REGISTER  |  SIGN IN
United States Patent5475843
Halviatti , ; et al.December 12, 1995
Title
System and methods for improved program testing
Abstract
The present invention includes a Computer-based Training system (CBT) having one or more Application Translation Units (ATUs), a Message Engine, and a Script Engine. For one or more target applications of interest, an ATU is provided for processing events specific to that application, thereby trapping events and translating them into abstract messages or "meta-messages" for conveying information about a particular event to the system. A computer-aided software testing embodiment of the present invention is also described. The system provides prefabricated building blocks for constructing a high-level model of an application's User Interface (UI). This high-level model serves as a middle ground between test scripts and the application being tested. The knowledge of how a given UI element is controlled or how it can be observed is retained in the model rather than in a test script. Consequently, the test script consists of easy-to-maintain, high-level testing commands only.
Inventors:Halviatti; Ramin L. (Santa Cruz, CA), Potts; Richard J.  (Palo Alto, CA)
Assignee:Borland International, Inc. (Scotts Valley, CA)
Appl. No.:140904
Filed:October 21, 1993
Current U.S. Class:717/124 717/125 717/127 719/329 717/108 
Field of Search:395/155,161,700,650
U.S. Patent Documents
4622013November 1986Cerchio
4789962December 1988Berry et al.
4901221February 1990Kodosky et al.
4947346August 1990Kamiya
4964077October 1990Eisen et al.
5103498April 1992Lanier et al.
5175812December 1992Krieger
5204968April 1993Parthasarathi
5239617August 1993Gardner et al.
Other References
Kepple, L., Testing GUI Applications: The Logic of Automation, Proceedings: 10th International Conference and Exposition on Testing Computer Software: Risk Driven Testing, Jun. 14-17, 1991, Washington, D.C., pp. 37-46. .
Wolters, W. High Automated Testing for GUI Displays, Proceedings: 10th International Conference and Exposition on Testing Computer Software: Risk Driven Testing, Jun. 14-17, 1991, Washington, D.C., pp. 111-116. .
Needleman, Raphael, Wizards' Make Works, Publisher Easy, Infoworld, Sep. 16, 1991, p. 78. .
Matthies, Kurt, W. G., Balloon Help Takes Off, MacUser, Dec. 1991, pp. 241-248..~
Primary Examiner: Heckler; Thomas M.
Attorney, Agent or Firm:Smart; John A.
Parent Case Text


The present application is a continuation-in-part of commonly-assigned application Ser. No. 07/970,724, filed Nov. 2, 1992now U.S. Pat. No. 5,432,940, the disclosure of which is incorporated herein by reference.
Claims

What is claimed is:
1. A computer-aided system for testing operation of an application program having a user interface, the system comprising:
means for storing a library of generic element models, each said generic element model encapsulating functionality and behavior for a single user interface element;
means for creating a test model of said application program by combining selected ones of said generic user interface elements into at least one application-specific testing model, each said at least one application-specific testing model representing a high-level model for a specific user interface component for the application program;
message engine means for monitoring events during operation of the application program; and
script engine means for specifying actions of said test model to be undertaken upon occurrence of selected ones of the events.

2. The system of claim 1, wherein user interface elements modeled by said generic element models include selected ones of a push button, a check box, a list box, and a text object.

3. The system of claim 1, wherein specific user interface components for the application program include selected ones of a menu, a toolbar, a dialog, a client-area window, and a status line.

4. The system of claim 1, wherein said events include system events generated by operating system software.

5. The system of claim 1, wherein said events include application-specific events which result from non-system events occurring within said application program.

6. The system of claim 1, wherein functionality for a user interface element includes how the element responds to occurrence of events during operation of the application program.

7. The system of claim 1, wherein behavior for a user interface element includes screen attributes specifying how the element is displayed on a screen during operation of the application program.

8. The system of claim 7, wherein said screen attributes include selected ones of screen position, size, color, and font.

9. The system of claim 7, wherein said screen attributes include a default state.

10. The system of claim 7, further comprising:
a database for storing expected values of screen attributes for user interface elements of the application program.

11. The system of claim 10, further comprising:
comparison means for comparing expected values stored in the database to actual values for user interface elements of the application program at runtime.

12. The system of claim 11, further comprising:
self-testing means for automatically invoking default functionality of each said at least one application-specific testing model for the application program and comparing actual behavior of each said at least one application-specific testing model with its expected behavior.

13. The system of claim 1, wherein said script engine means operates in response to a test script, said test script including high-level commands specifying a test in terms of actions to occur at selected ones of specific user interface components for the application program.

14. The system of claim 1, wherein said means for creating a test model includes:
means for decomposing the application program into irreducible user interface components.

15. The system of claim 14, wherein said means for decomposing includes:
means for reading resource information from the application program, said resource information specifying creation of user interface elements of the application program at runtime.

16. The system of claim 14, wherein said means for decomposing includes:
means for operating the application program and monitoring creation of user interface elements of the application program at runtime.

17. The system of claim 1, further comprising:
means for dynamically binding each said at least one application-specific testing model with the actual specific user interface component it represents.

18. The system of claim 1, wherein said specified actions include user events for simulating operation of the application program by a user.

19. The system of claim 18, wherein said user events include selected ones of keyboard device and mouse device input events.

20. In a computer system, a method for testing operation of an application program having a user interface, the method comprising:
(a) creating a test model of said application program by combining pre-supplied generic element models into at least one application-specific testing model, each said generic element model encapsulating functionality and behavior for a single user interface element, each said at least one application-specific testing model representing a high-level model for a specific user interface component for the application program;
(b) monitoring events during operation of the application program; and
(c) specifying actions of said test model to be undertaken upon occurrence of selected ones of the events, whereupon a specified action in the application program is effected in response to occurrence of said selected events.

21. The method of claim 20, wherein said generic element models represent selected ones of a push button, a check box, a list box, and a text object.

22. The method of claim 20, wherein specific user interface components for the application program include selected ones of a menu, a toolbar, a dialog, a client-area window, and a status line.

23. The method of claim 20, wherein said events include selected ones of events generated by operating system software and events occurring within said application program.

24. The method of claim 20, wherein behavior for a user interface element includes screen attributes specifying how the element is displayed on a screen during operation of the application program.

25. The method of claim 24, wherein said screen attributes include selected ones of screen position, size, color, and font.

26. The method of claim 24, further comprising:
storing in a database expected values of screen attributes for user interface elements of the application program.

27. The method of claim 26, further comprising:
comparing expected values stored in the database to actual values for user interface elements of the application program at runtime.

28. The method of claim 27, further comprising:
sequentially testing each specific user interface component which is modeled by invoking default functionality of each said at least one application-specific testing model for the application program and comparing actual behavior of each said at least one application-specific testing model with its expected behavior.

29. In a computer system, a method for testing operation of a graphical user interface, the method comprising:
(a) creating for the graphical user interface at least one model for testing user interface elements, each said at least one model being constructed from a set of generic objects representing basic user interface elements;
(b) providing each model a link to its actual user interface element;
(c) storing for each model a set of expected characteristics that its actual user interface element is to exhibit;
(d) trapping a request to create a user interface element; and
(e) in response to a trapped request, loading a model for the user interface element which is requested to be created and comparing its actual characteristics to said stored expected ones.

30. The method of claim 29, further comprising:
(f) providing each model with a set of methods for simulating operation of its user interface element; and
(g) simulating runtime operation of the graphical user interface by invoking methods for selected ones of the user interface elements.

31. The method of 30, wherein step (g) includes:
providing a test script, said test script having high-level commands for simulating operation of the graphical user interface.

32. The method of 31, wherein said high-level commands include a general format of:
where [object] specifies a user interface element and [action] specifies a simulated operation for the object.

33. The method of claim 29, wherein specific user interface elements include selected ones of a menu, a toolbar, a dialog, a client-area window, and a status line.

34. The method of claim 29, wherein step (d) includes:
registering an interest in a system message which indicates that a user interface element is about to be created; and
monitoring for occurrence of said system message.

35. The method of claim 34, wherein said graphical user interface includes Microsoft.RTM. Windows graphical user interface, and said system message is a Microsoft.RTM. Windows CreateWindow message.
Description
COPYRIGHT NOTICE

A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

BACKGROUND OF THE INVENTION

The present invention relates generally to system and methods for testing reliability of software programs. More particularly, the present invention relates to a computer-aided software testing system and methods which assist a Quality-Assurance engineer and software developers with the testing of Graphical User Interface (GUI) software programs operative on digital computers.

Development of software is largely a trial and error process. Accordingly, substantial development resources are allocated to the process of finding "bugs"--errors occurring in the program being developed. Expectedly, there is keen interest in finding ways to improving the testing of software.

As software is developed on and runs on computers, it is not surprising to find that many of the techniques for automating the testing of software have been implemented in digital computers. A common approach for testing software is the use of test suites. Test suites compare "known good" outputs of a program (for a given set of input) against the current output. Tests that check program file output are easy to implement and can be automated with shell scripts (e.g., Expect available on the Internet). For programs with user interfaces that communicate to standard input/output devices (stdin/stdout), a similar method may be employed. Capture/playback tools are available for recording keyboard input and program output as a person tests a program.

Much of the code written today is for software products with a graphical user interface (GUI), such as Microsoft.RTM. Windows.TM.. In fact, much of software development itself is done within a graphical user interface, with software tool vendors providing products which allow software developers to develop GUI software using visual programming techniques. The Quality Assurance (QA) engineer faces more complex problems when testing GUI software. In particular, GUI programs must behave correctly regardless of which video mode or operating environment is being employed.

Intuitively, testing user interfaces should not be as difficult as testing a complex internal engine, such as a compiler or a real-time, multi-user operating system. In practice, however, user interface (UI) testing is the most challenging part of the QA process. This problem stems largely from the difficulty in automating UI tests. Tests for complex engines, in contrast, are often command-line programs whose testing can easily be automated using simple batch execution. Thus despite the plethora of present day tools for automating program testing, the task of developing, maintaining and analyzing the results of UI tests remains an arduous task.

The basic steps traditionally employed to test user interfaces may be summarized as follows. First, the application being tested is controlled by placing it into a specific state using either pre-recorded keyboard or mouse device actions, or entering input through a test script. Next, the then-current state of the application is recorded by taking a screenshot (e.g., capturing a screen bitmap). Finally, the captured screenshot is compared with a baseline screenshot that is known to be valid.

The approach is far from ideal, however. Consider, for instance, the determination of whether the state of a check box is valid within a specific dialog box. Here, the QA engineer must take a screenshot of that check box and compare it with the expected image. Thus, testing of even the simplest component is laborious. Moreover, the approach itself is prone to error. A change of just a few pixels across all windows--a common occurrence in GUI software development--causes all tests to fail. Consequently, as software becomes more and more complex, it becomes less and less feasible to test user interface tasks with present-day screen comparison methodology. A better approach is needed.

SUMMARY OF THE INVENTION

The present invention includes, in a first embodiment, a Computer-based Training system (CBT) having one or more Application Translation Units (ATUs), a Message Engine, and a Script Engine. Specific operation of the system is directed by a series of user instructions, typically provided by a tutorial writer. Within the script, links are established between the CBT system and one or more target applications of interest. Specifically, within a particular application links are established with individual controls (e.g., menu items, button controls, dialog boxes, and the like) so that the script writer has complete control over the behavior and actions of the target application.

For each target application of interest, an ATU is provided for processing events specific to that application. A general Operating System (OS) ATU is also provided for trapping general system events. The ATUs function to trap events and translate them into abstract messages or "meta-messages" for conveying information about a particular event to the system. In this manner, low-level messages are abstracted into high-level, more meaningful messages which may be acted upon through the script.

Translated event messages are forwarded to the Message Engine for matching with event handlers. In a preferred embodiment, the Message Engine maintains a lookup table for matching messages with a desired target handler. System or application-specific messages which are not of interest are simply allowed to pass through. From the Message Engine, the individual handlers dispatch their respective messages to the Script Engine. The Script Engine, in turn, matches an incoming message with reserved words of the script. Appropriate action, based upon use of the reserved word within the script, is then effected.

A software testing automation embodiment of the present invention is also described. The system includes a Generic Element Models (GEMs) library, Application-specific Testing Models (ATMs), a Resource Database, one or more Model Generators, a Test Runtime Library, as well as the above-mentioned Script Engine, Message Engine, and Application Translation Units (ATUs).

The system employs the Model Generator to decompose the application under test to generate the ATMs. Each ATM is a high-level model for a specific component of the application being tested, such as a File Open dialog. ATMs describe the actual component which they represent in terms of Generic Element Models (stored in GEMs Library). A GEM encapsulates the behavior of irreducible user interface elements such as push buttons, checkboxes, listboxes, and the like. Knowledge of how a given UI element is controlled or how it can be observed is retained in the model rather than in a test script. This high-level model serves as a middle ground between test scripts and the application being tested. In this fashion, a script for testing operation of a program need only consist of easy-to-maintain, high-level testing commands.

During operation, the system maintains an in-memory Testing Model of a particular application under test. Overall operation of the system is directed by one or more test scripts. GEMs are instantiated when an ATM corresponding to an active state of the application under test is activated by the Model Manager. GEMs load their expected results from the Resource Database and are capable of binding themselves dynamically to the actual object on the screen which they represent. Each GEM therefore can perform a "self test" on itself by simply comparing its expected result (as stored in the Resource Database) to its actual behavior (as displayed on the screen at runtime). In this manner, an entire application can perform a self test by simply asking all its components to test themselves in turn.

GEMs of the present invention provide maximum controllability and observability over the actual screen objects that they represent. By breaking the user interface down into irreducible components which are modeled to provide maximum controllability and observability (over the actual screen objects that they represent), the Test Model created provides maximum controllability and observability. Accordingly, testing effectiveness is maximized.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1A is a block diagram of a computer system in which the present invention may be embodied.

FIG. 1B is a block diagram of a software system of the present invention, which includes operating system, application software, and user interface components.

FIG. 1C is a bitmap screenshot illustrating the basic architecture and functionality of a graphical user interface in which the present invention may be embodied.

FIG. 2 is a pair of flowcharts contrasting the operation of conventional modal architecture with that of event-driven architecture.

FIG. 3 is a block diagram of a Computer-based Training (CBT) system of the present invention.

FIG. 4A is a flowchart illustrating a method of the present invention for operating the CBT system of FIG. 3.

FIG. 4B is a flowchart illustrating the operation of hooks, dispatchers, and handlers of the method of FIG. 4A.

FIG. 5A is a flowchart illustrating the operation of an exemplary event handler of the present invention, which includes the dispatching of event information (EventInfo) objects.

FIG. 5B is a flowchart illustrating a method of the present invention for dispatching meta-messages.

FIG. 5C is a class hierarchy diagram illustrating the underlying structure for the EventInfo objects of FIG. 5A.

FIG. 6 is a block diagram of a computer-aided software testing system of the present invention.

FIG. 7A is a bitmap screenshot illustrating a screen dialog object (box), which includes children components (e.g., screen buttons).

FIG. 7B illustrates the relationship between the dialog of FIG. 7A and underlying resource information (which originally had been provided by the programmer creating the application).

FIG. 7C is a diagram illustrating the relationship between resource information (represented in Microsoft Windows resource format) and the runtime API Windows call (CreateWindow).

FIG. 8A is a bitmap screenshot illustrating Notepad--a simple Windows application for illustrating operation of the system of FIG. 6.

FIG. 8B is a diagram illustrating resource information for a screen menu of the application of FIG. 8A.

FIG. 8C is a flowchart illustrating a method of the present invention for decomposing (extracting) menu information for the Notepad application.

FIGS. 8D-E are diagrams illustrating the relationship between resource information (however extracted) and Application-specific Testing Models (ATMs) constructed by the system of the present invention.

FIG. 9A is a bitmap screenshot illustrating a File Open dialog of the Windows Notepad application.

FIG. 9B is a bitmap screenshot illustrating children components (e.g., screen buttons, combo boxes, and list boxes) of the dialog of FIG. 9A.

FIGS. 9C-D comprise a flowchart illustrating a method of the present invention for decomposing dialog information for the Notepad application.

GLOSSARY OF TERMS

CBT: Computer-based Training; the use of computers and specially designed tutorial programs for teaching.

CBT Message: A high-level or meta-message describing or encapsulating information about a particular event which has occurred, thereby allowing the user to abstract low level system messages into high level (and more meaningful) messages for script control.

CBT Object: An object, such as a C++ object, which can be placed in a Dynamic Link Library (DLL) and dynamically loaded when the tutorial is executed.

Control Window: A CBT object which is used for getting information from the user; typically, it includes a window having a number of dialog controls.

Interaction Objects: CBT objects which are used to interact with the user. These objects include Presentation windows, Control windows and message handlers.

Lesson Script: Script statements which control the execution of the CBT tutorial. Each lesson includes a collection of Scenes.

List: A container object which is used to hold unorganized data.

Message Handler: A CBT object which interacts with the target application. These objects are used to handle external events.

Object Property: An attribute or other data associated with a particular object, for example, name, title, color, and the like.

Performance: The execution of a CBT Scene by the CBT system.

Scene: A script object which describes the actions to perform at a particular point in the CBT lesson.

Script: A collection of statements which are understood by a Script Engine.

Windows Hook: A function installed between the Windows OS and an application to intercept Windows events before they are received by the application.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT

The following description will focus on the presently preferred embodiment of the present invention, which is operative in the Microsoft.RTM. Windows environment. The present invention, however, is not limited to any particular one application or any particular windows environment. Instead, those skilled in the art will find that the system and methods of the present invention may be advantageously applied to a variety of system and application software, including database management systems, wordprocessors, spreadsheets, and the like. Moreover, the present invention may be embodied on a variety of different platforms, including Macintosh, UNIX, NextStep, and the like. Therefore, the description of the exemplary embodiments which follows is for purposes of illustration and not limitation.

System Hardware

As shown in FIG. 1A, the present invention may be embodied in a computer system such as the system 100, which comprises a central processor 101, a main memory 102, an input/output controller 103, a keyboard 104, a pointing device 105 (e.g., mouse, track ball, pen device, or the like), a display device 106, and a mass storage 107 (e.g., hard disk). Additional input/output devices, such as a printing device 108, may be included in the system 100 as desired. As illustrated, the various components of the system 100 communicate through a system bus 110 or similar architecture. In a preferred embodiment, the computer system 100 includes an IBM-compatible personal computer, which is available from several vendors (including IBM of Armonk, N.Y.).

System Software

A. Overview

Illustrated in FIG. 1B, a computer software system 150 is provided for directing the operation of the computer system 100. Software system 150, which is stored in system memory 102 and on disk memory 107, includes a kernel or operating system (OS) 160 and a windows shell 180. One or more application programs, such as application software 170 or windows application software 190, may be "loaded" (i.e., transferred from storage 107 into memory 102) for execution by the system 100.

System 150 includes a user interface (UI) 165, preferably a graphical user interface (GUI), for receiving user commands and data. These inputs, in turn, may be acted upon by the system 100 in accordance with instructions from operating module
160, Windows 180, and/or application modules 170, 190. The UI 165 also serves to display the results of an operation, whereupon the user may supply additional inputs or terminate the session. Although shown conceptually as a separate module, the UI is typically provided by Windows shell 180, operating under OS 160. In a preferred embodiment, OS 160 is MS-DOS and Windows 180 is Microsoft.RTM. Windows; both are available from Microsoft Corporation of Redmond, Wash.

System 150 also includes a Computer-based Training (CBT) system 200 of the present invention for aiding users of the computer 100. As shown, the CBT system 200 interfaces with the system 100 through Windows shell 180, as well as interfacing directly through OS 160. Before undertaking a detailed description of the construction and operation of the CBT system 200 itself, however, it is helpful to first examine the general methodology of UI 165 and the event-driven architecture driving its operation.

B. Graphical User (Windowing) Interface

As shown in FIG. 1C, the system 100 typically presents UI 160 as a windowing interface or workspace 161. Window 161 is a rectangular, graphical user interface (GUI) for display on screen 106; additional windowing elements may be displayed in various sizes and formats (e.g., tiled or cascaded), as desired. At the top of window 161 is a menu bar 170 with a plurality of user-command choices, each of which may invoke additional submenus and software tools for use with application objects. Window 161 includes a client area 180 for displaying and manipulating screen objects, such as graphic object 181 and text object 182. In essence, the client area is a workspace or viewport for the user to interact with data objects which reside within the computer system 100.

Windowing interface 161 includes a screen cursor or pointer 185 for selecting and otherwise invoking screen objects of interest. In response to user movement signals from the pointing device 105, the cursor 185 floats (i.e., freely moves) across the screen 106 to a desired screen location. During or after cursor movement, the user may generate user-event signals (e.g., mouse button "clicks" and "drags") for selecting and manipulating objects, as is known in the art. For example, Window 161 may be closed, resized, or scrolled by "clicking on" (selecting) screen components 172, 174/5, and 177/8, respectively. Keystroke equivalents, including keyboard accelerators or "hot keys", are provided for performing these and other user operations through keyboard 104.

C. Event-driven Architecture

Underlying the windowing interface is a message or event-driven architecture. This model is perhaps best described by contrasting its operation with that of a modal or sequential architecture which has been traditionally employed. In this manner, the reader may appreciate the added flexibility of a message-driven system--flexibility which is employed by the CBT system of the present invention for providing hi-directional communication not only between the CBT system and a user but between the CBT system and a target application as well.

As shown in FIG. 2, a modal program 200 comprises a series of discrete operating blocks or modes 210, with a well-defined beginning, middle and end. In actual operation, such a program typically displays a series of input screens for receiving user information, for example, to create a written report. For instance, the first entry screen may require a customer name, the second a customer address, the third a part number, and so on. The program typically terminates in an output mode, for reporting results determined from the various inputs. Thus, the program 200 follows a fairly rigid sequence of operation, with each input or entry mode demanding successful completion before the program proceeds to the next step.

While a modal program is relatively easy to design and implement, it is not so easy to use. The design certainly ensures that all required information is entered, but only at the expense of forcing users to operation in a manner dictated by the program. Specifically, since the program is built around a pre-arranged set of modes, a user cannot get from one mode to another without first completing a previously-required mode. In the present example, for instance, a user must needlessly complete a customer name entry screen (and any other intervening input screens) just to access part number information. Any deviation from this sequence by the user is simply not permitted. Lacking flexibility, modal programs make a poor choice for handling real-world tasks.

As shown in the second half of FIG. 2, an event-driven architecture 250 eschews a pre-selected sequence, opting instead for an "event loop." The event loop 260 is a centralized mechanism for processing messages about user and system events. It includes an event queue 270 and mechanisms for retrieving 263 and dispatching 269 messages to various window classes 280. Before each of these components is described in detail, it is helpful to have a firm understanding of "messages."

In a typical modal environment, especially those typified by a character-based UI, a program reads from the keyboard by making an explicit call to a function, such as the C function getchar(). The function typically waits until the user presses a key before returning the character code to the program; all system activity ceases until completion of this one step. In a Windows environment, in contrast, the operating system uses messages to manage and synchronize multiple applications and hardware events, such as clicks of a mouse or presses of a keyboard, which are converted to messages by Windows event handlers.

From a programming perspective, a message is simply a data structure containing information about a particular event. In Microsoft Windows, a message is a 16-bit unsigned integer which serves as a symbolic constant for a particular event; packaged within this integer is a message identifier and message parameters, which vary with each event type represented. For example, messages from a window object might include information about creating (WM.sub.-- CREATE), closing (WM.sub.-- CLOSE), moving (WM.sub.-- MOVE), and resizing (WM.sub.-- SIZE) the window. The input messages are collected in a system-wide queue and then directed to the proper window. These messages, along with timer and screen paint messages, must be passed to the target application(s) of interest.

A mechanism is provided for retrieving messages from the system queue and dispatching them to the appropriate application which, in turn, may proceed to process any message that arrives. Each window belongs to a certain window type which defines certain characteristics common to all windows of that type. Associated with each type is a Windows function which processes all messages sent to windows of its type. An application queue is provided where windows may place messages that belong to a specific application. When the application is ready to receive input, it simply reads the awaiting messages. If none are found or if there exists a message for other applications with higher priority, Windows passes control to the other applications.

The message or event loop 260 shown in FIG. 2, for example, may be simply coded as:

______________________________________ while (GetMessage (&msg, NULL, 0, 0)) { TranslateMessage (&msg) ; DispatchMessage (&msg) ; } return msg.wParam ; ______________________________________

where a message (&msg) is retrieved by a call to GetMessage (step 263); if needed, the retrieved message may be translated by a call to TranslateMessage() and then dispatched by a call to DispatchMessage (step 269). This "while" loop continues until the GetMessage function returns a value of zero--indicating that the loop has read a WM.sub.-- QUIT message from the queue, telling the application to end (yes at step 265).

In addition to the messages generated by Windows, applications can create their own messages and place them in the application queues of other applications. The SendMessage and PostMessage functions let applications pass messages to their windows or to the windows of other applications. The PostMessage function directs Windows to post the message by placing it in the application queue. Control returns immediately to the calling application, and any action to be carried out as a result of the message does not occur until the message is read from the queue.

The SendMessage function, on the other hand, directs Windows to send a message directly to the given Windows function, thereby bypassing the application queue. Windows does not return control to the calling application until the Windows function that receives the message has processed it. Such an unqueued message, however, is generally a message that affects the window only.

The general mechanism for retrieving and dispatching messages in an event-based system, such as Microsoft Windows, is known in the art; see, e.g., Petzold, C., Programming Windows, Second Edition, Microsoft Press, 1990. Additional information can be found in Microsoft's Window Software Development Kit, including: 1) Guide to Programming, 2) Reference, Vols. 1 and 2, and 3) Tools, all available from Microsoft Corp. of Redmond, Wash. The disclosures of each of the foregoing are hereby incorporated by reference.

First embodiment: Computer-based Training System

The following description of the CBT system of the present invention will focus on the presently preferred embodiment which includes components implemented in an event-driven architecture with the C++ programming language. In particular, an object-oriented model is adopted whereby new objects may be created from existing objects (classes). The general features of C++, including data encapsulation, inheritance, and polymorphism, as well as C++ techniques for implementing class hierarchies and class methods are known; see e.g., Ellis, M. and Stroustrup, B., The Annotated C++ Reference Manual, Addison-Wesley, 1990. Additional information about object-oriented programming and C++ in particular can be found in Borland.RTM. C++3.1: 1) User's Guide, 2) Programmer's Guide, and 3) Library Reference, all available from Borland International of Scotts Valley, Calif. The disclosures of each of the foregoing are hereby incorporated by reference.

A. Overview

Referring now to FIG. 3, a Computer-based Training system 300 of the present invention includes one or more Application Translation Units (ATUs) 340, a Message Engine 350, and a Script Engine 330. Also shown, the system 300 includes a CBT window object 370 operably coupled to the Script Engine; CBT window object 370, in turn, may be linked to one or more custom Dynamic Link Libraries (DLLs) 380.

Driven by a set of instructions or script 320, the system 300 monitors various events of the Windows system and desired target applications. Messages from these events, including system messages 310 and target application messages, are trapped by the ATUs 340 and reported to the Message Engine 350 as CBT messages. The Message Engine, in turn, dispatches the messages according to handlers specified by the Script Engine 330, operating under the control of script 320. Messages of interest are processed as desired; others are simply allowed to pass through. The construction and operation of these components will now be described in further detail.

1. Application Translation Units and their Target Applications

In normal operation of the system 100, a user is using one or more Windows application programs, for example, programs 145 of FIG. 1B, which can be a spreadsheet, wordprocessor, database, or the like. For each application where CBT support is desired, an application-specific ATU 340 is provided for processing events specific to that application (now shown as target application 360 in FIG. 3). Thus, each ATU 340 is a module, preferably implemented as a Dynamic Link Library (DLL), for processing messages for a specific target application.

In addition to the application-specific DLLs, ATUs 340 include a Windows ATU. In contrast to the other ATUs which hook into specific target applications, the Windows ATU processes all Windows events, including system messages 310. In this manner, general system events, that is, ones not specific to any particular application, may be processed by the CBT system 300 as well.

Whether application specific or not, each ATU functions to trap events and convert them into "CBT messages"--a lingua franca or common language for all events, whether Windows or application-specific, occurring within the system. More particularly, a CBT message is a meta-message, that is, a high-level message describing one or more events which have occurred. For instance, instead of monitoring numerous, low-level system messages, such as WM.sub.-- RBUTTONDOWN, WM.sub.-- LBUTTONDOWN, WM.sub.-- RBUTTONUP, WM.sub.-- LBUTTONUP, and the like, the user/script writer need only monitor a single message CBT.sub.-- MOUSECLICK; the task of determining what the Windows system or application is doing is left to the ATUs. By abstracting low-level system messages into high-level (and more meaningful) CBT messages, the system of the present invention greatly simplifies script design and writing.

A CBT message, which is stored internally as an integer, represents a particular type of event. Information or data particular to each event, such as active window, cursor location, and the like, on the other hand, are packaged as a discreet data object (EventInfo object) fully describing the event. EventInfo objects, along with CBT messages, are dispatched from the ATU to the Message Engine 350 for further processing.

2. Message Engine and Scripts

The Message Engine 350 determines which of the many CBT messages it receives is of particular interest to the CBT system, as it operates under the direction of a script 320. At startup, each target application of interest to the script 320 is registered with the Message Engine. The script 320, which itself includes a collection of objects, may be compiled, pre-compiled, pseudocompiled, or the like, as desired for a particular platform. In a preferred embodiment, script 320 is pre-tokenized, whereby source listings and comments (optionally) are stored in a compact form which may be reconstructed into the original source. Thus, script 320 need only be a set of instructions; no particular format is required by the present invention.

For each application registered, the Message Engine maintains one or more "handlers" or modules for processing events specific to that application (as represented by the CBT messages). Thus, the Message Engine manages a list of target handlers for determining which CBT messages are of interest to the script 320, that is, which messages map into the list.

Messages which are of interest to the system, i.e., those specified by the script 320, are trapped and forwarded to a designated handler. Those messages which are not of interest are allowed to "fall through" (i.e., be passed on to) the individual target applications 360. In essence, the Message Engine filters the events for a particular application so that only those of interest, that is, those having a handler defined for the event, are processed.

B. System operation

1. Method: CBT session

Referring now to FIGS. 4A-B, the general method of operation for the system 300 is illustrated by a flowchart 400. While the general methodology remains the same from one CBT session to another, the reader should understand that the specific steps of any one CBT session are dictated by instructions and script objects defined in the script 320. For instance, the script 320 states which specific target applications will be registered with the Message Engine and which events of those applications will be dispatched. Thus, the following description illustrates the basic framework in which the CBT system operates.

Under the direction of the script 320, at step 410 the CBT system registers a target application of interest by creating a CBTApplication object. Serving as the main entry point for the CBT system, this object initializes and executes the script tutorial. Moreover, the object initiates sessions with the Message Engine and Script Engine and acts as a centralized dispatch point for all external functions and object method calls within each CBT lesson. From here individual CBT lessons are loaded and performed.

When a lesson script is loaded the CBTApplication object creates a CBTLesson object which is responsible for managing that particular lesson. The CBTLesson object reads the lesson script and builds a deck of CBTScene objects, with each corresponding to a scene described in the lesson script. Alternatively, each scene may be constructed on-the-fly, particularly in high-performance environments. The CBTLesson object resembles a state machine; it maintains the active scene (active state) and sends appropriate script scenes (CBTScene objects) to the Script Engine 330 for presentation. Each object is directly accessible to the script writer through script variables; for example, the CBTApplication object is accessed through a theCBTApp global variable, the CBTLesson object through a theCBTLesson global variable.

To complete the initialization of step 410, the target application is registered with the Message Engine 350. In particular, hooks are installed by a corresponding ATU 340 so that events within the target application of interest are trapped. As set forth below, these events will, in turn, be reported to the Message Engine 350 as CBT messages.

At step 420, script-specified links are established to individual resources of interest. Within the target application itself, various resources (e.g., buttons, menus, and the like) may be of interest to the script writer. For example, if one were interested in a particular button of the target application, such as a "help" button, that button may be registered with the Message Engine 350. Events associated with that button (e.g., "mouse enter" the button) are then trapped for processing by an ATU.

The links are specified within the CBT script 320 as follows. In a preferred embodiment, a link is established to an individual resource or control by indicating a selected one of its Windows class name, Window title, or Resource ID, all of which are readily accessible Windows data structures. Commonly, a link will be established by using the Resource ID. Particular links may be created or removed during a session, as desired.

At step 430, the system traps various events which are relevant to the established links. This operation, shown in further detail in FIG. 4B, occurs as follows. Every time that a message arrives at the message queue of a target application, a message filter hook function (WH.sub.-- MsgFilter) 431 is called. First, the hook function determines if the message represents a task which has been "hooked," that is a specific link for this event of the target application has been created. If not, the event is passed down the hook chain for processing by one of the other hooks (i.e., hooks 433, 435, 437). The Windows hook 437, for example, traps all window messages (WM.sub.--). In this fashion, which hook function is invoked depends on what type of message comes into the target application. By way of illustration, hooks may be installed as follows:

__________________________________________________________________________ BOOL installHooks(CbtEntry *pEntry) if( ! pEntry ) return FALSE; // Note that the fourth parameter below may receive a task handle of // a target application, whereupon the hook is installed in that // application. When receiving a NULL parameter, the hook installs // to applications. // CBT hook -- allow CBT system to stop events from progressing hCBTHook = SetWindowsHookEx(WH.sub.-- CBT, (HOOKPROC)cbtFilter, hInstance, NULL); // Msg filter hook -- dialog boxes and menus hMsgHook = SetWindowsHookEx(WH.sub.-- MSGFILTER, (HOOKPROC)msgFilter, hInstance, NULL); // Get msg hook hGetMsgHook = SetWindowsHookEx(WH.sub.-- GETMESSAGE, (HOOKPROC)getmsgFilter, hInstance, NULL); // Windows hook hCallWndProcHook = SetWindowsHookEx(WH.sub.-- CALLWNDPROC, (HOOKPROC)callWndProcFilter, hInstance,NULL); return( hCBTHook && hMsgHook && hCallWndProcHook && hGetMsgHook); } __________________________________________________________________________

As shown, a callback function is installed for each hook; each callback function, in turn, serves as an entry point into an ATU. Additional information on the installation and use of hooks in Microsoft Windows can be found in the Microsoft Windows Software Development Kit referenced above.

In addition to installing hooks to trap system messages, one or more application-specific hooks (callback functions) 439 are installed as well. For instance, a target application may be a spreadsheet which includes its own internal set of messages, for example, SS.sub.-- CELLSELECT, SS.sub.-- CELLEDIT, SS.sub.-- BLOCKDEFINE, and the like. To monitor these messages, an ATU may register an application-specific callback function with the target application, in effect dynamically binding the ATU to its target application. At runtime, the application invokes the callback function for dispatching internal or application-specific messages. Thus, the CBT system of the present invention is not limited to Windows events and their messages; instead, the CBT system may receive and translate any messages of interest, whether system-wide or strictly internal.

At step 440, events which are trapped by the Application Translation Units 340 are "dispatched" to the Message Engine 350 as CBT message/event information objects. As shown in particular detail in FIG. 4B, the dispatch module of each ATU includes a function corresponding to each Windows event. Thus, for the WM.sub.-- COMMAND, WM.sub.-- MENUSELECT, WM.sub.-- BUTTONDOWN, and WM.sub.-- SETCURSOR messages, the following translate functions may be defined:

int doWM.sub.-- Command (CbtEntry *pEntry, MSG *msg);

int doWM.sub.-- MenuSelect(CbtEntry ,pEntry, MSG *msg);

int doWM.sub.-- ButtonDown(CbtEntry *pEntry, MSG *msg);

int doWM.sub.-- SetCursor (CbtEntry *pEntry, MSG *msg);

each designed for processing its particular event.

The operation of an ATU dispatcher will be demonstrated by the processing of Windows messages for determining if the cursor has traversed a window boundary (i.e., entered a new window); this example illustrates how a multitude of Windows WM.sub.-- SETCURSOR messages can be converted into MOUSELEAVE and MOUSEENTER meta-messages. The dispatching of other events as CBT messages is set forth hereinbelow as Appendix A.

As shown in FIG. 5A, a doWM.sub.-- SetCursor dispatch method 500 is invoked whenever a SetCursor message is trapped by the ATU (i.e., captured by a hook function before the event has been received by the application). Here, the script writer is not interested in the screen cursor simply entering an already active window; thus, the method simply "allows" the Windows message to be passed to the target application at step 501 and returns. Specifically, since the screen cursor is entering a window which is already active, no particular CBT-generated message or other intervention is desired by the script writer at this point; hence, the WM.sub.-- SETCURSOR message is allowed to pass through.

Continuing the example, the script writer may specify that the event of a cursor leaving an active window is of interest and should be trapped. Since the cursor is not simply re-entering the active window (no at step 501), the window which the cursor is leaving should be notified of the event. The CBT system notifies the Message Engine of this action as follows. First, at step 502, the information pertaining to the window where the event occurred is encapsulated into a C++ object (which is derived from an EventInfo class hierarchy, described in further detail hereinbelow). At step 503, the information object and a "MouseLeave" message are dispatched to the previous (departing from) window, with the message being denoted as a "NOTIFY" message.

In a preferred embodiment, two different classes of messages are provided: CBT.sub.-- NOTIFY and CBT.sub.-- CONFIRM. Those messages which are purely informational, such as mouse movements, are CBT.sub.-- NOTIFY messages. Those which can be stopped before they reach the target application, on the other hand, are called CBT.sub.-- CONFIRM messages. Each is registered with Windows as an application-specific Windows event. Using two methods defined within EventInfo objects, the script 320
can allow or prevent a CBT.sub.-- CONFIRM type message from reaching the target application. Specifically, a stopMessage method is invoked which determines (based on script instructions) whether to allow the message to pass through to the target application.

After step 503, the method proceeds to alert the Message Engine that the cursor is entering a new window. In a manner similar to sending the "MouseLeave" message, the method first builds an EventInfo object at step 504. Next, a "MouseEnter" message of type CBT.sub.-- NOTIFY is dispatched to the application, along with the information for the event (EventInfo object), at step 505. At step 506, an active window flag is set; this is the flag that is read in step 501 to determine if the mouse cursor is entering the active window. Finally, the method concludes by passing the message on to the application (i.e., "allow" message) at step 507. At the conclusion of the method, memory for the EventInfo objects may be recovered (e.g., using manual or automatic garbage collection techniques).

For purposes of illustration, one may implement such a method in the C++ programming language as follows:

__________________________________________________________________________ int doWM.sub.-- SetCursor(CbtEntry *pEntry, MSG *msg) EventInfo *pObject = (EventInfo *)NULL; // init EventInfo if( pEntry->hActiveWnd == (HWND)(msg->wParam) ) // steps 510/507 return MSG.sub.-- ALLOW; pObject = new WindowInfo( pEntry->hActiveWnd ); // step 502 DispatchToCbt(pEntry, CBT.sub.-- NOTIFY, TM.sub.-- MOUSELEAVE, (LPARAM)pObject); // step 503 pObject = new WindowInfo( (HWND)(msg->wParam) ); // step 504 DispatchToCbt(pEntry, CBT.sub.-- NOTIFY, TM.sub.-- MOUSEENTER, (LPARAM)pObject); // step 505 pEntry->hActiveWnd = (HWND)(msg->wParam); // step 506 return MSG.sub.-- ALLOW; // step 507 // garbage collection performed by the system } __________________________________________________________________________

Here, pEntry is a pointer to a record, CbtEntry, which is updated. The record includes handles to the hooked task (application that has been hooked) and the currently active window:

______________________________________ typedef struct .sub.-- CbtEntry { HTASK hHookedTask; HWND hActiveWnd; } CbtEntry; // pEntry points to this ______________________________________

As shown, a meta-message may maintain its own data structures for tracking events at a higher level (e.g., the status of the active window).

The DispatchToCbt function, on the other hand, is conveniently viewed as two halves of a process. Specifically, the Message Engine registers a callback function with the ATU. The operation proceeds as follows. On the ATU side, the ATU passes to the DispatchCBTMessage method a task handle for identifying a particular application; since the system processes multiple applications, this mechanism serves to distinguish between different applications (and their instances):

__________________________________________________________________________ void DispatchToCbt(HTASK hHookedTask, UINT Msg, WPARAM wParam, LPARAM lparam) if( pCbtDispatchFunc ) (pCbtDispatchFunc)(hHookedTask, Msg, wParam, lparam); } __________________________________________________________________________

In this manner, an ATU filters or groups events by target application and communicates its events as meta-messages to other CBT components. At the completion of step 440 of FIG. 4A, the ATU has dispatched the CBT message, its type, and the EventInfo object to the Message Engine 350, thereby fully communicating an event which it has trapped.

Not all events are of interest, however. Thus, the events should be filtered so that only those of interest are acted upon. At step 450, the Message Engine performs this task by comparing the CBT message against known event handlers. In other words, the engine attempts to dispatch CBT messages of interest--ones having a handler define for the event. Thus on the Message Engine side, the Message Engine determines which entry in its table corresponds to that task:

__________________________________________________________________________ void CALLBACK MessageEngine::DispatchCBTMessage(HTASK hTarget, UINT cbtmsg, WPARAM wParam, LPARAM lParam) CbtSession *pSession = GetSessionFromTask(hTarget); if( pSession ) SendMessage(pSession->hMsgPane, CbtMessages[cbtmsg], wParam, lparam); // Notify application handler // where pSession is the current application session // (determined from hTarget); // CbtMessages[cbtmsg] is the table lookup for the // CBT message ("confirm" or "notify"; // wParam is the CBT message type (TM.sub.-- msg); and // lParam is a pointer to the EventInfo object. } __________________________________________________________________________

With particular reference to FIG. 4B, this process is illustrated. Message Engine filters CBT messages through a plurality of message handlers, including, for example, a TargetApplication handler 451, a TargetWindow handler 453, a Custom handler
455, and a Default handler 457; other exemplary handlers are set forth hereinbelow as Appendix B. CBT messages of interest will be passed to a particular handler. As shown in Appendix B, each message belongs to a particular Message Handler Class and is either informational (CBT.sub.-- NOTIFY) or preventable (CBT.sub.-- CONFIRM). A "mouseEnter" message, for example, belongs to a TargetWindow Handler Class; the message is therefore processed by the TargetWindow handler 453. An application specific event, such as an SS.sub.-- EDITCELL message from a spreadsheet target application, on the other hand, would typically be passed to the TargetApplication handler 451. Finally, messages without a handler, that is, those which the script writer has no particular interest, may be passed to a default handler (e.g., for ignoring, discarding, or otherwise invoking a desired default action); thus, the script writer need only enumerate handler methods for messages of interest.

If matched with a handler, the message is then dispatched. Specifically, the handler extracts properties for the message and the accompanying EventInfo object. For a message of TargetWindow handler class, for instance, available object properties include:

1) Name: Title string of the control;

2) Class: Windows class of the control;

3) ID: Resource ID of the control;

4) Style: Style flags of the control;

5) Enable: Whether the control is enabled;

6) Position: Coordinates of the control; and

7) EventInfo: Current EventInfo object, if any.

Additional exemplary properties which are available for various messages handler classes are set forth hereinbelow as Appendix C.

As an alternative to defining several event handlers, a more preferred embodiment provides only two basic event handlers: an Application Link handler and a Window Link handler. Each is an object having various handler methods for appropriately responding to each message passed. Based on the lookup performed by the Message Engine (i.e., NOTIFY or CONFIRM), an Application Link handler may, for example, effect the dispatching as follows:

__________________________________________________________________________ RESULT CALLBACK ApplicationLink::wndProc(HWND hwnd, UINT Message, WPARAM wParam, LPARAM lparam) // First, get handler (from link registered) ApplicationLink *pWindow = (ApplicationLink *)GetWindowLong(hWnd, 0); if( pWindow ) { if( Message == CBT.sub.-- notify ) // Msg is a NOTIFY msg { EventInfo *pInfo = (EventInfo *)lParam; // Recall that Event info includes one or more of a win class, // name, and resource ID. If exact information is not provided // (e.g., just "OK" button), do "fuzzy" match (i.e., match as // much as possible: pWindow->doDispatchNotify(wParam, pInfo); pInfo->Finished( ); return TRUE, } else if ( Message == CBT.sub.-- confirm // Msg is a CONFIRM msg { EventInfo *pInfo = (EventInfo *)lParam; // Event info pWindow->doDispatchConfirm(wParam, pInfo); pInfo->Finished( ); return TRUE, } } // on return, call WinProc instantiated w/ applic. link return DefWindowProc(hWnd, Message, wParam, lParam); } __________________________________________________________________________

Here, the doDispatch- methods communicate directly with the Script Engine. In turn, the Script Engine responds according to script objects defined within the active script. By invoking the stopMessage method for specifying whether an event is allowed, for example, events may be stopped from reaching a target application; in most instances, however, the script writer will simply specify the default--that the event should simply pass on through to the target application.

The script writer may provide methods for handling the various events of interest, or he or she may rely on default methods which are defined by the CBT system. In operation, a CBT message is passed to the objects. Each object, in turn, is responsible (through its method definitions) for knowing which messages are of interest, and how each one of interest should be processed. In a target application, for example, if the script writer hooks a window link up to a list box of the application, he or she should provide methods for handling the event (as communicated by CBT messages) of that list box.

Referring now to FIG. 5B, the overall method of dispatching messages is summarized. In a doDISPATCH method 520 of the present invention, a CBT message arrives at the Message Engine and is processed as follows. First, in step 521, an attempt is made to match the message to an application link handler. If the attempt is unsuccessful (no at step 522), then the message is simply allowed to "fall through" (i.e., left unprocessed, or processed by a default handler). Otherwise (yes at step 522), at step 523 the Message Engine forwards the CBT message (with EventInfo object) to the identified Application Link handler.

At step 524, the Application Link handler examines the EventInfo object and attempts to match it with a registered window link. If this step is unsuccessful (no at step 525), then a Default handler will be assigned for processing the event at step 526. At step 527, the message is forwarded to the Window Link handler. The Window Link handler, in turn, dispatches the message to the Script Engine at step 528. At this point, the Script Engine identifies the event by mapping the message into its set of known reserved words. At step 529, the Script Engine processes the message according to the instructions of the script (i.e., effects the action desired by the script writer, as indicated by the use of the matching reserve word). Upon completion of this step, the method has successfully dispatched the meta-message, with appropriate actions being effected in response to that message.

2. Building CBT Lessons

As a tutorial is designed, the CBT script writer creates a "storyboard" showing the visual appearance as well as the flow of the tutorial. The storyboard becomes the basis for the CBT lesson script.

CBT scripts are written in a simple language which contains both message handling and object-oriented features. Each lesson script is organized as a collection of scenes, with each scene describing the actions that take place at a particular point in the lesson. For example, a scene might instruct the CBT system to display a window containing some text when a particular menu item is chosen in the target application. As the lesson script proceeds, new scenes can be performed. This process continues until the user chooses to exit the CBT or the lesson is finished.

To control the target application, the CBT system intercepts all Windows events for the application and translates them into CBT messages. These messages will trigger any corresponding message handlers which are defined within a scene. When a message handler is triggered, its script is executed.

Within each scene, message handlers are defined for each UI control in the application which is of interest. For example, to respond to a button click within the script the following handler is defined:

______________________________________ script for Scene1 TargetButton theButton(120) on theButton.buttonClick theCBTLesson.perform("Scene2") end end ______________________________________

This hypothetical scene creates a TargetButton object which is linked to the UI control in the target application; the UI control Resource ID is 120. Next, a Message Handler is defined for the buttonClick message from the TargetButton object. When this message is received, the Script Engine performs a new scene, Scene2. Thus, the statement:

calls the perform method of the global object "theCBTLesson" (the CBT Lesson object).

In addition to controlling user actions, the CBT lesson also drives the target application autonomously by sending appropriate messages. Alternatively a sequence of events can be recorded (e.g., using a tool similar to MS-Windows Recorder) for later replay.

The script may also query the target application for its current properties. If a window link is established to a button, for instance, the script may query the button for its properties, such as its size, its title, and the like. One should note that the ability to query for properties operates independently of the processing of events and their messages. As another example, a target application could be asked to enumerate all the buttons of a dialog box. The script may, in turn, act on the queried information, including modifying selected resources. In this fashion, the resources of a target application may be dynamically varied, thereby providing the target application with an alternative user interface--one having UI components which may be altered on the fly.

Appended herewith as Appendix D is a source listing illustrating an exemplary script syntax and methodology for operating the CBT system of the present invention. Additional reference materials illustrating a preferred script syntax are appended herewith as Appendix F.

3. Multiple-application Lessons

As shown by FIG. 3, the system of the present invention is operative with one or more applications 360. More particularly, according to the present invention, a single script 320 may be employed to not only control multiple applications concurrently, but also control interaction between multiple applications. A script may be provided for tutoring the user in the operation of cutting and pasting between applications, for instance, cutting text from a word processor and pasting it into a database application. Thus, the CBT system 130 is not application bound; instead, it is a complete subsystem--one which may control multiple applications, including interactions between applications and/or the operating system, even launching additional applications as needed.

4. Event Information (EventInfo) Objects

An EventInfo object, which stores information about a particular event, is instantiated from an EventInfo class 550. FIG. 5C illustrates the EventInfo inheritance tree and lists the properties for each class. As shown, the EventInfo class hierarchy 550 includes nine derived EventInfo classes which contain the state information about the various standard CBT messages. At the root of the hierarchy is the EventInfo base class 551. In a preferred embodiment, this class may be constructed with the following C++ class definition:

______________________________________ class SHARED.sub.-- CLASS EventInfo : public CbtObject, public pEventInfo { public: ATOMTABLES(Keyword, 7) protected: HWND hwndTarget; BOOL bAllowMsg; BOOL bIsProcessing; public: EventInfo(HWND hwnd); virtual .about.EventInfo( ); virtual int Supports(hProtocol &Hdl) const; inline HWND WindowHandle( ) const; virtual const char * WindowName( ) const = 0; virtual const char * WindowClass( ) const = 0; virtual int WindowId( ) const = 0; virtual LONG WindowStyle( ) const = 0; virtual BOOL AllowMessage(BOOL bFlag, BOOL bState); virtual BOOL ProcessingMsg( ) const; virtual void Finished( ); inline void * operator new(unsigned size); inline void operator delete(void *p); ATOMMETHODS(Keyword) CLASSMETHODS(EventInfo, "EVENTINFO") }; ______________________________________

As shown, the EventInfo class 551 provides access to the Windows name and its class, its resource ID, its Windows style, and whether the message is allowed (according to the script 320).

Derived from EventInfo class 551 is WindowInfo class 552, a pure virtual base class for other EventInfo classes. The subclass affords the same four pieces of information which were provided by the base class 551. In addition, for a given window handle, the object will extract a window name, class, ID, and style. The class may be constructed with the following C++ class definition:

______________________________________ class SHARED.sub.-- CLASS WindowInfo : public EventInfo { protected: int iWindowId; LONG lWindowStyle; char * strWindowName; char * strWindowClass; public: Windowinfo(HWND hWnd); virtual .about.WindowInfo( ); virtual int Supports(hProtocol &Hdl) const; virtual const char * WindowName( ) const; virtual const char * WindowClass( ) const; virtual int WindowId( ) const; virtual LONG WindowStyle( ) const; CLASSMETHODS(WindowInfo, "WINDOWINFO") }; ______________________________________

In addition to the windowing information, other events are also of interest, particularly mouse and keyboard events. These other events are accessible through subclasses of WindowInfo class 552. Specifically, the WindowInfo class spawns five subclasses: WinHelpInfo class 561, WinPositionInfo class 563, WinShowInfo class 565, WinSelectInfo class 567, and KeyboardInfo class 569. As shown, objects may be instantiated from these subclasses for accessing help text, window position, menu information, keyboard events, and the like. WinHelpInfo, for instance, contains the text which was sent by the WinHelp engine to the CBT. This text can be a sequence of script statements which are executed or simply a string of text. WinPosition provides the coordinates of the window. WinShowInfo contains the SWP.sub.-- flags corresponding to the Windows ShowWindow() function. WinSelectInfo contains the name of the selected menu or control window item. KeyboardInfo contains the key that was pressed as well as any flags indicating if the <ALT>, <SHIFT>, or <CTRL> keys were also pressed.

Two classes, WinPositionInfo class 563 and WinSelectInfo class 567, spawn additional subclasses. As shown, MouseInfo class 571, which is derived from WinPositionInfo class 563, provides direct access to mouse events; it contains the mouse button that was clicked and whether it was a single or double click as well as the position of the mouse. WinSelectInfo class 567, on the other hand, spawns MenuInfo class 573 and ListInfo class 575. The former provides access to menu IDs and menu flags, the flags indicating whether the menu item is grayed, highlighted, or the like; the latter provides the index of the selected item, which is important if the list does not contain text.

Second embodiment: Computer-aided Software Testing System

A. Introduction

An alternative embodiment of the present invention, one adapted for automated testing of GUI applications, will now be presented. Before discussing the construction and operation of the computer-aided software testing embodiment in detail, it is helpful at the outset to briefly review parameters for quantifying the effectiveness of software testing.

By automating the tasks of "controlling" and "observing" the target software, the burden of testing for the QA engineer is substantially reduced. In other words, the more successfully the QA engineer can control and observe the software program, the more effective the test becomes to automate. This can be restated as a formula:

where Controllability is a rough measure of how effectively the program can be automatically controlled into a given state, and Observability is the measure of how effectively a bug can be automatically observed when the program being tested is in a faulty state. Test Automation Effectiveness is the product of these two factors for the program under examination.

Consider, for example, the task of testing a software compiler having both command-line and GUI components (such as found in commercially-available compilers from Borland and Microsoft). Conceptually, testing the GUI portion of the compiler should be easier than testing the command-line portion, owing to the latter's complex internal operations. In practice, however, it is far easier to test the command-line portion. The QA engineer need only develop test programs or data files that are run through the compiler using a simple batch file. Here, it is easy to achieve total control for the non-GUI portion. Observability is also high because return values or variables denoting the state of the compiler can be tested and written to a file using simple program statements that print the results.

The task of automating the control of a GUI program, in contrast, is particularly difficult. Specifically, the task of maintaining pre-recorded input (i.e., keyboard and mouse actions) and test scripts is not only very time consuming but also prone to error. If the degree of controllability and observability of the automated tests can be increased, the task of testing GUI software may be made easier.

According to the present invention, the QA engineer constructs a high-level model of an application's UI using prefabricated building blocks. This high-level model serves as a middle ground between test scripts and the application being tested. The knowledge of how a given UI element is controlled or how it can be observed is retained in the model rather than in a test script. Consequently, the test script consists of easy-to-maintain, high-level testing commands only. For instance, an exemplary script for opening a particular file (e.g., "testl.txt") may be defined as follows:

__________________________________________________________________________ TheApp.FileMenu.Open.Select( ) // Select File open menu // ActiveDialog now represents the File Open dialog ActiveDialog.Filename.Set("test1.txt") // Type in the file name ActiveDialog.OK.Click( ) // Click on the OK button __________________________________________________________________________

One need only include high-level statements defining the operation, such as shown above.

B. System overview

Referring to FIG. 6, a software testing automation system 600 of the present invention includes Generic Element Models (GEMs) library 610, Application-specific Testing Models (ATMs) 615, Resource Database 620, one or more Model Generators 625, Test Runtime Library 630 (for storing Test Objects or Test Functions, typically as one or more Dynamic Link Libraries (DLLs)), Script (interpreter) Engine 635, Message Engine with one or more Application Translation Units (ATUs) 640, and an in-memory Testing Model 655 (of the application under test), and one or more test scripts 660.

Communication directly with the application under test 665 is effected through Windows API 670 and/or Test Port 650. Using Windows API 670, various Windows messages may be sent or posted to the application for querying the application or effecting particular operations, as is known in the art. For instance, one may determine at runtime whether a checkbox screen object is checked (i.e., toggled to its "Checked" stated) by invoking the standard Windows API call:

where, hwndDlg is the handle of dialog box and idButton is the identifier of the button/checkbox. Documentation for the Windows API is provided in the Microsoft Windows Software Development Kit (SDK), available directly from Microsoft (and several other vendors). In addition to communication with the application under test 665 through the Windows API, application-specific messages may be registered with the Windows OS, as is known in the art. Test port 650 represents the use of such application-specific messages for communicating with the application under test 665.

In general operation, the system employs the Model Generator 625 to decompose the application under test 665 to generate the Application-specific Testing Models (ATMs) 615. Each ATM is a high-level model for a specific component of the application being tested, such as a File Open dialog. ATMs describe the actual component which they represent in terms of Generic Element Models (stored in GEMs Library 610), which will now be introduced.

A GEM encapsulates the behavior of irreducible user interface elements such as push buttons, checkboxes, listboxes, and the like. GEMs are instantiated when an ATM corresponding to an active state of the application under test is activated by the Model Manager 645. GEMs load their expected results from the Resource Database 620 and are capable of binding themselves dynamically to the actual object on the screen which they represent. Each GEM therefore can perform a "self test" on itself by simply comparing its expected result (as stored in the Resource Database 620) to its actual behavior (as displayed on the screen at runtime). In this manner, an entire application can perform a self test by simply asking all its components to test themselves in turn.

Driven by a test script 660, the Testing Model 655 of the application under test 665 employs both Windows APIs 670 and/or the Test Port 650 to control the application being tested into various states; in each state, the results generated may be observed for error. The Model Manager 645 monitors the state of the application under test using the Message Engine and Application Translation Units (ATUs) 640 (components which are the same as those described above for the first embodiment). ATUs translate low-level messages into high-level messages, dispatching those events that the Model Manager 645 has registered an interest in. Driven by the changes in the state of the application under test 665 and/or events occurring in the system, the Model Manager 645 instructs the Script Engine 635 to load and execute the appropriate ATM 615 which corresponds to the active state of the application under test. The test script 660 employs the Test Library 630 and the Testing Model of the Application
655 to carry out the test execution task. The construction and operation of the system components will now be described in detail.

C. Application-specific Testing Models

Application-specific Testing Models (ATMs) provide high-level representation for specific components of the application under test. An ATM for a File Open dialog, for instance, describes the actual dialog in terms of the fundamental, pre-constructed building blocks which it comprises. In this manner, an entire application can be described using a library of ATMs.

In a preferred embodiment, at least five categories of ATMs are automatically provided by the system: Menus, ToolBars, Dialogs, Client Area Windows, and a Status Line. Each of these categories corresponds to a base class pre-supplied with the system. Each base class encapsulates functionality associated with its category. A menu base class, for instance, includes data members for describing menus and methods operative on those data members. The base classes simplify the Model Generator's task of creating ATMs. Moreover, the base classes allow the QA engineer or software developer to construct custom ATMs by simply deriving specific instances of the pre-supplied base classes.

Consider, for instance, an application having a File Open dialog. An ATM representing such a dialog may be created by deriving a FileOpenDlg class from the pre-existing base class for dialogs. The pre-supplied base class for dialogs ("BaseDig" class) already encapsulates the behavior and properties of a generic dialog box. Thus, the task of modeling a File Open dialog is greatly simplified. In a like manner, other aspects of the application may be modeled using the base classes which are provided for the other categories. For instance, the application's menu and client area may be modeled using pre-supplied BaseMenu and BaseClientArea classes, respectively.

Before describing ATMs in further detail, it is helpful to review runtime construction of screen objects, such as dialog boxes in Microsoft Windows. Consider a typical dialog, such as confirmation message box dialog 700 shown in FIG. 7A. The dialog 700 comprises a screen window 705 which includes a plurality of elements, including caption bar 710, static text 715, and buttons 720, 725, 730. Each of these "children" may, in turn, include still further elements. Button 720, for instance, includes a caption text or label: "Yes".

FIG. 7B shows the resource or programming statements attendant to the creation of each element of dialog 700 (e.g., in Microsoft Windows). The first line of resource file 750, for instance, defines a DIALOG screen element, named DIALOG.sub.-- 1
(identifier), and includes screen coordinates (111, 98, 151, 59) specifying a starting location and size (ending location). The second line specifies various window attributes or "window styles" that the dialog 700 is to assume. Dialog 700, for instance, is to be a "popup" window (Microsoft window style=WS.sub.-- POPUP) with a caption bar (Microsoft window style=WS.sub.-- CAPTION) and a system menu (Microsoft window style=WS.sub.-- SYSMENU). A caption statement (CAPTION "Confirmation") specifies that the text "Confirmation" is to be displayed in the caption bar 710.

Also shown in the resource script are the children screen elements of the dialog 700. Specifically, the static text 715 is defined by the LTEXT statement, while buttons 720, 725, 730 are defined by PUSHBUTTON statements. The definitional statement for each of these includes information similar to that of the parent dialog window--namely, a caption or label, an identifier (resource identifier), a starting location and size, and window styles. Button 710, for instance, is defined by a resource statement having a caption of "&Yes" (& tells Windows to underline the immediately following character), an identifier of 101, screen coordinates of 24, 35, 24, 14, and window style of WS.sub.-- CHILD (child window), WS.sub.-- VISIBLE (visible), and WS.sub.-- TABSTOP (tab stop for setting input focus to this element with tab key).

The foregoing relationship between resource (.RC) script files and screen objects is well known in the field of Windows programming. For those readers unfamiliar with these concepts (i.e., non-Windows programmers), it is strongly suggested that the following references be consulted: Petzold, C., Programming Windows (particularly, Section III: Using Resources); Borland C++: Resource Workshop; and Microsoft Windows Software Development Kit/Microsoft Visual C++.

FIG. 7C illustrates that resource statements in a resource file are at runtime actually translated into Windows API calls. In particular, a given resource is created at runtime by calling Windows CreateWindow function, with the aforementioned attributes passed to Windows as parameters. CreateWindow is a standard Windows API call (and is therefore fully documented in the Petzold and Microsoft materials, as well as numerous other references). The "Yes" button 720, for instance, is created by the following CreateWindow call:

______________________________________ CreateWindow ( button, // window class name "&Yes", // window caption WS.sub.-- CHILD // window style WS.sub.-- VISIBLE WS.sub.-- TABSTOP, 247cxchar,4, // initial x position 351cyChar*8, // initial y position 241cxChar*4, // initial x size 141cyChar*8, // initial y size hDlg, // parent window handle NULL, // window menu handle hInstance, // Program instance handle NULL // creation parameters ); ______________________________________

At runtime, therefore, the Message Engine may trap each CreateWindow API call for determining each screen object which is about to be created.

Using known techniques, the Model Generator 625 may decompose the resources of an application to automatically construct a model specific for the application--the Application-specific Testing Model or ATM. Known techniques include reading Windows resource information. Since resources may change dynamically at runtime (e.g., menu item becoming "grayed out"), it is more preferable to dynamically poll resources at runtime. Thus in a preferred embodiment, the resource information is learned dynamically--t the application's runtime--using Windows API (calls) 670, or Test Port 650 (in the instance of a proprietary control object). Exemplary calls are provided below (with reference to the Notepad example of FIGS. 8 and 9). Once the application's resources are learned, the Model Generator proceeds to construct each ATM by declaring the individual resources in a class definition derived from BaseDlg--the generic dialog class which encapsulates all the properties and functionality found in an empty dialog box.

In an exemplary embodiment, BaseDlg may be constructed from the following class definition:

__________________________________________________________________________ // BaseDlg class definition class BaseDlg( ) //Begin Constructor Moniker = 0 //Live link to the actual dialog on the screen Id = 0 //Unique id used as index to fetch expected //values for this dialog from the resource dbase Caption = 0 //Dialog's expected caption ParentName = 0 //Parent Window's Caption or Name Style = 0 //This dialog's style Dimension = 0 //Coordinates and size Children = 0 //Reference to the components of this dialog NumOfKids = 0 //Number of components in this dialog ParentPtr = 0 //Reference to application which owns this dialog //End constructor function Init( Cl, Cap, App, ResId ) // Class, Caption, Application // reference and Unique Id //Establish a live link to the actual dialog on the screen //by instantiating a WindowProxy object called "Moniker" WindowProxy Moniker("", Cl, Cap ) //Set ParentPtr and Id ParentPtr = App Id = ResId //Fetch expected values from the resource database //using the "Id" key Defaults.LoadData( Id, Caption, ParentName, Style, Dimension //Bind each component to the corresponding element on the screen i = 0 while ( i < NumOfKids ) Child = Children [ i ] Child.Bind( ) //See BaseGem's Binds member function i + 1 end end function DetachChildren( ) // Detach from the actual dialog on the screen i = 0 while ( i < NumOfKids ) Child = Children[ i ] Child.Detach( ) i = i + 1 end Monker = 0 end function SelfTest( ) //Bist or Built-in-self-test controls verification and logs //messages to the appropriate file in a specified format Bist.Start( "Dialog selftest", Id ) Bist.Verify( "Caption", Caption, Moniker.Caption ) Bist.Verify( "Parent", ParentName, ParentPtr.Moniker.Caption ) Bist.Verify( "Size", Dimension, Moniker.GetSize( ) ) Bist.Verify( "Style", Style, Moniker.GetStyle( ) ) Bist.Verify( "Children",NumOfKids, Moniker.GetNumOfKids( ) ) //Instruct each component of the dialog to test itself i = 0 while ( i < NumOfKids ) Child = Children[ i ] Child.SelfTest( ) //Each Child has a SelfTest( ) //member function i = i + 1 end Bist.EndSection( ) end // Capture attributes of this dialog and its children function Capture( ) // . . . place holder for custom code; insert as desired end // Move window X pixels horizontally and Y pixels vertically function Move( X, Y ) // . . . place holder for custom code; insert as desired end // Size window X pixels horizontally and Y pixels vertically by // stretching the bottom-right corner. function Size( X, Y ) // . . . place holder for custom code; insert as desired end end __________________________________________________________________________

As shown, BaseDlg includes a Moniker data member. Moniker is a live link to the actual (corresponding) object on the screen. The link is created by calling WindowProxy, a C++ class which encapsulates Windows information for the object. In an exemplary embodiment, WindowProxy includes the following C++ class constructor:

__________________________________________________________________________ WindowProxy::WindowProxy(const char *strTask, const char *strClass, const char *strTitle) strWindowName = (char *)new char[MAX.sub.-- WINDOW.sub.-- NAME + 1]; strWindowClass = (char *)new char[MAX.sub.-- WINDOW.sub.-- NAME + 1]; strTaskName = (char *)new char[MAX.sub.-- WINDOW.sub.-- NAME + 1]; *strWindowClass = *strWindowName = *strTaskName = ' 0'; if( strTask ) STRCPY(strTaskName, strTask); hTargetTask = getTaskHandle(strTask); hWndTarget = FindWindow(strClass, strTitle); BindToWindow(hWndTarget); } __________________________________________________________________________

The constuctor includes a call to a BindToWindow method, which may be constructed as follows:

__________________________________________________________________________ BOOL WindowProxy::BindToWindow(HWND hWnd) BOOL bStatus; if( hWnd && iWindow(hWnd) ) { GetWindowText( hWnd, strWindowName, MAX.sub.-- WINDOW.sub.-- NAME ); GetClassName ( hWnd, strWindowClass, MAX.sub.-- WINDOW.sub.-- NAME ); iWindowId = GetWindowWord(hWnd, GWW.sub.-- ID); lWindowStyle = GetWindowLong(hWnd, GWL.sub.-- STYLE); hWndTarget = hWnd; bStatus = TRUE; } else { *strWindowName = *strWindowClass = ' 0'; iWindowId = 0; lWindowStyle = OL; hWndTarget = (HWND)NULL; bStatus = FALSE; } hCurrItem = (HWND)NULL; return bStatus; } __________________________________________________________________________

As shown, WindowProxy's constructor encapsulates Windows handle, task handle, caption, and the like for the object. In the event that the object is not a Windows control (e.g., it is a spreadsheet or other implementation-specific object), the information may be nevertheless encapsulated (as shown by the above "else" clause of BindToWindow). Other exemplary method prototypes of WindowProxy are set forth hereinbelow in Appendix E.

Regardless of how it is implemented, the Moniker, once created, is an encapsulation of the actual object on the screen. The properties of the Moniker fully describe the corresponding object on the screen, thus giving a live link to the actual screen object. The Resource Database, on the other hand, stores expected data for the object. The two may be compared against each other (e.g., during SelfTest) for detecting runtime errors, as well as modeling problems.

In addition to the Moniker, the BaseDlg class includes two other data members: Children--a reference to the components of the dialog, and NumOfKids--number of components (kids) in the dialog. These are employed in the SelfTest method of BaseDlg.

______________________________________ function SelfTest( ) i = 0 // init to first kid while ( i < NumOfKids ) Child = Children[ i ] // test this kid Child.SelfTest( ) i = i + 1 // increment to next kid end end ______________________________________

As shown, this allows a dialog object to test itself by asking its children to, in turn, test themselves (by calling their self test methods--Child. SelfTest()).

Moreover, SelfTest may include various levels of testing--each level denoting more extensive testing. For instance, maximum (all levels) testing may be performed by:

______________________________________ function SelfTest( MaxLevels ) i = 0 iLevel = 0 while ( iLevel < MaxLevels ) // step through all levels while ( i < NumOfKids ) Child = Children[ i ] Child.SelfTest( iLevel ) i = i + 1 end iLevel = iLevel + 1 // increment to next level end ______________________________________

In contrast, more expedient (but less thorough), high-level testing may be performed by:

______________________________________ function SelfTest( iLevel ) i = 0 while ( i < NumOfKids ) Child = Children[ i ] Child.SelfTest( iLevel ) // just test single level i = i + 1 end end ______________________________________

During creation of the dialog at runtime, the Message Engine traps the CreateWindow message for the dialog of interest (which the Model Manager has registered an interest in). In this manner, the Model Manager 645 receives notification from the Message Engine 640 that the dialog is about to be created. Upon receipt of the notification, the Model Manager locates an existing model of the dialog. If a model is found, the system is instructed to load the dialog and execute it, whereupon the Init() method of the BaseDlg class is invoked. Init initializes the Moniker, thereby creating the live link to the actual dialog. Also during Init (as shown above), the children are asked to bind themselves to their corresponding screen objects. If one of the children is a checkbox, for instance, it becomes bound to the actual checkbox on the screen.

The opposite of Init is Detach. Upon the Model Manager's receipt of a message that the dialog is about to be removed from the screen (upon occurrence of a DestroyWindow message for the dialog), the Detach method is invoked for freeing memory allocated for the BaseDlg instance which had been created. The Moniker for the dialog is also freed (set to 0).

This process will be illustrated for the confirmation message box dialog 700. For the dialog 700, a model may be constructed from the following class definition, derived from BaseDlg:

______________________________________ class ConfirmationDlg of BaseDlg Button Oui ( Self, 101 ) // line 2 Button No ( Self, 102 ) Button Cancel ( Self, 103 ) Static TheText ( Self, -1 ) // line 5 NumOfKids = 4 // Component array Components = [ NumOfKids ] Components[ 0 ] = Oui Components[ 1 ] = No Components[ 2 ] = Cancel components[ 3 ] = Text Children = Components end ______________________________________

The first line declares a class, ConfirmationDlg, which is derived from BaseDlg, the existing class. As shown above, each ATM includes a component array containing all its elements. The member functions of the BaseDlg class manipulate these components once they have been constructed. ConfirmationDlg inherits the data members and member functions of BaseDlg and instantiates the list of elements for FIG. 7A, as shown in lines 2 to 5 above. Oui, No, and Cancel represent objects within the confirmation dialog 700 which are instances of a Generic Element Model (GEM) of type Button.

In an exemplary embodiment, each GEM (described in detail below) takes two parameters to construct itself: a reference to the parent dialog (Self) and a unique identifier (either its pre-existing resource id, or one derived by the system). In the present example, the Yes button 710 on the message box 700 has been represented by an identifier called Oui in order to emphasize that there are no assumptions about textual information made in the ATM. The object Oui loads its expected label text (the text string "Yes") from the Resource Database during initialization time. Consequently, the ATM above can represent different language versions of the Confirmation dialog without any modifications. The language difference is encapsulated in the Resource Database (described below) which is generated automatically.

As shown above, ConfirmationDlg itself is a class definition. To use ConfirmationDlg at runtime, an instance is created. In an exemplary operation of the preferred embodiment, an instance (variable), ActiveDig, of the ConfirmationDlg class is created at runtime. More particularly, the instance is created when the Model Manager 645 detects that the actual Confirmation dialog has been activated, as shown by the following test script instructions:

______________________________________ ConfimationDlg ActiveDlg( ) // Instantiate ActiveDlg ActiveDlg.Init( ClassName, Caption, Application, ResourceId ) // Initialize ______________________________________

When instantiated, ActiveDig (or more specifically, its class constructor) loads the expected attributes for its components from the Resource Database 620. ActiveDig also loads its own attributes, such as its expected caption and location/size, from the Resource Database as well. In this manner, each component of ActiveDig (as well as ActiveDlg itself) binds itself at initialization to the actual screen object which it represents.

D. Resource Database

The Resource Database 620 stores expected or baseline information about the objects referred to in ATMs. Recall that the ATMs consist of instances of Generic Element Models. The Resource Database is where expected attributes of these GEM instances, such as label text, size, color, and the like, are stored. Thus, the database serves as an inventory of all resource information for an application under test.

By storing object attributes in a database, the need for "hard coding" textual or positional information within ATMs or within test scripts is eliminated. This is particularly advantageous because any textual or positional assumptions about a UI element often change, especially when the application is to be localized (i.e., translated to different languages) for several international versions. In this manner, the need for altering the ATMs or the test scripts every time the text label for a UI element changes is eliminated.

In a preferred embodiment, the Resource Database 620 itself is implemented using object database methodology. For instance, an exemplary resource database may be constructed using a single table defined as follows:

______________________________________ Field Name Type Description ______________________________________ UniqueId String Key field which uniquely identifies an object Label String The text associated with an object Parent String The object's parent name PreferredName String Variable name used by the Model Generator to instantiate this object as DefaultState Integer Default state of an object (checked, grayed, and the like) Dimension String Expected location of this object relative to its parent origin (x1,y1,x2,y2) format Attributes Binary Misc. attributes to be tested Picture Binary Screen shot of the object preferably in a device independent format ______________________________________

The table is keyed by (indexed on) the UniqueId and Label fields. Some fields are multi-purpose, depending on what object is stored in the record. For a menu item record, for instance, the Dimension field stores the order of the item within its popup menu. Further description of the fields is provided hereinbelow, during presentation of the Notepad example.

E. GEM Library

1. Overview

A Generic Element Model or GEM is perhaps best described by way of example. Consider an application program such as a Desktop Publisher (DTP) application. The application may be decomposed into its major UI components: Menus, Dialogs, ToolBar, Client Windows (which include scroll bars and system menus), and a Status Line. Each component can be broken down further. Menus, for instance, can be further reduced to four subtypes: top level pop-up menus, system menus, menu items, and side cars ("fly out" menus). The entire menu tree of any application can be described as instances of these four types of menus. Each one of these four types of menus can then be modeled by four corresponding types of Generic Element Models. Using these four GEMs, ATMs can be built which collectively represent the entire menu tree of the application. Thus given a small number of GEMs, one can build ATMs which represent the user interface of very complex applications.

As described above, a GEM encapsulates the behavior of irreducible user interface elements such as push buttons, checkboxes, listboxes, menu items, and the like. When a GEM is instantiated, it takes two parameters: a reference to its parent and a resource id which uniquely identifies this GEM among its siblings. During construction time, the GEM loads its expected results from the Resource Database using, in an exemplary embodiment, a key consisting of its parent's unique id concatenated with its own id. The GEM binds itself to the actual UI element on the screen which it represents, when requested to do so by its parent. At this point, the GEM can be instructed to run a self test ("SelfTest" method) on itself by simply comparing its expected attributes (loaded from the resource database) against its actual attributes (retrieved from the actual element on the screen which the GEM represents).

2. Detailed construction

In a preferred embodiment, all GEMs are derived from a base class, BaseGEM, which encapsulates the basic functionality offered by a empty or "faceless" UI element, such as processing a single or double mouse clicks, or getting input focus. BaseGEM may be constructed from the following class definition:

______________________________________ //class definition class BaseGem( Pops, ResId ) //Begin constructor ParentPtr = Pops //Reference to parent ParentWHndl = 0 //Parent Window Handle Id = ResId //Unique id for this GEM among its siblings Moniker = 0 //Live link to the actual UI element on screen ObjectType = 0 //Type of this GEM //These are expected values fetched from the resource data base Label = 0 ParentName = 0 DefState = 0 Location = 0 ShortCut = 0 //End constructor function Bind( ) //Load expected values from the resource database Defaults.LoadData( Pops.Id + "," + Id, Label, ParentName, DefState, Location ) Shortcut = GetNmemonic( Label ) //Bind to the actual UI element on screen Moniker = ParentPtr.Moniker.FindItem( Id ) ParentWHndl = ParentPtr.Moniker.WinHandle end function Detach( ) Moniker = 0 end function Select( ) if (Shortcut = = 0) return end KeyIn( ParentWHndl, "{ALT}" + ShortCut ) //KeyIn( ) resides in Test RTL 630 end function Click( ) MouseClick( Moniker.WinHandle, LeftButton ) //MouseClick( ) resides in Test RTL 630 end function DblClick( ) MouseDblClick( Moniker.WinHandle, LeftButton ) //MouseDblClick( ) resides in Test RTL 630 end function RightClick( ) MouseClick( Moniker.WinHandle, RightButton ) //MouseClick( ) resides in Test RTL 630 end function RightDblClick( ) MouseDblClick( Moniker.WinHandle, RightButton ) //MouseDblClick( ) resides in Test RTL 630 end KeyInto( Chars ) KeyIn( Moniker.WinHandle, Chars ) end function IsA( ) return ObjectType end function Capture( ) //Virtual function //Override w/ capture routine specific for object end function HasFocus( ) DlgHndl = FindTheWindow( ParentPtr.Caption ) return Moniker.Focus end function IsEnabled( ) return IsWindowEnabled( Moniker.WinHandle ) end function GEMTest( ) Bist.Start ( XsA( ), Id ) Bist.Verify( "Parent Name", ParentName, ParentPtr.Moniker.Caption ) Bist.Verify( "Location", Location, Moniker.GotLocation( ) ) end end //End class definition of BaseGem ______________________________________

As shown above, BaseGem includes data members of a parent pointer, a parent window handle, a unique identifier, and a "Moniker". The parent pointer or ParentPtr is passed in (to the constructor) at initialization; for a dialog, the dialog passes in "self" (i.e., passes in a pointer to itself). The parent window handle (ParentWHndl) is set to the window handle of the parent pointer's moniker during operation of the bind function. The unique identifier or Id is a way to identify a particular control among its siblings. This may simply be the resource identifier (provided for Windows), or it may be an identifier assigned by the system. Again, this is passed in by the dialog at creation. The Moniker is a handle or live link to the actual UI element on screen. The object type or ObjectType indicates a windows type, such as checkbox, pushbutton, and the like.

The class also includes data members for storing expected values for the object (fetched from the Resource Database): Label, ParentName, DefState, Location, and ShortCut. The values for these are fetched from the Resource Database upon instantiation by calling the Defaults.LoadData method; ShortCut is determined from analyzing the Label (e.g., menu item label of &File indicates a ShortCut keystroke of Alt-F). During self test, the actual values may be compared to these expected values.

Also shown, BaseGem includes class methods or functions. The Bind() function binds the object to its actual screen object. In particular, the method employs a Windows Id to return a unique handle or "Moniker" for the screen object. A corresponding Detach() function is provided for detaching from the screen object; the Moniker is reset (to zero). Other class methods are provided for simulating user events. For instance, the Click(), DblClick(), RightClick(), and RightDblClick() methods perform their respective mouse clicks, by sending the corresponding Windows event message to the target object. Select(), on the other hand, keys in the shortcut (e.g., keyboard accelerator) for the object.

Other types of GEMs may be derived from BaseGEM. A CheckBox class, for example, may be derived. CheckBox inherits all the capabilities of BaseGEM and adds any behavior unique to it. It may be constructed as follows:

______________________________________ class CheckBox( Pops, ResId ) of BaseGEM( Pops, ResId ) ObjectType = "CheckBox" // Get the state of the CheckBox function GetState( ) if ( IsDlgButtonChecked( ParentWHndl, Id ) == 1 ) // Note IsDlgButtonChecked( ) is std Win API (call) 670 return "1" else return "0" end end function SelfTest( ) GEMTest( ) Bist.Verify( "Label", Label, Moniker.Caption ) PrevState = GetState( ) Click( ) Bist.NonEqual( "Changing State", PrevState, GetState( ) ) Click( ) Bist.Done( ) end function Capture( ) dbDictionary Attributes( ) // dbDictionary is test object provided by Test RTL Attributes.AddEntry( "Label", Moniker.Caption ) Attributes.AddEntry( "Location", Moniker.Left + "," + Moniker.Top ) Attributes.AddEntry( "State", GetState( ) ) return Attributes end end ______________________________________

As shown, CheckBox is a derived class and includes state behavior specific for its type, namely "Checked" (state "1") or "Unchecked" (state "0") for a CheckBox. Since the CheckBox knows about its own behavior and attributes, it includes methods for handling these specifics. For instance, CheckBox includes a GetState() method which returns whether the checkbox instance is "checked" or "unchecked." CheckBox adds its own SelfTest() function, which runs the GemTest() function (standard self test) as well as running additional tests specific for a checkbox (i.e., clicking itself and comparing its resulting state with its previous state). Capture() function for capturing the attributes of itself. Specifically for a checkbox, this entails capturing information about its Label (i.e., text displayed with the checkbox) and its state (whether it is "checked" or "unchecked"). These may be stored as entries in a database.

3. GEM Functionality

GEMs will now be summarized in terms of their specific functionality. In a preferred embodiment, GEMs include the following fundamental functionality: Self load, Self test, Binding to screen objects, Attribute capture, Next state, User interaction simulation, and Resource tracking. Each will be described in turn.

(a) Self load

GEMs contain no information about their characteristics (e.g., text labels) in their declaration or definition. Instead, a GEM loads its characteristics from the Resource Database during construction. Each GEM (which is a derived instance of a BaseGem class, presented above) is constructed with a unique resource identifier (ResId). Thus when a GEM is instantiated, the object containing the GEM passes a unique key to the constructor of the GEM. The constructor uses this key to load in the GEM's expected properties from the Resource Database:

Each GEM is constructed only on an as-needed basis.

(b) Self test

All GEMs have a SelfTest member function. SelfTest enables the GEM to test itself by comparing its characteristics to the actual element on the screen that it represents. The basic test includes checking the position, label, parent name, and default state of the element. A CheckBox class, for instance, may include the following SelfTest class method:

______________________________________ function SelfTest( ) GEMTest( ) Bist.Verify( "Label", Label, Moniker.Caption ) PrevState = GetState( ) Click( ) Bist.MonEqual( "Changing State", PrevState, GetState( ) ) Click( ) Bist.Done( ) end ______________________________________

As shown, the Checkbox. SelfTest method begins by calling GEMTest(), which verifies the position and parent for the object. SelfTest then verifies the text label for the checkbox screen object. Finally, the SelfTest method proceeds to test the checkbox with mouse click events. Here, the current state is saved (to PrevState). Then, a click event is executed (by calling the inherited Click() method for the object). At this point, the SelfTest compares the state of the object with the previously-saved state. As a checkbox functions as a toggle, its now-current state (i.e., after receiving the click event) should differ from the previously-saved state. This is confirmed by:

If the states are instead equal (i.e., the checkbox on the screen did not toggle), the test will fail. Upon completion of the test, the checkbox object is restored to its previous state.

In this fashion, all checkboxes of a program may test themselves to assure they are not defective. In addition to testing display attributes and receipt of input events, additional tests can be added to GEMs at any time in the QA cycle. When a test is added to the self test section, all instances of that GEM will automatically include the new test as well.

(c) Binding to screen objects

GEMs can attach themselves to the actual object on the screen that they represent. This ability to bind to screen elements enables any given visual element on the screen to be examined at runtime, greatly increasing the level of observability offered by the test automation system.

(d) Attribute capture

A GEM can construct a container object which captures the attributes of the actual screen element the GEM represents. This container object is an instance of the standard C++ dictionary class with an extra feature of persistence. The attributes of the screen element can be stored in a database (e.g., in the Resource Database) for later reference and comparison.

(e) Next state

GEMs may contain knowledge of what the next state of the program will be after the control the GEM represents is selected. For example, when a menu item is selected, a dialog box might appear. In this case, the menu item and the dialog box are related. As another example, clicking an OK button should cause the button to tell its parent (i.e., the dialog box containing the button) to destroy itself.

Next state relationships among GEMs are established using links within ATMs. Maintaining links among related GEMs in practice usually proves problematic and inefficient however. Therefore the burden of "knowing" the next state can be disabled within GEMs and transferred to a higher level which in this system is the Model Manager.

(f) User interaction simulation

Each GEM can simulate any possible operation that a user would perform an any given UI element, such as a mouse click, to getting focus, pressing a key, and the like. As shown above (in the BaseGEM class definition), for instance, a test script may simulate a click event at an object by invoking that object's Click() method.

(g) Resource tracking

GEMs can monitor resource usage. If there is a sudden drop or leakage in overall system resources, for example, GEMs may log warning messages.

F. Model Manager

The Model Manager 645 functions to monitor the active state of the application under test and load the appropriate model of the application under test at any given state. The Model Manager provides a key benefit. Instead of keeping the entire model of an application under test in memory, only a model corresponding to the active or live portion of the application under test need be loaded and executed. This approach yields improved performance and decreased intrusion over the application being tested.

The approach also frees the script developer from knowing the instance name of every single component within the application. Consider, for example, the commands for bringing up a file open dialog and clicking the OK button:

In the above code, the script writer must know that after executing the first line, the resulting File Open dialog which was activated is referred to as FileOpenDialog. Keeping track of the names of every component within a large application is tedious and error-prone. Moreover, the instance FileOpenDialog must be instantiated even though the script may never use it.

In contrast, the following code performs the same as the one above, yet incurs much lower memory overhead and is easier for the script developer to maintain:

As shown, once the first line is executed, a global variable (ActiveDig) is automatically constructed by the Model Manager 645 by loading and executing the ATM which models this dialog. The Model Manager can also instruct the Model Generator to create a model on-the-fly if one does not already exist.

G. Model Generator

The Model Generator 625 generates ATMs 615 and the Resource Database 620. The Generator may employ resource information embedded in a program, live data on the screen, or a combination of both. In the event that all the necessary information is not available in a program resource, a control script may be constructed to take the application under test into necessary states so that the model generator may generate the appropriate ATM and Resource Database entries.

The Model Generator 625 also provides information about what changes were made between the current build and the previous run. In addition, the Model Generator shields old scripts from breaking (i.e., incurring errors) by using a field in the database call