Creating and Configuring Component-Based Applications Using a Text-Based Descriptive Attribute Grammar

1. Field of the Invention

The present invention relates generally to object-oriented programming, and more particularly, to a method and system for creating and configuring a component-based, object-oriented program using a text-based descriptive attribute grammar.

2. Identification of Copyright

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.

3. Description of the Background Art

Over the last two decades, object-oriented frameworks have become increasingly important in the design and development of software applications. Designing an application with a framework of objects, or "components," promises numerous benefits, including the ability to reuse those components in a variety of different applications. Object-oriented frameworks thus reduce development times and ensure more reliable code.

The promised benefits of object-oriented frameworks, however, have not always been easy to realize in practice. In particular, specifying how different components from a framework are initialized and how they interact with each other often requires trained programmers and expertise in arcane programming languages. Creating and configuring component-based applications is often a complex, time-consuming process. The ability to quickly and easily configure component based applications without requiring extensive programming expertise has thus become a much sought after goal of framework builders and end-users alike.

Concurrently, but independently from the above, the Standard Generalized Markup Language (SGML) was conceived for "marking up" text-based structured documents in a standardized and system-independent manner using only a descriptive attribute grammar and associated document type definition (DTD). Recently, one dialect is of SGML--"HTML"--has become the dominant grammar used on the World Wide Web. Because HTML has a simple, declarative syntax, it is relatively easy for non-programmers to learn. Moreover, HTML's standardized, declarative syntax makes it extremely easy to build tools and editors that manipulate HTML files in user-friendly ways. As a result of the Web's explosive growth and HTML's relative simplicity, HTML and HTML-oriented tools have become extremely popular and widely used by a variety of technical and non-technical end-users.

In many situations, it would be highly desirable to be able to create and configure sophisticated component-based applications using only a text-based, SGML-compliant application description language such as HTML. A text-based description language for component-based applications offers many advantages. For example, the text-based data file may be edited by a simple text editor and requires no special or proprietary editing tools. Similarly, SGML compliance allows such data files to be edited, parsed, verified and analyzed by a wide variety of standardized tools. A great number of third-party applications and tools have already been built for managing and manipulating HTML-based content on the Internet. Moreover, a descriptive syntax is much easier for non-programmers to learn and use than traditional programming language constructs. By basing an application description language on a standardized descriptive syntax already familiar to a wide variety of end-users, the language would be more quickly assimilated by developers and non-programmers than a proprietary syntax or traditional programming languages.

As component-based application design has increased in importance, many attempts have been made to create application frameworks that simplify the design process. However, none of those frameworks are based on the standardized, text-based descriptive grammar described above, nor realize the advantages resulting therefrom. Examples of these attempts include the following:

(1) integrated development environments for object-oriented languages such as Smalltalk.TM., Objective-C, C++ and Java.TM., which offer developers the ability to build and design complex component-based applications in a given programming language;

(2) component-based frameworks and protocols, such as ActiveX/COM.RTM., JavaBeans.TM. and CORBA, which provide standard rules and interfaces for regulating the interaction of heterogeneous components;

(3) multi-purpose "scripting" languages, such as JavaScript.TM., Visual Basic.RTM., Scheme, PERL, and the like, designed to facilitate the coordination and use of existing, external components;

(4) "visual" authoring tools, such as Sun Microsystem's.RTM. Javastudio.TM. and Symantec's.RTM. Visual Cafe which provide "easy-to-use" graphic user-interfaces for creating and configuring component-based applications in a given object-oriented language; and

(5) script-oriented "authoring tools" such as Apple's.RTM. Hypertalk or Macromedia's.RTM. Director.RTM., which provide user-friendly environments for developing multimedia applications using the tools' built-in functionality and single-purpose, proprietary scripting languages.

Although the above systems offer solutions to a variety of development problems, none, however, allow applications to be effectively created and configured using a text-based application description language based on an SGML-compliant syntax. For example, integrated development environments, component-based frameworks, and most visual authoring tools output object-code binary files which are interpreted and executed by the user's operating system or the given language's "virtual machine." Java "applet" builders, for instance, typically generate Java source or object code directly. Such binary code files are not text-based and thus are neither editable nor accessible to text-oriented tools.

Although interpreted scripting languages such as JavaScript, Hypertalk, and Lingo can be edited in a generic text editor, their syntax is procedural rather than declarative. As a result, the semantic content of these script files can only be productively inspected and analyzed by the specific tools and editors dedicated to that particular language. In short, while there are a wealth of powerful component-oriented application frameworks and authoring tools, none of them are based on a standardized, text-based descriptive grammar or realize the advantages which accompany being based on such a grammar.

Analogously, numerous efforts have been made to extend the ability of SGML and HTML to specify more sophisticated formatting instructions. For example, HTML's pending versions include numerous processing and formatting instructions to assist developers in creating "multimedia" documents that provide basic interactivity through external "scripting" languages that control "black box" components such as Java applets and/or ActiveX components. Originally, HTML was intended as a purely descriptive syntax that would be transparently portable and easy to understand by both technical and non-technical publishers. The newly evolving hybrid-HTML attempts to expand its functionality by adding simple scripting capabilities to HTML's original descriptive tags and of creating an "open" environment for the integration of third-party components.

However, the hybrid-HTML format has a number of limitations and disadvantages. For example, since hybrid-HTML mixes procedural scripts with descriptive markup tags, a scripting code is required in order to control the behavior of objects within the HTML document. Thus, HTML publishers must understand and master procedural programming in order to take advantage of HTML's newest features. Moreover, critical semantic information is now embedded in procedural code. These procedural "scripts" cannot be inspected or parsed by generic, SGML tools and often contain instructions proprietary to a particular browser vendor.

Another disadvantage is that, in the hybrid-HTML format, each component "plug-in" is a separate and independent "black box" with its own, idiosyncratic interface to the browser's scripting language. Even components developed in a common language, such as Java-based applets, are controlled through the particular interface exposed by a specific applet. Thus, it is difficult, particularly for non-programmers, to integrate a variety of components into a single component-based application.

Yet another problem is that, although HTML tags can be used to declare attributes associated with standard browser components and user-interface controls, there is little or no coordination between the structure of the markup tags and the underlying data structures employed by the browser application itself. For example, HTML elements do not "inherit" properties from "superclass" elements. Thus, hybrid-HTML does not provide a sufficiently sophisticated architecture to integrate large-scale component-based applications.

Finally, hybrid-HTML is limited by the rigidity of HTML itself. By design, HTML only allows for a fixed set of element and attribute tags. There is no provision for developers to add new functionality through application-specific element tags or attributes. Nor is there any ability to map new element tags to new components or to extend the functionality of the browser application through code-based subclassing. Fundamentally, HTML was designed and is still used as a document markup language that is "processed" by an autonomous application: the browser. The intent and design of HTML was to provide a standardized format for marking up "multimedia" documents. All sophisticated application processing is delegated to external resources, i.e. scripting languages, embedded "applets", server-side applications, and the like. Consequently, HTML-based browsers were not designed to actually create and configure a component-based application through descriptive tags alone.

The Extensible Markup Language (XML) proposal recently submitted to the World Wide Web Consortium (W3C) attempts to loosen some of these restrictions on HTML in order to create a simple, declarative syntax that can be used for a wider variety of data-oriented files. In particular, XML allows developers to add new element tags and attributes as long as these comply with the XML specification. Current initiatives using the proposed XML standard include Microsoft's "Channel Definition Format", Netscape's "Meta-Content Framework", and Peter Murray-Rust's "Chemical Definition Language."

While attempts have been made to integrate XML with such object-oriented applications and languages such as Java, C++, and Tc1, existing systems offer only a rudimentary parameterization of the resulting components and require external scripting or programming code in order to provide more sophisticated interactivity. XML has the flexibility and simplicity to provide the syntactic framework for configuring component-based applications. No existing method or system, however, has effectively tapped XML's potential.

What is needed, then, is a method and system for creating and fully configuring component-based applications using a text-based XML-compliant grammar. Moreover, what is needed is a system and method for modeling the component structure of an application through a declarative syntax, such that the component model instantiated by the application mirrors the element tag structure of the original application description file. Additionally, what is needed is a means for specifying the user-oriented interactive behavior of an application in purely descriptive form without resorting to procedural or functional language. Moreover, what is needed is a means for expressing how a component should respond to events within the limits of the XML syntax. Similarly, what is needed is a system and method that permits custom event messages to be created and processed at runtime by components without any special underlying support code on the part of the component. In addition, what is needed is means for specifying complex inter-component behaviors within the limits of a descriptive grammar and without resorting to procedural or functional language. Moreover, what is needed is a means for dynamically specifying the component types used by the application such that new component classes and functionality may be added to the system at runtime. Additionally, what is needed is method for referencing any given component in the application through a scoped, text-based name mechanism. What is also needed a system and method for dynamically controlling and interacting with any of the components of the given application through a single, top-level method call that takes only character arrays as its parameters.

The present invention overcomes the various limitations of the prior art by providing a method and system for creating and configuring a component-based application through simple, XML-compliant, text-based document. In accordance with a preferred embodiment of the present invention, a parse tree is created from an application description file. Thereafter, the parse tree is transformed into a plurality of components corresponding to instances of classes in an application framework. The components are then initialized and further processed to launch the component-based application.

Additionally, in accordance with a preferred embodiment of the present invention, an system for creating and configuring a component-based application includes a parser, a plurality of element processors, an element mapper, an attribute mapper, and a parameter converter.

FIG. 1 is a block diagram of a system for creating and configuring a component-based application in accordance with a preferred embodiment of the present invention;

FIG. 2 is a high-level dataflow diagram of a system in accordance with a preferred embodiment of the present invention;

FIG. 3A is a dataflow diagram of parsing an application description file in accordance with a preferred embodiment of the present invention;

FIG. 3B is a dataflow diagram of tranforming elements into component instances in accordance with a preferred embodiment of the present invention;

FIG. 3C is a dataflow diagram of processing component instances to launch a component-based application in accordance with a preferred embodiment of the present invention;

FIG. 4A is a high-level flowchart for a method of creating and configuring a component-based application in accordance with a preferred embodiment of the present invention;

FIG. 4B is a flowchart for a method of transforming the elements of the parse tree into uninitialized components in accordance with a preferred embodiment of the present invention;

FIG. 4C is a flowchart for a method of mapping element attributes 310 to component properties 320 in accordance with a preferred embodiment of the present invention;

FIG. 4D is a flowchart for a method of processing the components to launch the component-based application 214 in accordance with a preferred embodiment of the present invention; and

FIG. 5 is an illustration of an attribute mapper in accordance with a preferred embodiment of the present invention.

A preferred embodiment of the present invention is now described with reference to the Figures, where like reference numbers indicate identical or functionally similar elements. Also in the Figures, the left most digit of each reference number corresponds to the Figure in which the reference number is first used.

System Architecture

Referring now to FIG. 1, there is shown a block diagram of a system 100 for creating and configuring a component-based application in accordance with a preferred embodiment of the present invention. In one embodiment, the system 100 includes a computer 102, having a processor 104, addressable memory 106, output devices 108, input devices 110, a network interface 112, and a storage device 114.

The computer 102 is preferably a general purposes computer, such as an IBM PC or compatible machine. Thus, the processor 104 may include an Intel.RTM. Pentium.RTM. II or similar microprocessor. The memory 106 is implemented by a standard memory device, such as a random access memory (RAM). The output device 108 includes a CRT display, printer, or other suitable output device. The input device 110 includes a standard keyboard, mouse, or other similar input device. The network interface 112 is implemented as a conventional Ethernet adapter or other suitable interface. The storage device 114 includes a hard disk drive, CD-ROM, magneto-optical drive, DVD, flash memory, or other mass storage device. It should be recognized, however, that a number of alternative physical devices and implementations could be readily used without departing from the spirit of the invention.

In one embodiment, the memory 106 is used to store a number of software objects or modules, including a parser 116, a plurality of element processors 118, an element mapper 122, an attribute mapper 124, and a parameter converter 126. Throughout this discussion, the foregoing modules are assumed to be separate functional units, but those skilled in the art will recognize that the functionality of various units may be combined and even integrated into a single software application or device.

The processor 104 executes the parser 116, the element processors 118, the element mapper 122, the attribute mapper 124, and the parameter converter 126, including all of the operations thereof. In addition, the processor 104 executes a general purpose operating system 128, such as Windows 95, available from Microsoft Corporation of Redmond, Wash.

In a preferred embodiment, the object modules in the memory 106 are implemented using an object-oriented programming language such as Java, available from Sun Microsystems of Mountain View, Calif. The Java programming language was designed to allow component-based applications to be developed that are safe, platform-independent, robust, and dynamically instantiable. Consequently, Java provides a full-fledged, object-oriented programming language with single inheritance, class- and package-level encapsulation, strong data typing, garbage collection, dynamic class loading and run-time binding. One skilled in the art, however, will recognize that other languages providing similar functionality may also be used without departing from the spirit of the invention.

Additionally, an application framework 130 is provided, which includes a plurality of classes 132 in Java or another suitable object-oriented language. In one embodiment, the application framework is "JavaBeans," which is a specification for a standardized Java-based application framework designed by Sun Microsystems. The JavaBeans specification is a standard set of Application Program Interfaces (API's), and facilitates interoperability of "Bean" components through a standardized set of "metadata" queries. However, one skilled in the art will recognize that other frameworks could be used within the scope of the present invention.

System Dataflow

Referring now to FIG. 2, there is shown a high-level dataflow diagram of the system 100 in accordance with a preferred embodiment of the present invention. The parser 116 receives an application description file (ADF) 202, which, in one embodiment, resides in the storage device 114. Alternatively, the ADF 202 is generated externally by a software object, a server, or the like. Preferably, the ADF 202 is an XML-compliant text document that defines a component-based application using a descriptive attribute grammar. Two embodiments of an application description language (ADL) for the ADF 202 are detailed below.

The parser 116 parses the ADF 202 to generate an XML parse tree 204. Preferably, the parse tree 204 is stored in the memory 106 as a set of interconnected objects, although a variety of other configurations are possible within the scope of the invention. In one embodiment, the parser 116 is a conventional XML parser available from Microsoft and other sources. Alternatively, a custom-designed XML parser is used. The ADF 202 and the parse tree 204 are described in greater detail below with respect to FIG. 3A.

Thereafter, the parse tree 204 is transformed by the element processors 118 into a plurality of uninitialized components 212. In one embodiment, the components 212 are "bean" objects in the JavaBeans application framework 130. The process of mapping elements 306 to components 212 is discussed below in greater detail with reference to FIG. 3B.

After the components 212 are created, the element processors 118 process the uninitialized components 212 to launch the component-based application 214. As will be described in greater detail below with reference to FIGS. 3C and 4D, this is accomplished in one embodiment by initializing each of the components 212 and adding each child component 212 to its parent component 212.

Referring now to FIGS. 3A-C, there is shown a more detailed representation of the dataflow of the system 100 in accordance with a preferred embodiment of the present invention. FIG. 3A illustrates the process of parsing the ADF 202 into the parse tree 204. An exemplary ADF 202 is provided, showing one embodiment of a descriptive attribute grammar. In a preferred embodiment, the parser 116 uses recursive techniques to generate the parse tree 204, which is preferably a set of interconnected objects in the memory 106.

In one embodiment, the leaves or nodes of the parse tree 204 are of four basic types, corresponding to the four types of data in the ADF 202:

comments 302 (delimited in the ADF 202 by <!-- and -- >),

processing instructions 304 (delimited by <? and ?>).

elements 306 (all other constructs delimited by < and >), and

raw data 308.

The comments 302 are used for internal documentation of the ADF 202, and do not affect the operation of the system 100. Similarly, the processing instructions 304 are used to communicate with external applications and are not crucial to understanding the present invention.

The elements 306, however, comprise the primary structures of the ADF 202, and, as described below, are associated with components 212 in the application framework 130. Each element 306 has a name or "tag" 314 by which the element 306 is defined in the ADF 202. Only elements 306 can have children, which can be any of the four types discussed above. Additionally, the elements 306 may contain attributes 310, which correspond to properties of a component 212. The attributes 310 and their values are parsed by the parser 116 and placed in an attribute list 312 within the element 306.

FIG. 3B illustrates process of transforming the parse tree 204 into a plurality of components 212. In one embodiment, the set of components 212 is represented by an object graph 318, which preserves the hierarchical structure of the parse tree 204. The element processors 118 transform the parse tree 204 into uninitialized components 212, the process of which is described below.

By recursive descent of the parse tree 204, an element processor 118 is created for each of the elements 306. Each element processor 118 knows the meaning of its corresponding element 306, and understands how to process the element's attributes 310 and child elements 306, as well as the proper syntax for each. In one embodiment, a different element processor 118 is provided for each type of element 306. Alternatively, only a few "core" element processors 118 are provided, and the default component class 132 is overridden with a particular target component class 132. A variety of element types are possible, and examples are provided hereafter with respect to various embodiments of an application description language (ADL).

The element mapper 122 is used to select the appropriate element processor 118 for a given element 306. In one embodiment, the element mapper 122 is an object that maps the element's name or tag 314 to the corresponding ("target") class 132. Thereafter, the element processor 118 instantiates an object or component 212 of the target class 132. Additionally, an object called an attribute mapper 124 is used to map attributes 310 associated with the element 306 to properties 320 of the component 212.

FIG. 3C illustrates the processing of the components 212 in the object graph 318 to launch the component-based application 214. As will be described in greater detail below with reference to FIG. 4D, this is accomplished in one embodiment by initializing each of the components 212 and adding each child component 212 to its parent component 212.

Specifications for an Application Description Language

As noted above, the ADF 202 is preferably an Extensible Markup Language (XML) compliant text file that defines a component-based application using a descriptive attribute grammar. In a preferred embodiment, the ADF 202 conforms to an application description language (ADL) as described below. One skilled in the art will recognize that a variety of ADLs may be used within the scope of the present invention. Thus, the syntax of the ADL may vary without departing from the spirit of the invention.

Definition of an ADL "Document"

For purposes of following description, documents are data objects which contain text data made up of "characters" as defined by the XML specification. The physical storage characteristics of a document is irrelevant to its processing. As with any XML document, ADL documents have a logical structure, which includes element 306 declarations, attributes 310, comments 302 and character data 308 whose structure is specified through "markup" tags 314 in compliance with the XML specification.

Two exemplary ADLs are presented below. The first is called "BML." The second conforms to both "well-formed" and "valid" XML, and is called hereafter "ADML."

"Well-formed" ADL Documents

The definition of a well-formed ADL document is similar to that for a well-formed XML document. A well-formed ADL document 1) contains one or more elements 306; 2) has one root document element 306 which encompasses all other elements 306 in the given document; and 3) is structured such that all elements 306 other than the root documents are entirely contained within a single parent element 306.

In the currently preferred embodiment, the root document element is delimited through a pair of <BODY> and </BODY> tags. Well-formed BML documents do not require document type declarations (DTDs) or document version information. The following is a rudimentary but well-formed BML document:

Character Data and Markup

Like standard XML documents, BML documents consist of markup data and content character data within specific elements' closures. Valid markup structures allowed by BML include start-tags, end-tags, empty elements, character references, comments and CDATA sections. The current preferred embodiment of BML does not support processing instructions 304, document type declarations or entities though these could be easily added if appropriate in other embodiments. Comments 302 in BML are marked and processed identically to XML documents and are, therefore, ignored by parser 116. As with XML, BML comments 302 may not appear within declarations or tags.

Element and Attribute Declarations

Like XML, BML documents contain one or more elements 306. Elements 306 are declared and delimited by a start-tag and end-tag. Like XML, the name 314 in an element's end-tag matches the name in its start-tag. Empty elements 306 in BML may be specified using the standard XML "empty element" tag. Unlike XML, BML does not require the special "empty element" tag. The following are allowable as valid empty BML elements 306:

As with XML, the first character data inside an BML start tag is a valid element name or tag 314 which defines the element's type. Each element may have one or more attributes 310 which are declared in precisely the same name-value syntax as standard XML attributes. For example, in the "foo" empty element tag 314 above, the attribute 310 name "Att1" is set to the value "Value1," and the attribute 310 "Att2" set to the value "Value2".

BML attribute 310 names correspond to the constraints of XML attribute names. Unlike XML, BML does not require quote characters for attribute 310 values with no embedded white space characters. For example, the following are valid BML elements 306:

Unlike XML, BML does not disallow duplicate attributes 310. In other words, attributes 310 may appear more than once in the same start-tag. For example, the following is a valid BML element:

Only certain BML elements 306 may have child elements 306. Whether or not a given element type can contain children is not a property of the BML syntax itself but is determined solely by the properties of the underlying component class 132 to which the element name 314 is mapped. In general, only component classes 132 that implement a specific "container" interface may have children element tags 314 as will become apparent hereafter.

The following is valid as an element tag 314 with children if "Fizz" is mapped to a "container" component 212 interface:

All BML attribute 310 values are input and parsed as string data. The parser 116, however, allows these string-based values to be parsed into primitive and object data types according to the parameter specifications of the attribute's corresponding mutator. Allowable basic parameter types include: string, double, boolean, date, integer, and arrays. Parameter types may be extended at runtime by individual components and element processors. In addition, BML allows a special string-based parameter type called a "named object reference," which will be discussed in greater detail hereafter.

As with XML, BML element 306 and attribute 310 names are case insensitive. Attribute 310 values, however, may be case sensitive depending on the parameter type specified by the underlying component 212 method. Before attribute 310 values are passed to the underlying component 212, trailing and leading white space is removed unless explicitly declared within a pair of quotations.

Event Declarations

One object of the present invention is to provide a mechanism for defining complex component-based interactions within the limits of a descriptive attribute grammar. Thus, BML provides a special attribute 310 declaration, OnEvent, for specifying both user and application-generated "events". The OnEvent attribute 310 is used to define how a component 212 should "react" to specified event types as well as how event "messages" may be posted among the components 212 themselves.

The syntax for the OnEvent attribute 310 declaration requires a single string value with comma-separated fields as follows:

In one embodiment, the Event Tag field is required and may correspond to either a component-specific event tag (e.g. "MOUSEUP", "MOUSEENTER", etc) or to a custom event tag used among components in the application (e.g. "IMPORTANTEVENT", "aCustomEventTag", etc). The Action field is also required and should correspond to a valid, publicly accessible method on the target component 212. Preferably, Event tags and Action fields are both case insensitive. Similarly, the Target field is required and should be a scoped name reference to a component 212 accessible from the base component 212. The Parameter field is optional. Any text data after the Target field is treated as a single string and part of the Parameter field. The raw parameter string is not converted to any appropriate data type until the associated action message is executed.

Validation of the BML File

Unlike typical XML or SGML files, a BML file cannot be validated by reference to a pre-existing DTD. Instead, the validity of an BML file can only be determined in a specific application context and in reference to the component object model specified by the BML file. In a valid BML file, the resulting object model should have the following characteristics:

(1) each element 306 declaration in the BML file should be capable of being mapped to a corresponding application component class 132;

(2) each element 306 with children should result in a corresponding "container" component 212 with analogous children components 212;

(3) each attribute 310 declaration should be mapped to a corresponding property 320 value in the corresponding component 212 unless special processing for that attribute name has been specified; and

(4) all "Event Declarations" should be mapped to equivalent event-based "triggers" in the underlying components 212.

While transforming the elements of parse tree 204 into components 212, the element processors 118 are able to detect whether the four requirements above are met. Specifically, it can be determined whether a corresponding component class 132 exists for the specified element 306 and whether the attribute 310 declarations have a corresponding property 320 value. Thus, in a preferred embodiment, the element processors 118, in combination with the parser 116, may validate an BML file.

ADML has a more restricted syntax than BML, which allows a document to be validated against a document type definition (DTD). The main difference between ADML and BML is that the names of tags 314 and associated attributes 310 in ADML are fixed and cannot be added to. Thus, by allowing only certain tags 314 and certain attributes 310 for those tags, a DTD may be defined, and ADML is XML in the strictest sense, being both "well formed" and "validating" XML. A DTD for ADML is attached as Appendix A and is described in detail below.

The ADML DTD

A DTD is a special document that describes the syntax of a markup language. The XML specification requires that the processor of a dialect of XML have a DTD and enforce the syntax in order to be considered "validating" XML. ADML's syntax is enforced by the ADML runtime. Appropriate error messages are generated when a ADML fails to conform to the DTD's rules. To fully understand the document type definition of ADML in Appendix A, the following should be noted:

Types

In BML, the tag 314 name is used to identify the type of the component or "bean" 212. For instance, the <BUTTON> tag 314 is used to declare a button. Creation of new tags 314 is not permitted in ADML. Thus, the BML expression <BUTTON> is replaced by the ADML expression <COMPONENT Type="BUTTON">. COMPONENT is one standard tags 314 for declaring a bean 212, as described below. Although there are several different tags 314 in ADML for creating a bean 212, all are told what kind of bean 212 to create with a Type attribute 310, whose value should be the ID of a TYPE tag declared in the HEAD of the document. The standard types that are provided with the runtime can be found in a StandardHeader.ent file. These standard types do not have to be used, and can be replaced with custom designed types, but care must be taken that there is a TYPE for each bean's 212 Type attribute 310 or default type.

If the Type attribute 310 is missing in a bean tag 314, the runtime will look for a TYPE whose ID is the name of the bean's tag 314. For example, if a bean is declared using a CONTAINER tag 314, but no Type is specified, the runtime will look for a TYPE in the HEAD section called "CONTAINER".

An example of the differences in syntax between BML and ADML is illustrated below. For instance, a BML document might take the form:

whereas the same document in ADML would have the form:

Inheriting Types

In some situations, an existing type may be almost the desired type, but not quite. A user may be satisfied with the BUTTON type in the StandardHeader.ent file, but want to change the Tag, Class, Value, or Processor attributes. For example, a BUTTON type can normally only be used in a COMPONENT tag 314, because the Tag attribute 310 of the BUTTON type says "COMPONENT". Recognizing that the StandardHeader.ent file specifies a Swing JButton as the Class for the BUTTON type, if the user wants to take advantage of the fact that Swing buttons are actually containers, the user could create a type identical to BUTTON but usable in a CONTAINER tag 314. The user would create a new tag 314 derived from BUTTON and name it something new, such as:

INIT Macros

An INIT section can be associated with a TYPE. All beans 212 declared with that type will have all of the contents of that INIT section executed when the bean 214 is created. This can be used to set up default bean property 320 values for a type. For example, the BUTTON type in the StandardHeader.ent file might have an INIT section that sets the opaque property on the JButton to "true," because the user did not want all the buttons to be transparent. Thus, any component 212 whose type is BUTTON will be opaque. When a type is inherited, it inherits all of the inits from above, but the derived type will execute its inits last, allowing it to override the inits of a parent type. The bean's own INIT section is likewise executed after all inits inherited from the TYPE.

EVENT Macros

Similar to INIT macros, event-handling macros can be specified to create default behavior for an object 212 of the given type. For example, one could create a button type that responds to "mouseEntered" and "mouseExited" event tags to create buttons that become brighter when the mouse passes over them. Like INIT macros, derived types' event handlers execute after parent type's event handlers, and event handlers attached to the bean tag are executed last of all.

START and STOP Sections

There may also be default START and STOP sections for a type. These are used to specify start and stop behavior for all beans 212 of that type. For example, the animation in an Animator bean 212 could be started and stopped so that whenever the bean 212 becomes invisible, the animation automatically stops, thus saving CPU time.

Tags for Bean Creation and Declaration

There are several tags 314 which create and declare bean components 212 in the ADML language. While, for the most part, they share a similar syntax, each tag has different semantics. In one embodiment, the tags include BEAN, COMPONENT, CONTAINER, BODY, TEXT, STYLE, P, LAYOUT, and POSITION.

Most of the foregoing tags share the following common features:

Using the <LAYOUT> Tag

In general, a LayoutManager is a helper object that is associated with a Container that tells the container how to arrange its children within its area. Because they are objects, ADML allows them to be created and initialized using the same constructs as other beans 212. They are declared using the LAYOUT tag. The Type attribute 310 is used to identify which layout manager is being created. Like other beans 212, they can be configured using VALUE and INIT tags 314. If no LAYOUT is specified for a CONTAINER, the Container's default layout manager will be used (this will usually be FlowLayout). If LAYOUT is used, a Type should be specified, because it has no default type. Layout Types, like any other types, are specified in the HEAD section of the document. Some examples of the foregoing include:

To use a declared layout type, its ID should be specified in the LAYOUT tag's Type attribute. For example, to use ADML's layout manager, the RLayout type declared in StandardHeader.ent should be used thus:

Using the <POSITION> Tag

A POSITION specifies the information, called "constraints", that constraints-based LayoutManagers need for a child of the container in order to know how the child is to be laid out. For example, ADML's built-in layout manager calls for one of three kinds of objects. The most useful one is used to specify relative positioning. Its Type, declared in StandardHeader.ent, is "RPosition". BorderLayout, on the other hand, requires a String.

The RPosition type currently only supports the relative positioning portion of BML's built-in layout manager. It implements all of the BML layout attributes except x, y, width and height. RAbsolute and RPreferred are two other less common constraints types accepted by RLayout.

If no POSITION tag is specified for a child, it is up to the LayoutManager to determine how to deal with the child. For example, if the RLayout type is used, and no POSITION is given for a child, the child will be stretched to fill the entire container.

Because some LayoutManagers use Strings as constraints, String has been set up as the default type for a POSITION (if StandardHeader.ent is used). Therefore, if a POSITION tag with no Type is specified, the parent's LAYOUT should accept Strings for its constraints.

Bean Creation Expressions

Like Sun's beanbox, BML is limited to using the default constructor for a bean 212. ADML, on the other hand, allows a user to specify an expression for creating the bean 212. This expression can be a call to a non-default constructor (i.e. a constructor with arguments), a "factory" method that produces objects, or a CONSTANT value such as a String or a Color, an ARRAY, or the like. The following example uses a constructor for a JButton that takes a single Icon argument:

The Method="new" tells the runtime to call a constructor. A Bean creation expression can be specified as part of a TYPE. The following example creates a type called SpecialButton. Components 212 of that type will be buttons that contain the text, "Click Me!". It calls the constructor for JButton that takes a single String argument.

Expression Handling Tags in BML and ADML

Table 1 provides a summary of expression handling tags in both BML and ADML.

Text Constructs in ADML

Part of the ADML design is to allow text to be embedded within beans 212 and beans 212 within text and to allow both to interact with each other through ADML events. The text features center around three tags defined in the ADML DTD: TEXT, P, and STYLE.

The TEXT tag is a tag 314 from the <COMPONENT>, <CONTAINNER> <BEAN> family of components 212. It resembles them in most respects in that it can generate events, be laid out within its parent with POSITION constraints, etc. The ADML event types available for the TEXT component 212 are the names of all the methods of all the listeners capable of being registered with a JTextComponent. A TEXT tag should not be empty because it contains at least one paragraph.

A Paragraph is defined by the P tag 314. The paragraph can contain text, beans 212, and STYLE tags 314 that in turn contain more text and beans 212. It has certain bean 212 properties that correspond to paragraph attributes. In one embodiment, paragraphs do not have an INIT section, so these properties 320 should be set inside of the INIT section for the paragraph's type. Examples of text properties 320 are shown below in Table 2.

Additionally, any properties from STYLE are available to paragraphs. For example, to make a paragraph that contains red, centered text, with 1/2 inch left and right indents, a paragraph type is initially made that declares these properties as shown below:

Thereafter, the paragraph of that type is declared:

Styles

A Style is defined by the STYLE tag. The paragraph can contain text and beans 212. It has certain bean properties 320 that correspond to text attributes 310. In one embodiment, since styles cannot have an INIT section, these properties should be set inside of the INIT section for the style's type. An exemplary list of the style properties 320 is provided below in Table 3.

For example, to make a style for large, green text, a style type that declares these properties 320 is shown below:

Thereafter, a style may be declared of that type:

Hypertext Events

There are three special event types defined for styles related to hypertext. These event types include:

hyperlinkActivated--called when text within that style block is clicked,

hyperlinkEntered--called when mouse enters that style block, and

hyperlinkexited--called when mouse exits that style block.

Such event types can be used to create hypertext links. To create a hypertext link that is normally blue and underlined, but turns red when the mouse is over it, a style is first declared that is underlined and blue:

Thereafter, a text block is created with a style of that type, along with some EVENT tags:

Although the foregoing provides a number of specific embodiments of an application description language (ADL), one skilled in the art will recognize that a variety of ADLs may be used within the scope of the present invention. Thus, the specific syntax of the ADL is not crucial and may vary without departing from the spirit of the invention.

Method of Operation

Referring now to FIG. 4A there is shown a high-level flow diagram for a method of creating and configuring a component-based application in accordance with a preferred embodiment of the present invention. The method begins by creating 402 the XML parse tree 204. Preferably, a conventional parser 116 is used to parse the ADF 202, as illustrated in FIG. 3A.

Thereafter, the method continues by transforming 404 the parse tree 204 into uninitialized components 212. This step will be described in greater detail below with reference to FIG. 4B. Nevertheless, as noted above, each element 306 in the parse tree 204 is mapped to a corresponding target class 132, after which an uninitialized component 212 is instantiated. In addition, each attribute 310 of the element 306 is mapped to a corresponding property 320 of the component 212 by the attribute mapper 124.

The method continues by processing 406 the components 212 to launch the component-based application 214. This step will be described in greater detail with respect to FIG. 4D. Nevertheless, in one embodiment, this is accomplished by initializing each of the components 212 and adding each child component 212 to its parent component 212.

Referring now to FIG. 4B, there is shown a method of transforming the elements 204 of the parse tree 204 into uninitialized components 212 in accordance with a preferred embodiment of the present invention. As noted above, this is accomplished by recursively descending the parse tree 204 and creating an element processor 118 for each elements 306. Preferably, the parse tree 204 is traversed using a post-order recursive traversal algorithm, such that each child component 212 is created before its parent component 212. In one embodiment, the nodes of the parse tree 204 other than elements 306 are either ignored (in the case of comments 302 and processing instructions 304) or are processed appropriately (in the case of data 308).

Starting with the root element 306 of the parse tree 204, the method begins by mapping 408 the element 306 to a target class 132. Preferably, an element mapper 112 is provided, which is an object used to map the string tag 314 of an element 314 with the corresponding element processor 118 and Java component class 132. In one embodiment, the element mapper 122 is implemented by an ElementTagManagerIntf instance, the public interface of which is provided below in Java pseudo-code:

The mapping information is preferably provided to the ElementTagManagerIntf instance 122 at runtime through the call, addProcessorTag(String theElementTag,String theProcessorClassName). This call dynamically associates the given element string tag 314 with the specific element processor class 132 name. In one embodiment, a single global instance 122 of ElementTagManagerIntf is used. However, one skilled in the art will recognize that several such ElementTagManagerIntf instances 122 could be used, depending on the requirements of a particular application 214.

Following step 408, a determination 410 is made whether the element 306 was successfully mapped to a class 132. If not, an error message is generated 412, and the method is complete. Otherwise, the method continues by creating 414 the element processor 118 for the particular class 132. In one embodiment, the element processor 118 implements the Java interface, ElementProcessorIntf, an example of which provided below in Java pseudo-code:

In one embodiment, a different element processor 118 is used for each distinct element tag 314. Alternatively, only a few "core" element processors 118 are used. In that case, instead of registering separate element processors 118 for every element tag 314, only the basic core element tags 314 have distinct element processors 118. Other element tags 314 simply reuse one of the core processors 118, but override the default component class 132 with their particular target class 132. In one embodiment, the target class 132 for a given element tag 314 can be specified in one of three ways:

1. Each element processor 118 has a "default" target class 132 associated with it. Unless otherwise overridden, an element tag 314 which makes use of this element processor 118 will create an instance of the given target class 132.

2. The target class 132 associated with a given element tag 314 may be "globally" overridden through the ElementTagManagerIntf method, addComponentClassTag(String theElementTag, String theDefaultElementTag, String theJavaClassName). This call specifies that the element processor 118 already registered with theDefaultElementTag should also be used with theElementTag but that instead of creating an instance of the default target class 132 for that processor 118, an instance is of theJavaClassName should be created instead.

3. In one embodiment, it is also possible to locally "override" a processor's target class 132 through a special markup attribute 310, Class. The value of the Class attribute 310 is expected to be the fully qualified name of a target class 132 which can be created and used by the given element processor 118 for a specific element 306 instance. In one embodiment built upon the Java programming language, the element processor 118 expects the value of the Class attribute 310 to be the fully qualified name of an accessible Java class 132 (what constitutes a valid, accessible Java class 132 will depend on the end-user's system and network environment as well as the other resources provided by the application 214). When the element processor 118 finds the Class attribute 310 declaration in a particular element 306 declaration, it is instructed to create and use the specified Java class 132 instead of the Java class 132 normally registered for that element 306. For example, in one embodiment, the following element 306 declaration instantiates an instance of the class 132, "my.SpecialClass", rather than the default class 132 specified for the given element processor 118:

For the root element 306 of the parse tree 204, a top-level element processor 118 is preferably created. As explained in greater detail below, the top-level processor 118 is used to create a root component 212. In one embodiment, the root component 212 manages the "event loop" of application 214 and thus receives update and user event messages generated by the platform-specific operating system 128.

One skilled in the art will recognize that the above-mentioned process is only one embodiment of the present invention, other variations are possible and may be advantageous under certain circumstances. For example, instead of using intermediary element processors 118, a variation could create Java components 212 directly and expect each component 212 to implement its own "processing" logic.

After the element processor 118 is created, it instantiates 416 an uninitialized component 212, which is preferably a default instance of the associated Java class 132. Alternatively, the component 212 may be created in a non-default way, if so specified in special attributes 310 or child elements 306 dedicated to that purpose.

After the component 212 has been instantiated, the element processor 118 maps 418 each attribute 310 in the attribute list 312 to a corresponding property 320 in the component 212. In one embodiment, the element processor 118 uses an attribute mapper 124 for this purpose. A detailed explanation of this process is provided below.

Preferably, each element processor 118 is free to treat attribute 310 declarations in its own manner. A given element processor 118, for example, may choose to ignore certain attributes 310 or use specified attribute values in non-conventional ways. However, the default, "core" element processors 118 used in one embodiment of the invention treat the attribute lists 310 in a simple, canonical manner as defined hereafter.

The element processors 118 expect the target component 212 to provide an exact description of its publicly accessible methods and properties 320 such that string-based attribute keys can be mapped to executable "mutator" and "accessor" methods on the target component 212. In one embodiment, this information is obtained through an attribute mapper 124, such as the BeanInfo instance as defined by Sun's JavaBeans specifications. The BeanInfo object 124 is used to map attribute 310 names to corresponding properties 320 in the underlying component 212. If the present invention is implemented in a different object-oriented languages such as Smalltalk or C++, analogous metadata frameworks could be used.

As defined by the JavaBeans specification, each BeanInfo instance 124 may describe all its publicly accessible properties through an array of PropertyDescriptor instances 502, which are illustrated in FIG. 5. Each PropertyDescriptor instance 502 supplies a method, getDisplayName(), which returns the localized, display name for the given property 320, as well as a set of accessor and mutator methods (referred to herein as "write methods" 504) to be used in obtaining and setting the value of the property 320 in the component 212.

Referring now to FIG. 4C, there is shown a flow diagram for a method of mapping element attributes 310 to component properties 320. The method begins by selecting 420 the next attribute 310 in the attribute list 312. Thereafter, the element processor 118 searches 422 the given BeanInfo instance 124 for a PropertyDescriptor instance 502 whose display name matches the given attribute 310 name on a case-insensitive basis. A determination 424 is then made whether a match was found. If a match is found, the write method 504 of the property 320 is obtained 428 through a getWriteMethod() call, and the method continues at step 430. If a match is not found, or if the property is hidden (i.e. PropertyDescriptor's isHidden() message returns a "true" value), an exception is generated 426 and the attribute 310 is ignored.

If a valid write method 504 is obtained, the element processor 118 continues by checking 430 the expected parameters of the write method 504 against the string value of the attribute 310. A determination 432 then is made whether the attribute 230 value needs to be converted. If conversion is necessary, the method proceeds to step 434; otherwise, the method continues with step 436.

In one embodiment, the element processors 118 use a parameter converter 126 for converting string values into a variety of basic data types, such as integers, doubles, dates, booleans, and the like. Methods for such type conversions are well known to those skilled in the art. Additionally, the parameter converter 126 may convert string-based component references into the corresponding component 212 instances. For example, in one embodiment, a set of canonical object names may be used anywhere within a scoped name path. These canonical object names include:

A number of other naming conventions may be provided within the scope of the present invention.

If conversion is necessary, the attribute 320 string values are converted 434 by means of the parameter converter 126. A determination 438 is then made whether the conversion was successful. If so, the method proceeds to step 436; otherwise, control passes to step 426, wherein an exception is generated 426 and the attribute 310 is ignored.

After the attribute is converted, or if no conversion is necessary, the write method 504 is executed 436 with the attribute 310 value being passed as a parameter. Thereafter, a determination 440 is made whether more attributes 310 in the element 306 remain to be mapped to the component 212. If more attributes 310 remain, the method returns to step 420 to select the next attribute; otherwise, the method is complete.

As will be apparent from the foregoing discussion, for each attribute 310 in a given element 306, the element processor 118 set the properties 320 of the target component 212. An example of this process in Java pseudo-code is provided below:

The setProperty() pseudo-method identified above takes a component object 212, a property 320, and a value, and attempts to call the property's write method 504, converting the value(s) to the correct data type(s), as follows:

In one embodiment, execution of the method is performed using the standard reflection functionality built into Java as of version 1.1 as specified by Sun Microsystems. Preferably, the element processors 118 expect all information about the underlying component 212 to be provided explicitly by the BeanInfo instance 124. Alternatively, runtime introspection could be used. One skilled in the art will recognize, however, that a variety of other element mappers 124 may be used within the scope of the present invention, and a Beaninfo instance 124 may be optional or unnecessary.

Referring again to FIG. 4B, after all of the attributes 310 in the attribute list 312 are mapped, a determination 416 is made whether an exception was signaled as a result of extraneous attributes 310, or whether any required attributes 310 were missing from the element 306. If either case is true, an error message is generated 412 and the method is complete.

Referring now to FIG. 4D, there is shown a method of processing 406 the components 212 to launch the component-based application 214 in accordance with a preferred embodiment of the present invention. Preferably, an object graph 318 including a plurality of components 212 was created in step 404. Starting with the root component 212, the associated element processor 118 is invoked, which applies 450 any specified initial properties 320 to the component 212. In one embodiment, such properties 320 may be specified in an <INIT> section of the corresponding element 306.

Thereafter, the element processor 118 calls 452 any initialization method for the current component 212. In one embodiment, the initialization method may be specified in an <INIT> section of the corresponding element 306. Alternatively, a default method, such as init(), may be used, as in the standard Java Applet class.

After the component 212 is initialized, the method continues by invoking the element processors 118 associated with any child elements, and instructing the element processors 118 to recursively perform each of the steps 450-460 of the current method. Because the initialization step 452 is performed using pre-order recursion, parent components 212 are initialized before their child components 212.

When a component 212 without children (a leaf component 212) is encountered, the method continues by determining 456 whether the parent component 212 is of the appropriate type to have child components 212 (i.e., it is a "container element"). This will depend on the underlying specification of the component 212 within the framework 130. If the component cannot have children, an error is generated 458. If, however, the component is a container element, the method continues by adding 460 the child component 212 to the parent component 212. After either step 460 or 458, control is returned to the parent element processor 118. When complete, all child components 212 are added to their parents 212, from the bottom of the object graph 318, upwards.

In one embodiment of the invention, after all of the components 212 have been processed, the init() and start() methods of the root component 212 are invoked in order to begin actual execution of the application 214. Preferably, the root component 212 implements the interface, MinAppIntf which is the minimal interface that is expected for any root-level component 212 to implement in order to be correctly instantiated and launched. In one embodiment, the MinAppIntf's interface is patterned after Java's standard Applet class. An example of the interface is provided below in Java pseudo-code:

The MinAppIntf also specifies a setWindow(MainWindowIntf theWindow) method which is used to set the underlying window instance. In one embodiment, each application has one and only one top-level main window. The top-level main window may be platform-specific (i.e. Windows 95, Macintosh, etc.). It manages the application's "event loop" and thus receives update and user event messages generated by the platform-specific operating system. A given application may have an unlimited number of "child" windows.

The application's main window is set prior to the init() sequence described above. The application's main window is responsible for communicating user and "paint" events to the application through its handleEvent(Event e) call. Finally, the window should be capable of supplying a graphics context on which the application's visible surface may be painted. The root component 212 class uses the start() method within the current processor 104 thread to launch the application 214 itself. After the start() call is made and finishes, the control passes to the created application 214.

The above description is included to illustrate the operation of the preferred embodiments and is not meant to limit the scope of the invention. The scope of the invention is to be limited only by the following claims. From the above discussion, many variations will be apparent to one skilled in the art that would yet be encompassed by the spirit and scope of the present invention. ##SPC1##