If an error occurs in the DECLARE section of a block immediately the enclosing block is searched

Remarks

This class is the base class for all exceptions. When an error occurs, either the system or the currently executing application reports it by throwing an exception that contains information about the error. After an exception is thrown, it is handled by the application or by the default exception handler.

In this section:

Errors and exceptions
Try/catch blocks
Exception type features
Exception class properties
Performance considerations
Re-throwing an exception
Choosing standard exceptions
Implementing custom exceptions

Errors and exceptions

Run-time errors can occur for a variety of reasons. However, not all errors should be handled as exceptions in your code. Here are some categories of errors that can occur at run time and the appropriate ways to respond to them.

• Usage errors. A usage error represents an error in program logic that can result in an exception. However, the error should be addressed not through exception handling but by modifying the faulty code. For example, the override of the Object.Equals(Object) method in the following example assumes that the obj argument must always be non-null.

  using System; public class Person { private string _name; public string Name { get { return _name; } set { _name = value; } } public override int GetHashCode() { return this.Name.GetHashCode(); } public override bool Equals(object obj) { // This implementation contains an error in program logic: // It assumes that the obj argument is not null. Person p = (Person) obj; return this.Name.Equals(p.Name); } } public class Example { public static void Main() { Person p1 = new Person(); p1.Name = "John"; Person p2 = null; // The following throws a NullReferenceException. Console.WriteLine("p1 = p2: {0}", p1.Equals(p2)); } } // In F#, null is not a valid state for declared types // without 'AllowNullLiteralAttribute' [] type Person() = member val Name = "" with get, set override this.GetHashCode() = this.Name.GetHashCode() override this.Equals(obj) = // This implementation contains an error in program logic: // It assumes that the obj argument is not null. let p = obj :?> Person this.Name.Equals p.Name let p1 = Person() p1.Name <- "John" let p2: Person = null // The following throws a NullReferenceException. printfn $"p1 = p2: {p1.Equals p2}" Public Class Person Private _name As String Public Property Name As String Get Return _name End Get Set _name = value End Set End Property Public Overrides Function Equals(obj As Object) As Boolean ' This implementation contains an error in program logic: ' It assumes that the obj argument is not null. Dim p As Person = CType(obj, Person) Return Me.Name.Equals(p.Name) End Function End Class Module Example Public Sub Main() Dim p1 As New Person() p1.Name = "John" Dim p2 As Person = Nothing ' The following throws a NullReferenceException. Console.WriteLine("p1 = p2: {0}", p1.Equals(p2)) End Sub End Module

  The NullReferenceException exception that results when obj is null can be eliminated by modifying the source code to explicitly test for null before calling the Object.Equals override and then re-compiling. The following example contains the corrected source code that handles a null argument.

  using System; public class Person { private string _name; public string Name { get { return _name; } set { _name = value; } } public override int GetHashCode() { return this.Name.GetHashCode(); } public override bool Equals(object obj) { // This implementation handles a null obj argument. Person p = obj as Person; if (p == null) return false; else return this.Name.Equals(p.Name); } } public class Example { public static void Main() { Person p1 = new Person(); p1.Name = "John"; Person p2 = null; Console.WriteLine("p1 = p2: {0}", p1.Equals(p2)); } } // The example displays the following output: // p1 = p2: False // In F#, null is not a valid state for declared types // without 'AllowNullLiteralAttribute' [] type Person() = member val Name = "" with get, set override this.GetHashCode() = this.Name.GetHashCode() override this.Equals(obj) = // This implementation handles a null obj argument. match obj with | :? Person as p -> this.Name.Equals p.Name | _ -> false let p1 = Person() p1.Name <- "John" let p2: Person = null printfn $"p1 = p2: {p1.Equals p2}" // The example displays the following output: // p1 = p2: False Public Class Person Private _name As String Public Property Name As String Get Return _name End Get Set _name = value End Set End Property Public Overrides Function Equals(obj As Object) As Boolean ' This implementation handles a null obj argument. Dim p As Person = TryCast(obj, Person) If p Is Nothing Then Return False Else Return Me.Name.Equals(p.Name) End If End Function End Class Module Example Public Sub Main() Dim p1 As New Person() p1.Name = "John" Dim p2 As Person = Nothing Console.WriteLine("p1 = p2: {0}", p1.Equals(p2)) End Sub End Module ' The example displays the following output: ' p1 = p2: False

  Instead of using exception handling for usage errors, you can use the Debug.Assert method to identify usage errors in debug builds, and the Trace.Assert method to identify usage errors in both debug and release builds. For more information, see Assertions in Managed Code.

• Program errors. A program error is a run-time error that cannot necessarily be avoided by writing bug-free code.

  In some cases, a program error may reflect an expected or routine error condition. In this case, you may want to avoid using exception handling to deal with the program error and instead retry the operation. For example, if the user is expected to input a date in a particular format, you can parse the date string by calling the DateTime.TryParseExact method, which returns a Boolean value that indicates whether the parse operation succeeded, instead of using the DateTime.ParseExact method, which throws a FormatException exception if the date string cannot be converted to a DateTime value. Similarly, if a user tries to open a file that does not exist, you can first call the File.Exists method to check whether the file exists and, if it does not, prompt the user whether they want to create it.

  In other cases, a program error reflects an unexpected error condition that can be handled in your code. For example, even if you've checked to ensure that a file exists, it may be deleted before you can open it, or it may be corrupted. In that case, trying to open the file by instantiating a StreamReader object or calling the Open method may throw a FileNotFoundException exception. In these cases, you should use exception handling to recover from the error.

• System failures. A system failure is a run-time error that cannot be handled programmatically in a meaningful way. For example, any method can throw an OutOfMemoryException exception if the common language runtime is unable to allocate additional memory. Ordinarily, system failures are not handled by using exception handling. Instead, you may be able to use an event such as AppDomain.UnhandledException and call the Environment.FailFast method to log exception information and notify the user of the failure before the application terminates.

Try/catch blocks

The common language runtime provides an exception handling model that is based on the representation of exceptions as objects, and the separation of program code and exception handling code into try blocks and catch blocks. There can be one or more catch blocks, each designed to handle a particular type of exception, or one block designed to catch a more specific exception than another block.

If an application handles exceptions that occur during the execution of a block of application code, the code must be placed within a try statement and is called a try block. Application code that handles exceptions thrown by a try block is placed within a catch statement and is called a catch block. Zero or more catch blocks are associated with a try block, and each catch block includes a type filter that determines the types of exceptions it handles.

When an exception occurs in a try block, the system searches the associated catch blocks in the order they appear in application code, until it locates a catch block that handles the exception. A catch block handles an exception of type T if the type filter of the catch block specifies T or any type that T derives from. The system stops searching after it finds the first catch block that handles the exception. For this reason, in application code, a catch block that handles a type must be specified before a catch block that handles its base types, as demonstrated in the example that follows this section. A catch block that handles System.Exception is specified last.

If none of the catch blocks associated with the current try block handle the exception, and the current try block is nested within other try blocks in the current call, the catch blocks associated with the next enclosing try block are searched. If no catch block for the exception is found, the system searches previous nesting levels in the current call. If no catch block for the exception is found in the current call, the exception is passed up the call stack, and the previous stack frame is searched for a catch block that handles the exception. The search of the call stack continues until the exception is handled or until no more frames exist on the call stack. If the top of the call stack is reached without finding a catch block that handles the exception, the default exception handler handles it and the application terminates.

F# try..with Expression

F# does not use catch blocks. Instead, a raised exception is pattern matched using a single with block. As this is an expression, rather than a statement, all paths must return the same type. To learn more, see The try...with Expression.

Exception type features

Exception types support the following features:

• Human-readable text that describes the error. When an exception occurs, the runtime makes a text message available to inform the user of the nature of the error and to suggest action to resolve the problem. This text message is held in the Message property of the exception object. During the creation of the exception object, you can pass a text string to the constructor to describe the details of that particular exception. If no error message argument is supplied to the constructor, the default error message is used. For more information, see the Message property.

• The state of the call stack when the exception was thrown. The StackTrace property carries a stack trace that can be used to determine where the error occurs in the code. The stack trace lists all the called methods and the line numbers in the source file where the calls are made.

Exception class properties

The Exception class includes a number of properties that help identify the code location, the type, the help file, and the reason for the exception: StackTrace, InnerException, Message, HelpLink, HResult, Source, TargetSite, and Data.

When a causal relationship exists between two or more exceptions, the InnerException property maintains this information. The outer exception is thrown in response to this inner exception. The code that handles the outer exception can use the information from the earlier inner exception to handle the error more appropriately. Supplementary information about the exception can be stored as a collection of key/value pairs in the Data property.

The error message string that is passed to the constructor during the creation of the exception object should be localized and can be supplied from a resource file by using the ResourceManager class. For more information about localized resources, see the Creating Satellite Assemblies and Packaging and Deploying Resources topics.

To provide the user with extensive information about why the exception occurred, the HelpLink property can hold a URL (or URN) to a help file.

The Exception class uses the HRESULT COR_E_EXCEPTION, which has the value 0x80131500.

For a list of initial property values for an instance of the Exception class, see the Exception constructors.

Performance considerations

Throwing or handling an exception consumes a significant amount of system resources and execution time. Throw exceptions only to handle truly extraordinary conditions, not to handle predictable events or flow control. For example, in some cases, such as when you're developing a class library, it's reasonable to throw an exception if a method argument is invalid, because you expect your method to be called with valid parameters. An invalid method argument, if it is not the result of a usage error, means that something extraordinary has occurred. Conversely, do not throw an exception if user input is invalid, because you can expect users to occasionally enter invalid data. Instead, provide a retry mechanism so users can enter valid input. Nor should you use exceptions to handle usage errors. Instead, use assertions to identify and correct usage errors.

In addition, do not throw an exception when a return code is sufficient; do not convert a return code to an exception; and do not routinely catch an exception, ignore it, and then continue processing.

Re-throwing an exception

In many cases, an exception handler simply wants to pass the exception on to the caller. This most often occurs in:

• A class library that in turn wraps calls to methods in the .NET Framework class library or other class libraries.

• An application or library that encounters a fatal exception. The exception handler can log the exception and then re-throw the exception.

The recommended way to re-throw an exception is to simply use the throw statement in C#, the reraise function in F#, and the Throw statement in Visual Basic without including an expression. This ensures that all call stack information is preserved when the exception is propagated to the caller. The following example illustrates this. A string extension method, FindOccurrences, wraps one or more calls to String.IndexOf(String, Int32) without validating its arguments beforehand.