Usenet 1994 January

home *** CD-ROM | disk | FTP | other *** search

/ Usenet 1994 January / usenetsourcesnewsgroupsinfomagicjanuary1994.iso / answers / object-faq / part1 < prev next >

Wrap

Text File | 1993-12-14 | 60KB | 1,196 lines

Newsgroups: comp.object,comp.answers,news.answers Path: senator-bedfellow.mit.edu!bloom-beacon.mit.edu!nic.hookup.net!swrinde!cs.utexas.edu!howland.reston.ans.net!agate!msuinfo!uchinews!news From: Bob Hathaway <rjh@geodesic.com> Subject: Comp.Object FAQ Version 1.0.5 (12-13) Part 1/8 Message-ID: <1993Dec14.043920.17948@midway.uchicago.edu> Followup-To: comp.object Summary: Frequently Asked Questions (FAQ) List and Available Systems For Object-Oriented Technology Sender: news@uchinews.uchicago.edu (News System) Organization: Geodesic Systems Date: Tue, 14 Dec 1993 04:39:20 GMT Approved: news-answers-request@MIT.Edu Lines: 1181 Xref: senator-bedfellow.mit.edu comp.object:13848 comp.answers:2989 news.answers:15747 Archive-name: object-faq/part1 Last-Modified: 12/13/93 Version: 1.0.5 COMP.OBJECT FAQ Version: 1.0.5 Date: 12/13/1993 Author: Bob Hathaway Geodesic Systems rjh@geodesic.com Anonymous FTP Sites and Hypertext Server: anonymous@zaphod.uchicago.edu:/pub/comp.object-faq.1.0.5(.Z) (128.135.72.61) anonymous@rtfm.mit.edu:/pub/usenet/comp.object/*_Part_* http://cui_www.unige.ch/OSG/FAQ/OO-FAQ/index.html Mail Server: (See also section 1.24) mail mail-server@rtfm.mit.edu Subject: send usenet/comp.object/* Zaphod is preferred over rtfm for anonymous ftp retrieval, as it provides a single file. Rtfm contains the FAQ as posted. To use the hypertext system, see APPENDIX E, entries 27. There is currently no comp.object archive. Copyright 1992-1994 Bob Hathaway All rights reserved. Permission is granted to freely copy and distribute this document but only at no cost to others and with the exception of a nominal distribution fee, if any. No sale, resale or reprinting is granted without the explicit written permission of the author. Contributors: Per Abrahamsen, Stewart Clamen, Edwardo Casais, Dennis De Champeaux, Mike DeVaney, Eric Dujardin, Piercarlo Grandi, Brian Henderson- Sellers, Urs Hoelzle, Paul Johnson, Bill Kinnersley, James Odell, Eoin Woods and many others whose contributions have helped this document to fulfull its objective of bringing object-oriented concepts and systems to everyone. Objective: In the spirit of other FAQs, to provide a simple document to answer the most frequently asked and recurring questions and to allow new users to understand frequently discussed topics and terms used in comp.object. This should bring new comp.object readers and/or writers to at least an introductory level of comprehension as soon as possible. Other goals (hopes) are to provide a quick and current reference on available systems such as object- oriented languages, CASE, OODB and etc. and to provide good references to current and relevant OO systems, groups, texts and literature. Disclaimer: This document does not necessarily reflect the opinions of the author's or any contributor's companies. There are no explicit or implicit guarantees implied by this document. While object systems are a constantly changing and moving target with a broad diversity of often conflicting methodologies, constructs, terminologies, approaches, languages, implementations and etc. and comp.object has a wide diversity of readers and writers ranging from students, professors and researchers in academia to beginners, professionals, top-notch experts and leaders in industry with a broad range of experience and backgrounds ranging across many paradigms, this FAQ can certainly not aspire to satisfy all of them completely but instead attempts to provide the most well-rounded treatment of object-oriented concepts and realizations primarily from the mainstream and popular authors and systems and further to provide a collection of available systems and tools in the appendices. It should be understood that it is not possible to present a complete, unbiased, and perfect document on the subject to be everything to all people; however, herein lies at least one effort to provide a reasonable compromise. Several improvements are planned for future FAQs, including a glossary. SECTION 1: BASICS 1.1) What Is An Object? 1.2) What Is Object Encapsulation (Or Protection)? 1.3) What Is A Class? 1.4) What Is A Meta-Class? 1.5) What Is The Infinite Regress Of Objects And Classes? 1.6) What are MOPs and Reflection? 1.7) What Is Inheritance? 1.8) What Is Multiple Inheritance? 1.9) Does Multiple Inheritance Pose Any Additional Difficulties? 1.10) What Is Dynamic Inheritance? 1.11) What Is Shared (Repeated) Inheritance? 1.12) Why Use Inheritance? 1.13) Why Don't Some People Like Inheritance? 1.14) What Is Specialization/Generalization/Overriding? 1.15) What Is The Difference Between Object-Based And Object-Oriented? 1.16) Is A Class An Object? 1.17) Is An Object A Class? 1.18) What Is A Method? (And Receiver And Message) 1.19) What Are Multi-Methods And Multiple-Polymorphism? 1.20) What Is OOP? 1.21) What Is OOA/OOD (And Where Can I Get What I Need On It)? 1.22) Where Did Object-Orientation Come From? 1.23) What Are The Benefits Of Object-Orientation? 1.24) What Other FAQs Are available? SECTION 2: TYPING 2.1) What Is Polymorphism? 2.2) What Does Polymorphism Boil Down To In OO Programming Languages? 2.3) What Is Dynamic Binding? 2.4) Is There A Difference Between Being A Member Or Instance Of A Class? 2.5) What Is This I Read About ML And Functional Programming Languages? 2.6) What Is the Difference Between Static And Dynamic Typing? 2.7) What Is A Separation Between Type And Class (Representation)? 2.8) What Are Generics And Templates? SECTION 3: GENERAL 3.1) What Is The "Classical" Object-Oriented Paradigm? 3.2) What Is The "Delegation/Prototyping" Object-Oriented Paradigm? 3.3) Are There Any Other Object-Oriented Paradigms? 3.4) What Are The Major Object-Oriented Programming Languages Today? 3.5) What Are Object-Oriented Databases And Persistence? 3.6) What Are Object-Oriented Operating Systems? 3.7) What Are The Current Object-Oriented Methodologies? 3.8) What Is The OMG/OMA/ORB/CORBA? 3.9) Why Is Garbage Collection A Good Thing? 3.10) What Can I Do To Teach OO To The Kids? 3.11) What Is Available On Object-Oriented Testing? 3.12) What Distributed Systems Are Available? 3.13) What Is The MVC Framework? 3.14) What is Real-Time? SECTION 4: COMMONLY ASKED LANGUAGE SPECIFIC QUESTIONS 4.1) What Is Downcasting? 4.2) What Are Virtual Functions? 4.3) Can I Use Multiple-Polymorphism Or Multi-Methods In C++? 4.4) Can I Use Dynamic Inheritance In C++? ANNOTATED BIBLIOGRAPHY APPENDIXES APPENDIX A VIPS APPENDIX B OBJECT-ORIENTED DATABASES AND VENDORS APPENDIX C OBJECT-ORIENTED LANGUAGES AND VENDORS APPENDIX D OBJECT-ORIENTED CASE (OOA/D/P TOOLS) AND VENDORS APPENDIX E ANONYMOUS FTP SITES APPENDIX F MAGAZINES, JOURNALS AND NEWSLETTERS [Another appendix on miscellaneous commercial object-oriented systems should be added soon for completeness] SECTION 1: BASICS ================== Suggested Readings: [Booch 91, 94] Others to be added... 1.1) What Is An Object? ----------------------- There are many definitions of an object, such as found in [Booch 91, p77]: "An object has state, behavior, and identity; the structure and behavior of similar objects are defined in their common class; the terms instance and object are interchangeable". This is a "classical languages" definition, as defined in [Coplien 92, p280], where "classes play a central role in the object model", since they do not in prototyping/delegation languages. "The term object was first formally applied in the Simula language, and objects typically existed in Simula programs to simulate some aspect of reality" [Booch 91, p77]. Other definitions referenced by Booch include Smith and Tockey: "an object represents an individual, identifiable item, unit, or entity, either real or abstract, with a well-defined role in the problem domain." and [Cox 91]: "anything with a crisply defined boundary" (in context, this is "outside the computer domain". A more conventional definition appears on pg 54). Booch goes on to describe these definitions in depth. [Martin 92, p 241] defines: "An "object" is anything to which a concept applies", and "A concept is an idea or notion we share that applies to certain objects in our awareness". [Rumbaugh 91] defines: "We define an object as a concept, abstraction or thing with crisp boundaries and meaning for the problem at hand." [Shlaer 88, p 14] defines: "An object is an abstraction of a set of real-world things such that: * all of the real-world things in the set - the instances - have the same characteristics * all instances are subject to and conform to the same rules" and on identifying objects: "What are the *things* in this problem? Most of the things are likely to fall into the following five categories: Tangible things, Roles, Incidents, Interactions, and Specifications." [Booch 91, 4.3] covers "Identifying Key Abstractions" for objects and classes based on an understanding of the problem domain and [Jacobson 92] provides a novel approach to identifying objects through use-cases. The implementation of objects could roughly be categorized into descriptor- based, capability-based, and simple static-based approaches. Descriptor- based approaches (e.g. Smalltalk handles) allow powerful dynamic typing, as do the capability-based approaches which are typically found in object- oriented databases and operating systems (object id's). A "proxy" based approach with an added layer of indirection to Smalltalk's handles is found in Distributed Smalltalk which allows transparent, distributed, and migrating objects [Kim 89, ch 19 and Yaoqing 93]. Simple static approaches are found in languages such as C++, although the new RTTI facility will supply simple dynamic typing, which is also present in new versions of Eiffel. Descriptor-based approaches can have pointer semantics and can be statically typeless (or just "typeless", as in Smalltalk) where references (variables) have no type, but the objects (values) they point to always do. An untyped pointer (such as void* in C++) and an embedded dynamic typing scheme are used in more conventional languages to fully emulate this style of dynamically typed programming (see sections 2.3, 4.3, and [Coplien 92]). Below is a simple example to show a most trivial case of OO implementation. It is primarily intended to introduce new terms. See [Cardelli 85] for another semantic definition of OO using functions for methods and for a view of types as sets of values. Simple statically-typed objects (static and auto vars and temps in C++ and expanded types in Eiffel) can be viewed as instances of a record type, whose record fields are called instance variables (Smalltalk) or member data (C++). The record (class) may also contain operations which are called methods (Smalltalk) or member functions (C++) which are equivalent to a function taking an object of the record type, called the receiver, as the first parameter. The receiver is called self (Smalltalk) or this (C++). Members will denote both instance variables and methods. Inheritance is roughly equivalent to a loosely coupled variant record, with derived classes as variant parts and with multiple-inheritance concatenating several records to serve as a base. A virtual member in statically typed languages is a base class member that can be set or respecified by a derived class. This is roughly equivalent to a pointer or function pointer in the base class being set by the derived class. [Stroustrup 90] covers the implementation details of virtual member functions in C++, which also involve an offset for the receiver to handle multiple- inheritance. This is an example of dynamic binding, which replaces a switch statement on variant parts with a single call, reducing code size and program complexity (fewer nested programming constructs) and allowing variants to be added without modifying client code (which causes higher defect injection rates during maintanance and debugging). Virtual members in dynamically typed languages are more flexible because static typechecking requirements are dropped. See section 2.5. The terms method/member function, instance variable/member data, subclass/ derived class, parent class/base class, and etc. will be used interchangeably. As pointed out in [Stroustrup 90, p197], the base/derived class terminology may be preferable to the sub/super-class terminology, and is preferred in this document also. Delegation/prototyping languages [Kim 89, ch3; Ungar 87, Sciore 89] have a more flexible kind of object which can play the role of classes in classical OO languages. Since there is no separate class construct in these languages, and only objects, they are referred to as single-hierarchy, or 1 Level systems. Objects contain fields, methods and delegates (pseudo parents), whereas classical object-oriented languages associate method, field and parent definitions with classes (and only associate state and class with objects, although vtables of function pointers for dynamic binding is an exception). However, one-level objects often play the role of classes to take advantage of sharing and often instances will simply delegate to parents to access methods or shared state, otherwise idiosyncratic objects, a powerful and natural concept, will result. Typical 1 Level objects can contain any number of fields, methods and parents and any object can be used as a template/exemplar, thus performing the classical role of a class. In typical prototyping systems, parents (as any other member) can be added or changed dynamically, providing dynamic multiple inheritance (or more typically simple delegation). Here, the term "Prototype" usually refers to prototype theory, a recent theory of classification where any object can be inherited from or cloned to serve as a prototype for newly created instances. [The Author also uses the term for languages providing high quality support for rapid prototyping, although this usage is atypical] See [Booch 94, pp 154-155] for a brief discussion of prototype theory in the context of OOA and OOD. It is common in such systems for an object to "become" another kind of object by changing its parent. A good example is a window becoming an icon, since window and icon objects display different behavior (although cognitive differences are significant too:-) Delegation refers to delegating the search for an attribute to a delegate, and is therefore more of a pure message passing mechanism (as with dynamic scoping) than inheritance, which also typically specifies non-shared state when used for representation. Chambers has proposed an interesting variation called "Predicate Classes" [Chambers 93] as a part of his Cecil language. These classes will only be parents when certain predicates are true. This can support a types/classes as collections of objects view, which is the same as the types as sets of values view taken by [Cardelli 85]. [Martin 92] provides some examples of this view applied during OOA. 1 level systems therefore provide the most flexible and powerful capabilities. Self is a good example of a delegation-based single hierarchy language [Ungar 87]. 1.2) What Is Object Encapsulation (Or Protection)? --------------------------------------------------- [Booch 91, p. 45] defines: "Encapsulation is the process of hiding all of the details of an object that do not contribute to its essential characteristics." [Coad 91, 1.1.2] defines: "Encapsulation (Information Hiding). A principle, used when developing an overall program structure, that each component of a program should encapsulate or hide a single design decision... The interface to each module is defined in such a way as to reveal as little as possible about its inner workings. [Oxford, 1986]" Some languages permit arbitrary access to objects and allow methods to be defined outside of a class as in conventional programming. Simula and Object Pascal provide no protection for objects, meaning instance variables may be accessed wherever visible. CLOS and Ada allow methods to be defined outside of a class, providing functions and procedures. While both CLOS and Ada have packages for encapsulation, CLOS's are optional while Ada's methodology clearly specifies class-like encapsulation (Adts). However most object-oriented languages provide a well defined interface to their objects thru classes. C++ has a very general encapsulation/protection mechanism with public, private and protected members. Public members (member data and member functions) may be accessed from anywhere. A Stack's Push and Pop methods will be public. Private members are only accessible from within a class. A Stack's representation, such as a list or array, will usually be private. Protected members are accessible from within a class and also from within subclasses (also called derived classes). A Stack's representation could be declared protected allowing subclass access. C++ also allows a class to specify friends (other (sub)classes and functions), that can access all members (its representation). Eiffel 3.0 allows exporting access to specific classes. For another example, Smalltalk's class instance variables are not accessible from outside of their class (they are not only private, but invisible). Smalltalk's methods are all public (can be invoked from anywhere), but a private specifier indicates methods should not be used from outside of the class. All Smalltalk instance variables can be accessed by subclasses, helping with abstract classes and overriding. Another issue is per-object or per-class protection. Per-class protection is most common (e.g. Ada, C++, Eiffel), where class methods can access any object of that class and not just the receiver. Methods can only access the receiver in per-object protection. This supports a subtyping model, as any object other than the receiver is only satisfying an abstract type interface, whereby no method or object structure can be inferred in the general case. 1.3 What Is A Class? -------------------- A class is a general term denoting classification and also has a new meaning in object-oriented methods. Within the OO context, a class is a specification of structure (instance variables), behavior (methods), and inheritance (parents, or recursive structure and behavior) for objects. As pointed out above, classes can also specify access permissions for clients and derived classes, visibility and member lookup resolution. This is a feature-based or intensional definition, emphasizing a class as a descriptor/constructor of objects (as opposed to a collection of objects, as with the more classical extensional view, which may begin the analysis process). Original Aristotlean classification defines a "class" as a generalization of objects: [Booch 91, p93] "a group, set, or kind marked by common attributes or a common attribute; a group division, distinction, or rating based on quality, degree of competence, or condition". [Booch's definition in the context of OOD] "A class is a set of objects that share a common structure and a common behavior." "A single object is simply an instance of a class." The intension of a class is its semantics and its extension is its instances [Martin 92]. [Booch 94, 4.2] proposes 3 views of classification as useful in OO analysis and design: classical categorization (common properties), conceptual clustering (conceptual descriptions), and prototype theory (resemblance to an exemplar). He advocates starting with the former approach, turning to the second approach upon unsatisfactory results, and finally the latter if the first two approaches fail to suffice. 1.4) What Is A Meta-Class? --------------------------- [See also section 1.6] A Meta-Class is a class' class. If a class is an object, then that object must have a class (in classical OO anyway). Compilers provide an easy way to picture Meta-Classes. Classes must be implemented in some way; perhaps with dictionaries for methods, instances, and parents and methods to perform all the work of being a class. This can be declared in a class named "Meta-Class". The Meta-Class can also provide services to application programs, such as returning a set of all methods, instances or parents for review (or even modification). [Booch 91, p 119] provides another example in Smalltalk with timers. In Smalltalk, the situation is more complex. To make this easy, refer to the following listing, which is based on the number of levels of distinct instantiations: 1 Level System All objects can be viewed as classes and all classes can be viewed as objects (as in Self). There is no need for Meta-Classes because objects describe themselves. Also called "single-hierarchy" systems. There is only 1 kind of object. 2 Level System All Objects are instances of a Class but Classes are not accessible to programs (no Meta-Class except for in the compiler and perhaps for type-safe linkage, as in C++). There are 2 kinds of distinct objects: objects and classes. 3 Level System All objects are instances of a class and all classes are instances of Meta-Class. The Meta-Class is a class and is therefore an instance of itself (really making this a 3 1/2 Level System). This allows classes to be first class objects and therefore classes are available to programs. There are 2 kinds of distinct objects (objects and classes), with a distinguished class, the metaclass. 5 Level System What Smalltalk provides. Like a 3 Level System, but there is an extra level of specialized Meta-Classes for classes. There is still a Meta-Class as in a 3 Level System, but as a class it also has a specialized Meta-Class, the "Meta-Class class" and this results in a 5 Level System: object class class class (Smalltalk's Meta-Classes) Meta-Class Meta-Class class The "class class"es handle messages to classes, such as constructors and "new", and also "class variables" (a term from Smalltalk), which are variables shared between all instances of a class (static member data in C++). There are 3 distinct kinds of objects (objects, classes, and metaclasses). 1.5) What Is The Infinite Regress Of Objects And Classes? ---------------------------------------------------------- In the authors opinion, a myth. The story goes an object is an instance of a class (Meta-Object), a class is an instance of a Meta-Class, which must also be an instance of a Meta-Meta-Class, which must also be an instance of a Meta- Meta-Meta-Class, ... Closure can be achieved with an instance-of loop, as with a Meta-Class being an instance of itself or with a "Meta-Class - Meta-Class class" instance-of loop (as in Smalltalk). 1.6) What Are MOPs And Reflection? ----------------------------------- MOP is an acronym for Meta-Object Protocol. This is a system with Meta-Classes accessible to users [Kiczales 92]. In CLOS terminology, an introspective protocol provides a read only capability (e.g. what is this object's class, give info on this class, etc.) and an intercessory protocol provides a write capability which allows system modification (e.g. add the following method or instance to this class, perform inheritance this way, etc.). Because inheritance can be used to perform differential changes, intercessory protocols allow users to not only define new frameworks but to specialize existing system frameworks differentially without affecting them and their extant objects. Thus, many frameworks can interoperate together simultaneously. This is a good example of object-oriented reuse, since the compiler itself is reused thru specialization to provide new frameworks. "Reflective" systems are systems with MOPs (not to be confused with reflexive systems, which often refer to systems implemented in terms of themselves, or bootstrapped). Reflective systems are inevitably reflexive (as are most quality compilers), providing a direct program interface to the system. 1.7) What Is Inheritance? -------------------------- Inheritance provides a natural classification for kinds of objects and allows for the commonality of objects to be explicitly taken advantage of in modeling and constructing object systems. See the example below using computers. Inheritance is a relationship between classes where one class is the parent (base/superclass/ancestor/etc.) class of another. Inheritance provides programming by extension (as opposed to programming by reinvention [LaLonde 90]) and can be used as an is-a-kind-of (or is-a) relationship or for differential programming. Inheritance can also double for assignment compatibility (see section 2.7). In delegation languages, such as Self, inheritance is delegation where objects refer to other objects to respond to messages (environment) and do not respecify state by default. Inherited parents can specify various flavors of state. Delegation languages don't specify new state by default (to do so requires cloning), C-based (C++, Eiffel, Objective-C, etc.), lisp-based (CLOS, Flavors, Scheme, etc.), and Pascal-based (Ada-9x, Modula-3, Object Pascal, etc.) OO languages do, but with multiple-inheritance can also share parents within a class lattice (CLOS and Eiffel provide this as a default at the level of slots and features, respectively). Inheritance also provides for member lookup, or internal environment. Various schemes exist, for example C++ finds the closest match within a scope but causes an ambiguity error iff more than one parent has match, CLOS creates a linear precedence list, Self provides parent priorities, and Eiffel forces renaming for any parent member conflicts. Defining inheritance (with a thorough description or denotational semantic definition, or both) can avoid confusion about which inheritance scheme is being used (especially in OOD), because inheritance has many variations and combinations of state and environment (sometimes with complex rules). Inheritance can also be used for typing, where a type or class can be used to specify required attributes of a matching object (see sections 2.1, 2.7 and [Cardelli 85]). It would be more judicious to have discussions on how inheritance should be defined instead of over what it is, since it has many existing uses and semantics. An example of the is-a-kind-of relationship is shown below. Is-a is often used synonymously, but can be used to show the "object is-a class" instantiation relationship. In classical OO, inheritance is a relationship between classes only. In one-level systems, is-a (object instantiation) and is-a-kind-of (inheritance) are merged into one [Ungar 87, Madsen 93, Sciore 89]. Computer / | \ Mainframe Mini Personal / \ ... / \ Data Proc Scientific PC Workstation Class hierarchies are subjective [Booch 91, 4.2; Lakoff 87] and usually drawn with the parent class on top, but more demanding graphs (as is often the case in [Rumbaugh 91]) allow any topology, with the head of an arrow indicating the base class and the tail indicating the derived class. Differential programming is the use of inheritance to reuse existing classes by making a small change to a class. Creating a subclass to alter a method or to add a method to a parent class is an example. 1.8) What Is Multiple Inheritance? ----------------------------------- Multiple Inheritance occurs when a class inherits from more than one parent. For example, a person is a mammal and an intellectual_entity, and a document may be an editable_item and a kind of literature. Mixin's is a style of MI (from flavors) where a class is created to provide additional attributes or properties to other classes. They are intended to be inherited by any class requiring them. Method combination, or calling sequences of before, after, and around methods or even several primary methods [Kim 89, ch 4], make good use of mixins by invoking their methods without explicitly calling them, allowing client class code to remain unchanged [Booch 91, p 113]. 1.9) Does Multiple Inheritance Pose Any Additional Difficulties? ----------------------------------------------------------------- Yes, it does. Any name can be simply resolved to a class member with single inheritance by simply accessing the first name encountered for data members and by accessing the first signature match (or ambiguity) encountered for methods (at least one way, C++ hides some member functions). Since several distinct parents can declare a member within a multiple inheritance hierarchy, which to choose becomes an issue. Eiffel forces derived classes to rename parent members that conflict. Self prioritizes parents. CLOS merges member "slots" (instance variables) with the same name into a single slot, as did the earlier flavors. C++ declares an error iff a conflict arises, but a class qualifier can be used to explicitly disambiguate. Smalltalk renders same names for instance variables of subclasses illegal. On the other hand, multiple-inheritance can be seen as required for basic object-oriented programming, because many objects in the real world belong to several classes. In classical systems without MI, a class which should inherit from more than one class must textually include all but one of those classes in its interface, causing code duplication (and a messy interface). 1.10) What Is Dynamic Inheritance? ----------------------------------- Dynamic inheritance allows objects to change and evolve over time. Since base classes provide properties and attributes for objects, changing base classes changes the properties and attributes of a class. A previous example was a window changing into an icon and then back again, which involves changing a base class between a window and icon class. More specifically, dynamic inheritance refers to the ability to add, delete, or change parents from objects (or classes) at run-time. Actors, CLOS, and Smalltalk provide dynamic inheritance in some form or other. Single hierarchy systems, such as Self, provide dynamic inheritance in the form of delegation [Ungar 87]. See also [Kim 89, chs 1, 3] for a discussion and [Coplien 92] for some implementation discussion in C++. 1.11) What Is Shared (Repeated) Inheritance? --------------------------------------------- Multiple Inheritance brings up the possibility for a class to appear as a parent more than once in a class graph (repeated inheritance), and there is then a potential to share that class. Only one instance of the class will then appear in the graph (as is always the case in CLOS, because all *members* with the same name will be shared (receive a single slot) with the greatest common subtype as its type). C++ provides an alternative, where only parents specified as virtual (virtual bases) are shared within the same class lattice, allowing both shared and non-shared occurrences of a parent to coexist. All "features" in Eiffel (C++ members) of a repeated parent that are not to be shared must be renamed "along an inheritance path", else they are shared by default. This allows a finer granularity of control and consistent name resolution but requires more work for parents with many features. 1.12) Why Use Inheritance? --------------------------- Inheritance is a natural way to model the world or a domain of discourse, and so provides a natural model for OOA and OOD (and even OOP). This is common in the AI domain, where semantic nets use inheritance to understand the world by using classes and concepts for generalization and categorization, by reducing the real-world's inherent complexity. Inheritance also provides for code and structural reuse. In the above Computer class diagram, all routines and structure available in class Computer are available to all subclasses throughout the diagram. All attributes available in Personal computers are also available to all of its subclasses. This kind of reuse takes advantage of the is-a-kind-of relationship. Class libraries also allow reuse between applications, potentially allowing order-of-magnitude increases in productivity and reductions in defect rates (program errors), as library classes have already been tested and further use provides further testing providing even greater reliability. With differential programming, a class does not have to be modified if it is close to what's required; a derived class can be created to specialize it. This avoids code redundancy, since code would have to be copied and modified otherwise. See [Raj 89] for an alternative approach as found in Jade. Polymorphism is often explicitly available in many OO languages (such as C++, CLOS, Eiffel, etc.) based on inheritance when type and class are bound together (typing based on subclassing, or subclass polymorphism), since only an object which is a member of (inherits from) a class is polymorphically assignment compatible with (can be used in place of) instances or references of that class. Such assignment can result in the loss of an object's dynamic type in favor of a static type (or even loss of an object's representation to that of the static class, as in C++ slicing). Maintaining the dynamic type of objects can be provided (and preferred); however, C++ provides both sliced and non- sliced replacement in a statically typed environment (see section 2.1). 1.13) Why Don't Some People Like Inheritance? ---------------------------------------------- Some people complain that inheritance is hierarchical (which is what most object-oriented languages provide). They would also like to see more operations available (set operations are quite common in specialized systems). The former is a kind of language dependent feature commonly found in object- oriented languages which are then associated with the term "inheritance" (although they don't need to be. For example, delegation languages allow graph inheritance stuctures). Some don't like the coupling of classes (as in Jade), but in the author's opinion many of their complaints are easily answered. In systems that provide inheritance, inheritance provides a simple and elegant way to reuse code and to model the real world in a meaningful way. Others complain multiple inheritance is too complicated because it brings up the issues of shared bases and member conflict resolution. But most modern systems support Multiple Inheritance by employing semantic resolution strategies or renaming, and most consider MI to be highly desirable. See the latter part of section 1.9 for an example of why MI is important. Some prefer association to MI, claiming "roles" (as defined in [Rumbaugh 91]) should be associations and inheritance should be reserved for a single hierarchy "creation" mechanism, however this loses polymorphism and loses the use of inheritance for typical classification. Representation "roles" can be supported by dynamic multiple inheritance (DMI) in many situations. 1.14) What Is Specialization/Generalization/Overriding? -------------------------------------------------------- To create a subclass is specialization, to factor out common parts of derived classes into a common base (or parent) is generalization [Booch 91, p56]. Overriding is the term used in Smalltalk and C++ for redefining a (virtual in Simula and C++) method in a derived class, thus providing specialized behavior. All routines in Smalltalk and Eiffel are overridable (although in Eiffel they must be "redefined" in a derived class). Whenever a method is invoked on an object of the base class, the derived class method is executed overriding the base class method, if any. Overriding in Simula is a combination of overloading and multiple-polymorphism because parameters do not have to be declared. Eiffel and Beta are examples of languages allowing any member to be redefined and not just methods, as is typical. 1.15) What Is The Difference Between Object-Based And Object-Oriented? ----------------------------------------------------------------------- Object-Based Programming usually refers to objects without inheritance [Cardelli 85] and hence without polymorphism, as in '83 Ada and Modula-2. These languages support abstract data types (Adts) and not classes, which provide inheritance and polymorphism. Ada-9x and Modula-3; however, support both inheritance and polymorphism and are object-oriented. [Cardelli 85, p481] state "that a language is object-oriented if and only if it satisfies the following requirements: - It supports objects that are data abstractions with an interface of named operations and a hidden local state. - Objects have an associated type. - Types may inherit attributes from supertypes. object-oriented = data abstractions + object types + type inheritance These definitions are also found in [Booch 91, Ch2 and Wegner 87]. [Coad 91] provides another model: Object-Oriented = Classes and Objects + Inheritance + Communication with messages Stroustrup's first edition of [Stroustrup 91, '86 p. 37] defines object based as: "... storing type identification in each object, brings us to a style of programming often referred to as "object based"", which is quite different from C+W's. A more modern definition of "object-oriented" includes single-hierarchy languages and perhaps object id's for unique objects. Object id's support the modern notion of relocatable, persistent and distributed objects that can even migrate across machines. Distributed Smalltalk's proxy objects [Kim 89, ch 19 and Yaoqing 93] provide another example of a distributable and migratable object facility. Separate type system support is another extension. [Booch 94, 2.2] proposes 7 "Elements of the Object Model"; 4 major and 3 minor: Major: Abstraction Encapsulation Modularity Hierarchy (Inheritance) Minor: Typing Concurrency Persistence 1.16) Is A Class An Object? ---------------------------- In C++ no, because C++ classes are not instances of an accessible class (a Meta-Class) and because C++ classes are not accessible to programs. Classes are objects in 3 Level Systems and above because classes are instances of meta-classes. But classes play a dual role, because objects can only be declared to be instances of a class (and class objects instances of a meta-class). In 1 Level (single-hierarchy) systems, all classes are objects. 1.17) Is An Object A Class? ---------------------------- In a Level 3 System and above yes, but only instances of a Meta-Class are Classes. Instances of a Class (ordinary objects) are not classes (excluding hybrid systems). However, all objects may be classes in single hierarchy systems, since any object may act as a class (provide object instantiation or act as a shared parent). 1.18) What Is A Method? (And Receiver And Message) --------------------------------------------------- A method implements behavior, which is defined by [Booch 91, p80]: Behavior is how an object acts and reacts, in terms of its state changes and message passing. A method is a function or procedure which is defined in a class and typically can access the internal state of an object of that class to perform some operation. It can be thought of as a procedure with the first parameter as the object to work on. This object is called the receiver, which is the object the method operates on. An exception exists with C++'s static member functions which do not have a receiver, or "this" pointer. The following are some common notations for invoking a method, and this invocation can be called a message (or message passing, see below): receiver.message_name(a1, a2, a3) receiver message_name: a1 parm1: a2 parm3: a3 Selector would be another good choice for message_name in the above examples, although keywords (or formal parameter names, like named parameters) are considered part of the selector in Smalltalk (and hence Objective-C). If done statically, this can be referred to as invocation, and message passing if done dynamically (true dynamic binding). Statically typed dynamic binding (e.g. C++ and Eiffel) is really in between (checked function pointers). See also section 1.19 below for a discussion on the functional (prefix) verses message based (receiver based) notation. 1.19) What Are Multi-Methods And Multiple-Polymorphism? -------------------------------------------------------- Multi-methods involve two primary concepts, multiple-polymorphism and lack of encapsulation. These issues are orthogonal. Multiple-polymorphism implies more than one parameter can be used in the selection of a method. Lack of encapsulation implies all arguments can be accessed by a multi-method (although packages can be used to restrict access, as in CLOS). Multi-methods can also imply a functional prefix notation, although the CLOS designers (who coined the term "multi-method") consider the functional and receiver based forms (messages) equivalent. Functional syntax was chosen "in order to minimize the number of new mechanisms added to COMMON LISP" [Kim ch 4, p70 (D. Moon)]. [Chambers 93] discusses multi-methods in his new OO language, Cecil. Multiple-polymorphism allows specialized functions or methods to be defined to handle various cases: +(int, int) +(int, float) +(int, complex) +(int, real) +(float, complex) +(float, real) +(float, float) The above functions are specialized to each of the cases required allowing single, highly cohesive and loosely coupled functions to be defined. This is also the true essence of object-oriented polymorphism, which allows objects to define methods for each specific case desired. In addition to better coupling and cohesion, multiple-polymorphism reduces program complexity by avoiding coding logic (switch statements) and because small methods further reduce complexity, as code complexity doesn't grow linearly with lines of code per method, but perhaps exponentially. This should be distinguished from double dispatch, a fancy name for single dispatch after a call, which only provides switching on a single argument per call (but for 2 levels), consistently ignoring the inherent type of parameters in messaging. Double dispatch is used in languages with static typing for efficiency considerations. If all of the above types are Numbers, code can be written without concern for the actual classes of objects present: fn(one, two: Number): Number return one + two; The addition expression above will invoke the correct "+" function based on the inherent (true, actual, or dynamic) types of one and two. Only the inherent type of "one" would be used with double dispatch! In the author's opinion, this is a serious shortcoming. Further, double dispatch would only allow switching to the "fn" function based on the type of "one" also. This could lead to the use of switch statements based on type or complex coding in many real-world programming situations, unnecessarily. In the author's opinion, this should only be used as necessary, e.g. if the implementation language doesn't support multiple-polymorphism and either efficiency considerations dominate and double dispatch can be suffered, or an embedded dynamic typing scheme is used. Why do multi-methods allow open access to parameters? It allows for efficient handling, like C++ friends, usually by allowing representation details of more than one object to be exposed. See [Kim ch 4, pp70-71 (D. Moon)] for an alternative explanation. While open access can be useful in some cases, it typically isn't recommended as a general OO practise (see section 1.15, C+W's requirement 1 for OO languages) and also violates subtype polymorphism, because only subclass polymorphism is based on representation and not type. Polymorphic languages can be statically typed to provide strong type checking, efficiency, and to support a static programming idiom, but require restrictions in many cases, such as requiring overriding methods to have identical signatures with the methods they substitute (as in C++) or allowing covariant parameters but limiting base class usage (as in Eiffel). If these restrictions are dropped, multiple-polymorphism results. Thus a single overridable function declared in a base class may have several functions overriding it in a derived class differentiated only by their formal argument types. This therefore requires both static and dynamic typing, because no formal argument differentiation is possible without static types, as in Smalltalk, and no actual argument differentiation is possible without dynamic types (as in C++ and Eiffel). See section 2.3 for another example of multiple-polymorphism. There is some concern about the efficiency of run-time method selection as can occur with multiple-polymorphism (or even dynamic message passing). However, static analysis optimizations are commonly available in the literature, potentially providing a single static selection in many cases [See Agrawal 91, Chambers 92, Mugridge 91, and etc.]. But coupling the two cases of selector variables (as found in CLOS, Objective-C, and etc.) and several possible known selectors together with the general undecidability of dynamic types at compile-time renders dynamic typing and run-time selection (or checking) as unavoidable in the general case [a point often mistaken in comp.object. E.g. simple statically/strongly typed multi-methods still require dynamic types!] See [Booch 91], multiple-polymorphism, for a good CLOS example. 1.20) What Is OOP? ------------------- OOP stands for Object-Oriented Programming, the usual programming/hacking and etc. most programmers think of. Modern software engineering methodologies; however, consider OOP as the implementation/evolution of an OOD. 1.21) What Is OOA/OOD (And Where Can I Get What I Need On It)? --------------------------------------------------------------- See also section 3.7, the Annotated Bibliography, and APPENDIX D. The classified bibliography in [Booch 94] also contains entries on OOA(B), OOD(F) and OOP(G). [Booch 91] "In OOA, we seek to model the world by identifying the classes and objects that form the vocabulary of the problem domain, and in OOD, we invent the abstractions and mechanisms that provide the behavior that this model requires." [Coad 91] "OOA is the challenge of understanding the problem domain, and then the system's responsibilities in that light". "To us, analysis is the study of a problem domain, leading to a specification of externally observable behavior; a complete, consistent, and feasible statement of what is needed; a coverage of both functional and quantified operational characteristics (e.g. reliability, availability, performance)". "Design. The practise of taking a specification of externally available behavior and adding details needed for actual computer system implementation, including human interaction, task management, and data management details." And on Domain Analysis: "Whereas OOA typically focuses upon one specific problem at a time, domain analysis seeks to identify the classes and objects that are common to all applications within a given domain, [...]". - [Booch 91] [The following quotes on domain analysis are from [Berard 93]] "An investigation of a specific application area that seeks to identify the operations, objects, and structures that commonly occur in software systems within this area. - Dan McNicholl "Systems analysis states what is done for a specific problem in a domain while domain analysis states what can be done in a range of problems in a domain. ...A domain analysis is only useful in many similar systems are to be built so that the cost of the domain analysis can be amortized over the cost of all the systems. The key to reusable software is captured in domain analysis in that it stresses the reusability of analysis and design, not code. - Jim Neighbors "The process of identifying, collecting, organizing, and representing the relevant information in a domain based on the study of existing systems and their development histories, knowledge captured from domain experts, underlying theory, and emerging technology within the domain." - Kang et al. Object-oriented domain analysis (OODA) seeks to identify reusable items localized around objects, e.g., classes, instances, systems of interacting objects, and kits [frameworks]. OORA analysts and OOD designers will interact on a fairly frequent basis with the domain analysis effort. OOA and OOD stand for Object-Oriented Analysis and Object-Oriented Design, respectively. OOA strives to understand and model, in terms of object-oriented concepts (objects and classes), a particular problem within a problem domain (from its requirements, domain and environment) from a user-oriented or domain expert's perspective and with an emphasis on modeling the real-world (the system and its context/(user-)environment). The product, or resultant model, of OOA specifies a complete system and a complete set of requirements and external interface of the system to be built. [Shlaer 88] is often credited as the first book on OOA, although their method adds OO techniques to the traditional structured analysis principles of Yourdon and Constantine. Their complete approach ([Shlaer 88, 92]) consists of information modeling and recursive design, or OOA/RD and represents a recent addition to the structured analysis family (as does Martin and Odell). [Yourdon 92] provides a critique, although may only refer to their earlier work. Many other methodologies including Rumbaugh's OMT, Martin and Odell's OOA/D, and many others, also share common ground with SA and other existing analysis methodologies with such constructs as associations (E-R), functional models, and even DFD's. Booch, Jacobson, and Wirfs-Brock are examples of OO methodologies representing a greater departure from the conventional "structured" techniques, with greater emphasis on objects. OOram [Reenskaug 91] provides support and emphasis on types and roles as guiding principles, which is quite powerful. [Booch 94] presents a methodology which is an evolutionary step beyond the first edition by incorporating a collection of the best features from several of the major OO methodologies, as does HP's new FUSION methodology. The usual progression is from OOA to OOD to OOP (implementation) and this Universal Process Model roughly corresponds to the Waterfall Model [Royce 70]. See [Humphrey 89] and [Yourdon 92] for a few of many discussions on software life-cycle models and their use. Humphrey also details Worldy and Atomic Process Models for finer grained analysis and design in the Defined Process (see below) and discusses other alternatives to the task oriented models. He also provides the following critisisms on the Waterfall Model which had led to Boehm's seminal work on the Spiral Model: * It does not adequately address changes * It assumes a relatively uniform and orderly sequence of development steps * It does not provide for such methods as rapid prototyping or advanced languages Modern OO methodologies directly address these points and emphasize the incremental, iterative, evolutionary, concurrent and situational nature of software development. [Boehm 86] presents a seminal spiral life-cycle model with a risk-driven prototyping approach. [Booch 91, 6.1] proposes a "round- trip gestalt" design with analyze-design iterations and an overall system perspective and [Berard 93] proposes an (incremental) "parallel-recursive design" with analyze-design-implement-test iterations. [Coad 91b] presents the following development cycle breakdown: Waterfall- Analysis Design Programming Spiral- Analysis, prototyping, risk management Design, prototyping, risk management Programming, prototyping, risk management [Boehm, 1988] Incremental- A little analysis A little design A little programming Repeat [Gilb, 1988] [Author's note: The spiral model may be incremental in many ways and ends with a waterfall.] Since classes and objects are used in all phases of the OO software life-cycle, the process is often referred to as seamless, meaning there is no conceptual gap between the phases as is often the case in other software development methodologies, such as the analysis (DFD's) to design (structure charts) to programming gaps found in traditional structured analysis and design. Seamlessness together with naturalness is a big advantage for consistency. A problem domain has many realizations, or differing OOAs. An OOA has many realizations, or differing OODs, but a similar notation is often used for the two. An OOD also has many realizations, or differing OOPs, but allows a selection from among various languages for implementation (choosing the best language to implement the design). But some, such as Bjarne Stroustrup, don't like OOA and OOD getting too far from OOP (implementation independent), for fear that great discrepancies could occur between OOD and OOP by losing sight of the implementation language, which in some cases is predetermined. See also [Stroustrup 91]. From a greater perspective, the SEI has developed the Software Process Maturity Model (SPMM), a 5 level model classifying the progressive levels of an organization's software development [Humphrey 89]. Level 1: Initial: Every project is handled differently Level 2: Repeatable: Every project is handled similarly Level 3: Defined: A standard process is now well defined for all projects Level 4: Managed: A measurable basis for all improvements to the process Level 5: Optimizing: Emphasis on defect prevention and optimizing/continually improving the process See also [Yourdon 92], [Wilkie 93], and [Booch 94] for discussions on this often cited model. There is also an ISO 9000 standard on software quality. Object-Oriented Information Engineering's "Enterprise Modeling" [Martin 92], also found in [Jacobson 92], and some recent business "reengineering" efforts place information systems within an organizational perspective by modeling entire organizations or a large part of them, with the information processing system and software products development as integrated components. [Yourdon 92] even calls for "global modeling"! 1.22) Where Did Object-Orientation Come From? ---------------------------------------------- Simula was the first object-oriented language providing objects, classes, inheritance, and dynamic typing in 1967 (in addition to its Algol-60 subset). It was intended as a conveyance of object-oriented design. Simula 1 was a simulation language, and the later general-purpose language Simula 67 is now referred to as simply Simula. Smalltalk was the next major contributor including classes, inheritance, a high-powered graphical environment and a powerful dynamic typing mechanism (although these existed to some extent in Simula). Self is somewhat of a Smalltalk-based next generation language, as is Beta a followup to Simula (by its original designers). [Meyer 88] contains a brief summary and history of Simula and Smalltalk, among other OO languages. 1.23) What Are The Benefits Of Object-Orientation? --------------------------------------------------- Reuse, quality, an emphasis on modeling the real world (or a "stronger equivalence" with the RW than other methodologies), a consistent and seamless OOA/OOD/OOP package, naturalness (our "object concept") and etc. [Harmon 93]: Faster development Increased Quality Easier maintenance Enhanced modifiability [Booch 94]: Exploit power of OOPs Reuse of software and designs, frameworks Systems more change resilient, evolvable Reduced development risks for complex systems, integration spread out Appeals to human cognition, naturalness 1.24) What Other FAQs Are Available? ------------------------------------- FAQ's are cross-posted to news.answers and are archived on anonymous ftp from: rtfm.mit.edu:/pub/usenet (also usenet-by-hierarchy, etc.) rtfm archives several FAQs pertinent to OO (most are also archived elsewhere): comp.lang.ada ajpo.sei.cmu.edu:public/comp-lang-ada/cla-faq[12] comp.lang.c++ sun.soe.clarkson.edu:pub/C++/FAQ [128.153.12.3] comp.lang.clos comp.lang.eiffel ftp.cm.cf.ac.uk:/pub/eiffel/eiffel-faq comp.lang.oberon comp.lang.objective-c comp.lang.sather ftp.ICSI.Berkeley.EDU:pub/sather [not on rtfm] comp.lang.scheme ftp.think.com:/public/think/lisp/scheme-faq.text comp.lang.smalltalk xcf.Berkeley.EDU:misc/smalltalk/FAQ/SmalltalkFAQ.entire comp.object zaphod.uchicago.edu:/pub/comp.object-faq.1.0.5(.Z) comp.object.logic ftp.cs.cmu.edu:*prg_1.faq,prg_2.faq [128.2.206.173] comp.software-eng [xcf.Berkeley.EDU is 128.32.138.1] * /afs/cs.cmu.edu/project/ai-repository/ai/pubs/faqs/prolog/ A new C++ libraries FAQ is posted monthly to comp.lang.c++ and should be on rtfm soon. Contact cpplibs@trmphrst.demon.co.uk. It contains anonymous ftp sites and commercial libraries and may be merged with this FAQ soon. Many FAQs are also available from mail-servers, however most can be accessed by the rtfm mail-server. Mail to mail-server@rtfm.mit.edu with help and index in the body with no leading spaces and on separate lines for more information. Example Unix Command (will retrieve this FAQ in about 26 pieces (and growing)): mail mail-server@rtfm.mit.edu Subject: send usenet/comp.object/* There is also a great ftp site for sci.virtual-worlds on: stein.u.washington.edu (140.142.56.1) - home of sci.virtual-worlds, huge faq w/ great info! - if unable to use try ftp.u.washington.edu /public/virtual-worlds [While vr may not be directly related to comp.object, it is most interesting! - The Author] SECTION 2: TYPING ================== There are many definitions of type (and class and related concepts). Many authors define the terms as applied by their particular approach or language, however we shall proceed in the face of this diversity. References [Blair 89] Some typing topics. [Booch 91] Small Section on Typing. [Cardelli 85] Discussion on Object-Oriented Typing. [Kim 89, ch1] Discussion on Some Research Topics. 2.1) What Is Polymorphism? --------------------------- Polymorphism is a ubiquitous concept in object-oriented programming and is defined in many ways, so many definitions are presented from: Websters', Author, Strachey, Cardelli and Wegner, Booch, Meyer, Stroustrup, and Rumbaugh. Polymorphism is often considered the most powerful facility of an OOPL. > Webster's New World Dictionary: Polymorphism 1. State or condition of being polymorphous. 2. Cryall. crystallization into 2 or more chemically identical but crystallographically distinct forms. 3. Zool., Bot. existence of an animal or plant in several forms or color varieties. polymorphous adj. having, assuming, or passing through many or various forms, stages, or the like. Also, polymorphic. [<Gk polymorphous multiform] > Author's Definition: Polymorphism is the ability of an object (or reference) to assume (be replaced by) or become many different forms of object. Inheritance (or delegation) specifies slightly different or additional structure or behavior for an object, and these more specific or additional attributes of an object of a base class (or type) when assuming or becoming an object of a derived class characterizes object-oriented polymorphism. This is a special case of parametric polymorphism, which allows an object (or reference) to assume or become any object (possibly satisfying some implicit or explicit type constraints (parametric type), or a common structure), with this common structure being provided by base classes or types (subclass and subtype polymorphism, respectively).