Introduction to Exception Handling

Exceptions are unforeseen errors that happen in your programs. Most of the time, you can, and should, detect and handle program errors in your code. For example, validating user input, checking for null objects, and verifying the values returned from methods are what you expect, are all examples of good standard error handling that you should be doing all the time.

            However, there are times when you don't know if an error will occur. For example, you can't predict when you'll receive a file I/O error, run out of system memory, or encounter a database error. These things are generally unlikely, but they could still happen and you want to be able to deal with them when they do occur. This is where exception handling comes in.

            When exceptions occur, they are said to be "thrown". What is actually thrown is an object that is derived from the System.Exception class. In the next section, I'll be explaining how thrown exceptions are handled with try/catch blocks.

            The System.Exception class provides several methods and properties for obtaining information on what went wrong. For example, the Message property provides summary information about what the error was, the Stacktrace property provides information from the stack for where the problem occurred, and the ToString() method is overridden to reveal a verbose description of the entire exception.

            Identifying the exceptions you'll need to handle depends on the routine you're writing. For example, if the routine opened a file with the System.IO.File.OpenRead() method, it could throw any of the following exceptions:

SecurityException

ArgumentException

ArgumentNullException

PathTooLongException

DirectoryNotFoundException

UnauthorizedAccessException

FileNotFoundException

NotSupportedException

            It's easy to find out what exceptions a method can raise by looking in the .NET Frameworks SDK Documentation. Just go to the Reference/Class Library section and look in the Namespace/Class/Method documentation for the methods you use. The exception in the list above were found by looking at the OpenRead() method definition of the File class in the System.IO namespace. Each exception identified has a hyperlink to its class definition that you can use to find out what that exception is about. Once you've figured out what exceptions can be generated in your code, you need to put the mechanisms in place to handle the exceptions, should they occur.

try/catch Blocks

When exceptions are thrown, you need to be able to handle them. This is done by implementing a try/catch block. Code that could throw an exception is put in the try block an exception handling code goes in the catch block. Listing shows how to implement a try/catch block. Since an OpenRead() method could throw one of several exceptions, it is placed in the try block. If an exception is thrown, it will be caught in the catch block. The code in Listing will print message and stack trace information out to the console if an exception is raised.

using System;
using System.IO;

class tryCatchDemo
{
            static void Main(string[] args)
            {
                        try                        {
                                    File.OpenRead("NonExistentFile");
                         }
                        catch(Exception ex)
                        {
                                    Console.WriteLine(ex.ToString());
                         }
            }
}

Although the code in Listing only has a single catch block, all exceptions will be caught there because the type is of the base exception type "Exception". In exception handling, more specific exceptions will be caught before their more general parent exceptions. For example, the following snippet shows how to place multiple catch blocks:

catch(FileNotFoundException fnfex)
 {
            Console.WriteLine(fnfex.ToString());
 }
 catch(Exception ex)
 {
            Console.WriteLine(ex.ToString());
 }

            If the file doesn't exist, a FileNotFoundException exception will be thrown and caught by the first catch block. However, if a PathTooLongException exception was raised, the second catch part would catch the exception. This is because there isn't a catch block for the PathTooLongException exception and the generic Exception type catch block is the only option available to catch the exception.

            Exceptions that are not handled will normally bubble up the stack until a calling routine in the call chain handles them. If you forget to include try/catch blocks in a part of your code and there aren't any try/catch blocks earlier in the call chain, your program will abort with a message describing the exception. To your users this would be very cryptic and uncomfortable. It is good practice to provide exception handling in your programs

Finally Blocks

            An exception can leave your program in an inconsistent state by not releasing resources or doing some other type of cleanup. A catch block is a good place to figure out what may have went wrong and try to recover, however it can't account for all scenarios. Sometimes you need to perform clean up actions whether or not your program succeeds. These situations are good candidates for using a finally block.

            Listing illustrates the usefulness of a finally block. As you know, a file stream must be closed when you're done with it. In this case, the file stream is the resource that needs to be cleaned up. In Listing, outStream is opened successfully, meaning the program now has a handle to an open file resource. When trying to open the inStream, a FileNotFoundException exception is raised, causing control to go immediately to the catch block.

            It's possible to close the outStream in the catch block, but what if the algorithm executed successfully without an exception?  On success, the file would never be closed. Fortunately, we've included a finally block in Listing which will always be executed. That's right, regardless of whether the algorithm in the try block raises an exception or not, the code in the finally block will be executed before control leaves the method.

using System;
using System.IO;

class FinallyDemo
{
 static void Main(string[] args)
 {
            FileStream outStream = null;
            FileStream inStream = null;

            try            {
                        outStream = File.OpenWrite("DestinationFile.txt");
                        inStream = File.OpenRead("BogusInputFile.txt");
            }
            catch(Exception ex)
            {
                        Console.WriteLine(ex.ToString());
            }
            finally            {
                        if (outStream != null)
                        {
                                    outStream.Close();
                                    Console.WriteLine("outStream closed.");
                        }
                        if (inStream != null)
                        {
                                    inStream.Close();
                                    Console.WriteLine("inStream closed.");
                        }
            }
 }
}

            A finally block is not required and you may ask what happens if you just put code after the catch block. true, under normal circumstances, if the exception is caught, all code following the catch will be executed. However, try/catch/finally is for exceptional circumstances and it is better to plan for the worst to make your program more robust. For example, if one of the catch handlers rethrew an exception or caused another exception, the code following the catch block (not in a finally block) would never be executed. Also, if you don't catch the exception at all, program flow would immediately do a stack walk looking for an exception handler that fits and the code following the catch blocks would not be executed. Since there is too much potential for code in an algorithm to not be executed, a finally block is your insurance for executing those critical actions you need.