type of the exception object using one or a series of catch clauses. A catch clause consists of three parts: the keyword catch, the declaration of the exception type within parentheses, and a set of statements within curly braces that actually handle the exception.

catch clauses are associated with try blocks. A try block begins with the try keyword, followed by a sequence of program statements enclosed within curly braces. The catch clauses are positioned at the end of the try block. For example, consider the following code sequence:

StreamReader freader = openFile( fname );
string textline;

while (( textline = freader.ReadLine() ) != null )

We know that openFile() throws three possible exceptions. ReadLine() throws one exception, that of the IOException class. As written, this code does not handle any of those four exceptions. To correct that, we place the code inside a try block and associate the relevant set of catch clauses:

try
{
   StreamReader freader = openFile( fname );
   string textline;

   while (( textline = freader.ReadLine() ) != null )
   {
         // do the work here ...
   }
   catch ( IOException ioe )
   { ... }

   catch ( ArgumentNullException ane )
   { ... }

   catch ( ArgumentException ae )
   { ... }

   catch ( Exception e )
   { ... }

What happens when an exception is thrown? The exception-handling mechanism looks at the site of the throw expression and asks, “Has this occurred within a try block?” If it has, the type of the exception object is compared against the exception type declaration of each associated catch clause in turn. If the types match, the body of the catch clause is executed.

This represents a complete handling of the exception, and normal program execution resumes. If the catch clause does not specify a return statement, execution begins again at the first statement following the set of catch clauses. Execution does not resume at the point where the exception was thrown.

What if the type of the exception object does not match one of the catch clauses or if the code does not occur within a try block? The currently executing function is terminated, and the exception-handling mechanism resumes its search within the function that invoked the function just terminated.

If one of the three if statements of openFile() throws an exception, the assignment of freader and the remainder of the try block are not executed. Rather the exception-handling mechanism assumes program control, examining each associated catch clause in turn, trying to match the exception type.

What if the chain of function calls is unwound to the Main() program entry point, and still no appropriate catch clause is found? The program itself is then terminated. The unhandled exception is propagated to the runtime debugger. The user can either debug the program or let it prematurely terminate.

In addition to a set of catch clauses, we can place a finally block after the last catch clause. The code associated with the finally block is always executed before the function exits:

• If an exception is handled, first the catch clause and then the finally clause is executed before normal program execution resumes.

• If an exception occurs but there is no matching catch clause, the finally clause is executed; the remainder of the function is discarded.

• If no exception occurs, the function terminates normally. The finally clause is executed following the last non-return statement of the function.

The primary use of the finally block is to reduce code duplication due to differing exception/no-exception exit points that need to execute the same code.