Exceptions (Delphi)

From RAD Studio
Jump to: navigation, search

Go Up to Classes and Objects Index

This topic covers the following material:

  • A conceptual overview of exceptions and exception handling
  • Declaring exception types
  • Raising and handling exceptions

About Exceptions

An exception is raised when an error or other event interrupts normal execution of a program. The exception transfers control to an exception handler, which allows you to separate normal program logic from error-handling. Because exceptions are objects, they can be grouped into hierarchies using inheritance, and new exceptions can be introduced without affecting existing code. An exception can carry information, such as an error message, from the point where it is raised to the point where it is handled.

When an application uses the SysUtils unit, most runtime errors are automatically converted into exceptions. Many errors that would otherwise terminate an application - such as insufficient memory, division by zero, and general protection faults - can be caught and handled.

When To Use Exceptions

Exceptions provide an elegant way to trap runtime errors without halting the program and without awkward conditional statements. The requirements imposed by exception handling semantics impose a code/data size and runtime performance penalty. While it is possible to raise exceptions for almost any reason, and to protect almost any block of code by wrapping it in a try...except or try...finally statement, in practice these tools are best reserved for special situations.

Exception handling is appropriate for errors whose chances of occurring are low or difficult to assess, but whose consequences are likely to be catastrophic (such as crashing the application); for error conditions that are complicated or difficult to test for in if...then statements; and when you need to respond to exceptions raised by the operating system or by routines whose source code you don't control. Exceptions are commonly used for hardware, memory, I/O, and operating-system errors.

Conditional statements are often the best way to test for errors. For example, suppose you want to make sure that a file exists before trying to open it. You could do it this way:

try
    AssignFile(F, FileName);
    Reset(F);     // raises an EInOutError exception if file is not found
except
    on Exception do ...
end;

But you could also avoid the overhead of exception handling by using:

if FileExists(FileName) then    // returns False if file is not found; raises no exception

begin
    AssignFile(F, FileName);
    Reset(F);
end;

Assertions provide another way of testing a Boolean condition anywhere in your source code. When an Assert statement fails, the program either halts with a runtime error or (if it uses the SysUtils unit) raises an SysUtils.EAssertionFailed exception. Assertions should be used only to test for conditions that you do not expect to occur.

Declaring Exception Types

Exception types are declared just like other classes. In fact, it is possible to use an instance of any class as an exception, but it is recommended that exceptions be derived from the SysUtils.Exception class defined in SysUtils.

You can group exceptions into families using inheritance. For example, the following declarations in SysUtils define a family of exception types for math errors:

type
   EMathError = class(Exception);
   EInvalidOp = class(EMathError);
   EZeroDivide = class(EMathError);
   EOverflow = class(EMathError);
   EUnderflow = class(EMathError);

Given these declarations, you can define a single SysUtils.EMathError exception handler that also handles SysUtils.EInvalidOp, SysUtils.EZeroDivide, SysUtils.Overflow, and SysUtils.EUnderflow.

Exception classes sometimes define fields, methods, or properties that convey additional information about the error. For example:

type EInOutError = class(Exception)
       ErrorCode: Integer;
     end;

Raising and Handling Exceptions

To raise an exception object, use an instance of the exception class with a raise statement. For example:

raise EMathError.Create;

In general, the form of a raise statement is

raise object at address

where object and at address are both optional. When an address is specified, it can be any expression that evaluates to a pointer type, but is usually a pointer to a procedure or function. For example:

raise Exception.Create('Missing parameter') at @MyFunction;

Use this option to raise the exception from an earlier point in the stack than the one where the error actually occurred.

When an exception is raised - that is, referenced in a raise statement - it is governed by special exception-handling logic. A raise statement never returns control in the normal way. Instead, it transfers control to the innermost exception handler that can handle exceptions of the given class. (The innermost handler is the one whose try...except block was most recently entered but has not yet exited.)

For example, the function below converts a string to an integer, raising an SysUtils.ERangeError exception if the resulting value is outside a specified range.

function StrToIntRange(const S: string; Min, Max: Longint): Longint;
begin
    Result := StrToInt(S);   // StrToInt is declared in SysUtils
    if (Result < Min) or (Result > Max) then
       raise ERangeError.CreateFmt('%d is not within the valid range of %d..%d', [Result, Min, Max]);
end;

Notice the CreateFmt method called in the raise statement. SysUtils.Exception and its descendants have special constructors that provide alternative ways to create exception messages and context IDs.

A raised exception is destroyed automatically after it is handled. Never attempt to destroy a raised exception manually.

Note: Raising an exception in the initialization section of a unit may not produce the intended result. Normal exception support comes from the SysUtils unit, which must be initialized before such support is available. If an exception occurs during initialization, all initialized units - including SysUtils - are finalized and the exception is re-raised. Then the exception is caught and handled, usually by interrupting the program. Similarly, raising an exception in the finalization section of a unit may not lead to the intended result if SysUtils has already been finalized when the exception has been raised.

Try...except Statements

Exceptions are handled within try...except statements. For example:

try
   X := Y/Z;
   except
     on EZeroDivide do HandleZeroDivide;
end;

This statement attempts to divide Y by Z, but calls a routine named HandleZeroDivide if an SysUtils.EZeroDivide exception is raised.

The syntax of a try...except statement is:

try statements except exceptionBlock end

where statements is a sequence of statements (delimited by semicolons) and exceptionBlock is either:

  • another sequence of statements or
  • a sequence of exception handlers, optionally followed by
else statements

An exception handler has the form:

on identifier: type do statement

where identifier: is optional (if included, identifier can be any valid identifier), type is a type used to represent exceptions, and statement is any statement.

A try...except statement executes the statements in the initial statements list. If no exceptions are raised, the exception block (exceptionBlock) is ignored and control passes to the next part of the program.

If an exception is raised during execution of the initial statements list, either by a raise statement in the statements list or by a procedure or function called from the statements list, an attempt is made to 'handle' the exception:

  • If any of the handlers in the exception block matches the exception, control passes to the first such handler. An exception handler 'matches' an exception just in case the type in the handler is the class of the exception or an ancestor of that class.
  • If no such handler is found, control passes to the statement in the else clause, if there is one.
  • If the exception block is just a sequence of statements without any exception handlers, control passes to the first statement in the list.

If none of the conditions above is satisfied, the search continues in the exception block of the next-most-recently entered try...except statement that has not yet exited. If no appropriate handler, else clause, or statement list is found there, the search propagates to the next-most-recently entered try...except statement, and so forth. If the outermost try...except statement is reached and the exception is still not handled, the program terminates.

When an exception is handled, the stack is traced back to the procedure or function containing the try...except statement where the handling occurs, and control is transferred to the executed exception handler, else clause, or statement list. This process discards all procedure and function calls that occurred after entering the try...except statement where the exception is handled. The exception object is then automatically destroyed through a call to its Destroy destructor and control is passed to the statement following the try...except statement. (If a call to the Exit, Break, or Continue standard procedure causes control to leave the exception handler, the exception object is still automatically destroyed.)

In the example below, the first exception handler handles division-by-zero exceptions, the second one handles overflow exceptions, and the final one handles all other math exceptions. SysUtils.EMathError appears last in the exception block because it is the ancestor of the other two exception classes; if it appeared first, the other two handlers would never be invoked:

try
  ...
except
  on EZeroDivide do HandleZeroDivide;
  on EOverflow do HandleOverflow;
  on EMathError do HandleMathError;
end;

An exception handler can specify an identifier before the name of the exception class. This declares the identifier to represent the exception object during execution of the statement that follows on...do. The scope of the identifier is limited to that statement. For example:

try
  ...
except
  on E: Exception do ErrorDialog(E.Message, E.HelpContext);
end;

If the exception block specifies an else clause, the else clause handles any exceptions that aren't handled by the block's exception handlers. For example:

try
  ...
except
  on EZeroDivide do HandleZeroDivide;
  on EOverflow do HandleOverflow;
  on EMathError do HandleMathError;
else
  HandleAllOthers;
end;

Here, the else clause handles any exception that isn't an SysUtils.EMathError.

An exception block that contains no exception handlers, but instead consists only of a list of statements, handles all exceptions. For example:

try
   ...
except
   HandleException;
end;

Here, the HandleException routine handles any exception that occurs as a result of executing the statements between try and except.

Re-raising Exceptions

When the reserved word raise occurs in an exception block without an object reference following it, it raises whatever exception is handled by the block. This allows an exception handler to respond to an error in a limited way and then re-raise the exception. Re-raising is useful when a procedure or function has to clean up after an exception occurs but cannot fully handle the exception.

For example, the GetFileList function allocates a TStringList object and fills it with file names matching a specified search path:

function GetFileList(const Path: string): TStringList;
var
  I: Integer;
  SearchRec: TSearchRec;
begin
  Result := TStringList.Create;
  try
    I := FindFirst(Path, 0, SearchRec);
    while I = 0 do
      begin
          Result.Add(SearchRec.Name);
          I := FindNext(SearchRec);
      end;
  except
      Result.Free;
      raise;
  end;
end;

GetFileList creates a TStringList object, then uses the FindFirst and FindNext functions (defined in SysUtils) to initialize it. If the initialization fails - for example because the search path is invalid, or because there is not enough memory to fill in the string list - GetFileList needs to dispose of the new string list, since the caller does not yet know of its existence. For this reason, initialization of the string list is performed in a try...except statement. If an exception occurs, the statement's exception block disposes of the string list, then re-raises the exception.

Nested Exceptions

Code executed in an exception handler can itself raise and handle exceptions. As long as these exceptions are also handled within the exception handler, they do not affect the original exception. However, once an exception raised in an exception handler propagates beyond that handler, the original exception is lost. This is illustrated by the Tan function below:

type
   ETrigError = class(EMathError);
   function Tan(X: Extended): Extended;
   begin
      try
        Result := Sin(X) / Cos(X);
      except
        on EMathError do
        raise ETrigError.Create('Invalid argument to Tan');
      end;
   end;

If an SysUtils.EMathError exception occurs during execution of Tan, the exception handler raises an ETrigError. Since Tan does not provide a handler for ETrigError, the exception propagates beyond the original exception handler, causing the SysUtils.EMathError exception to be destroyed. To the caller, it appears as if the Tan function has raised an ETrigError exception.

Try...finally Statements

Sometimes you want to ensure that specific parts of an operation are completed, whether or not the operation is interrupted by an exception. For example, when a routine acquires control of a resource, it is often important that the resource be released, regardless of whether the routine terminates normally. In these situations, you can use a try...finally statement.

The following example shows how code that opens and processes a file can ensure that the file is ultimately closed, even if an error occurs during execution:

Reset(F);
try
   ... // process file F
finally
   CloseFile(F);
end;

The syntax of a try...finally statement is

try statementList1 finally statementList2 end

where each statementList is a sequence of statements delimited by semicolons. The try...finally statement executes the statements in statementList1 (the try clause). If statementList1 finishes without raising exceptions, statementList2 (the finally clause) is executed. If an exception is raised during execution of statementList1, control is transferred to statementList2; once statementList2 finishes executing, the exception is re-raised. If a call to the Exit, Break, or Continue procedure causes control to leave statementList1, statementList2 is automatically executed. Thus the finally clause is always executed, regardless of how the try clause terminates.

If an exception is raised but not handled in the finally clause, that exception is propagated out of the try...finally statement, and any exception already raised in the try clause is lost. The finally clause should therefore handle all locally raised exceptions, so as not to disturb propagation of other exceptions.

Standard Exception Classes and Routines

The SysUtils and System units declare several standard routines for handling exceptions, including ExceptObject, ExceptAddr, and ShowException. SysUtils, System and other units also include dozens of exception classes, all of which (aside from OutlineError) derive from SysUtils.Exception.

The SysUtils.Exception class has properties called Message and HelpContext that can be used to pass an error description and a context ID for context-sensitive online documentation. It also defines various constructor methods that allow you to specify the description and context ID in different ways.

See Also