Object-oriented programming

From Wikipedia, the free encyclopedia

Object-oriented programming (OOP) is a programming paradigm based on the concept of "objects", which are data structures that contain data, in the form of fields, often known as attributes; and code, in the form of procedures, often known as methods. A distinguishing feature of objects is that an object's procedures can access and often modify the data fields of the object with which they are associated (objects have a notion of "this"). In object-oriented programming, computer programs are designed by making them out of objects that interact with one another.^[1][2] There is significant diversity in object-oriented programming, but most popular languages are class-based, meaning that objects are instances of classes, which typically also determines their type.
Many of the most widely used programming languages are multi-paradigm programming languages that support object-oriented programming to a greater or lesser degree, typically in combination with imperative, procedural programming. Significant object-oriented languages include C++, Objective-C, Smalltalk, Delphi, Java, C#, Perl, Python, Ruby and PHP.

Overview

Object-oriented programming attempts to provide a model for programming based on objects.^[3] Object-oriented programming integrates code and data using the concept of an "object". An object is an abstract data type with the addition of polymorphism and inheritance. An object has both state (data) and behavior (code).

Objects sometimes correspond to things found in the real world. For example, a graphics program may have objects such as "circle," "square," "menu." An online shopping system will have objects such as "shopping cart," "customer," and "product." The shopping system will support behaviors such as "place order," "make payment," and "offer discount."

Objects are designed in hierarchies, commonly a class hierarchy. For example, with the shopping system there might be high level classes such as "electronics product," "kitchen product," and "book." There may be further refinements for example under "electronic products": "CD Player," "DVD player," etc. These classes and subclasses correspond to sets and subsets in mathematical logic. Rather than utilizing database tables and programming subroutines, the developer utilizes objects the user may be more familiar with: objects from their application domain.^[4]

Object orientation uses encapsulation and information hiding. Object-orientation essentially merges abstract data types with structured programming and divides systems into modular objects which own their own data and are responsible for their own behavior. This feature is known as encapsulation. With encapsulation, the data for two objects are divided so that changes to one object cannot affect the other. Note that all this relies on the various languages being used appropriately, which, of course, is never certain. Object-orientation is not a software silver bullet.^[5]

The object-oriented approach encourages the programmer to place data where it is not directly accessible by the rest of the system. Instead, the data is accessed by calling specially written functions, called methods, which are bundled with the data. These act as the intermediaries for retrieving or modifying the data they control. The programming construct that combines data with a set of methods for accessing and managing that data is called an object. The practice of using subroutines to examine or modify certain kinds of data was also used in non-OOP modular programming, well before the widespread use of object-oriented programming.

Defining software as modular components that support inheritance is meant to make it easy both to re-use existing components and to extend components as needed by defining new subclasses with specialized behaviors. This goal of being easy to both maintain and reuse is known in the object-oriented paradigm as the "open/closed principle." A module is open if it supports extension (e.g. can easily modify behavior, add new properties, provide default values, etc.). A module is closed if it has a well defined stable interface that all other modules must use and that limits the interaction and potential errors that can be introduced into one module by changes in another.^[6]

History

Terminology invoking "objects" and "oriented" in the modern sense of object-oriented programming made its first appearance at MIT in the late 1950s and early 1960s. In the environment of the artificial intelligence group, as early as 1960, "object" could refer to identified items (LISP atoms) with properties (attributes);^[7][8] Alan Kay was later to cite a detailed understanding of LISP internals as a strong influence on his thinking in 1966.^[9] Another early MIT example was Sketchpad created by Ivan Sutherland in 1960–61; in the glossary of the 1963 technical report based on his dissertation about Sketchpad, Sutherland defined notions of "object" and "instance" (with the class concept covered by "master" or "definition"), albeit specialized to graphical interaction.^[10] Also, an MIT ALGOL version, AED-0, linked data structures ("plexes", in that dialect) directly with procedures, prefiguring what were later termed "messages", "methods", and "member functions".^[11][12]

The formal programming concept of objects was introduced in the 1960s in Simula 67, a major revision of Simula I, a programming language designed for discrete event simulation, created by Ole-Johan Dahl and Kristen Nygaard of the Norwegian Computing Center in Oslo.^[13] Simula 67 was influenced by SIMSCRIPT and C.A.R. "Tony" Hoare's proposed "record classes".^[11][14] Simula introduced the notion of classes and instances or objects (as well as subclasses, virtual methods, coroutines, and discrete event simulation) as part of an explicit programming paradigm. The language also used automatic garbage collection that had been invented earlier for the functional programming language Lisp. Simula was used for physical modeling, such as models to study and improve the movement of ships and their content through cargo ports. The ideas of Simula 67 influenced many later languages, including Smalltalk, derivatives of LISP (CLOS), Object Pascal, and C++.

The Smalltalk language, which was developed at Xerox PARC (by Alan Kay and others) in the 1970s, introduced the term object-oriented programming to represent the pervasive use of objects and messages as the basis for computation. Smalltalk creators were influenced by the ideas introduced in Simula 67, but Smalltalk was designed to be a fully dynamic system in which classes could be created and modified dynamically rather than statically as in Simula 67.^[15] Smalltalk and with it OOP were introduced to a wider audience by the August 1981 issue of Byte Magazine.

In the 1970s, Kay's Smalltalk work had influenced the Lisp community to incorporate object-based techniques that were introduced to developers via the Lisp machine. Experimentation with various extensions to Lisp (such as LOOPS and Flavors introducing multiple inheritance and mixins) eventually led to the Common Lisp Object System, which integrates functional programming and object-oriented programming and allows extension via a Meta-object protocol. In the 1980s, there were a few attempts to design processor architectures that included hardware support for objects in memory but these were not successful. Examples include the Intel iAPX 432 and the Linn Smart Rekursiv.

In 1985, Bertrand Meyer produced the first design of the Eiffel language. Focused on software quality, Eiffel is among the purely object-oriented languages, but differs in the sense that the language itself is not only a programming language, but a notation supporting the entire software lifecycle. Meyer described the Eiffel software development method, based on a small number of key ideas from software engineering and computer science, in Object-Oriented Software Construction. Essential to the quality focus of Eiffel is Meyer's reliability mechanism, Design by Contract, which is an integral part of both the method and language.

Object-oriented programming developed as the dominant programming methodology in the early and mid 1990s when programming languages supporting the techniques became widely available. These included Visual FoxPro 3.0,^[16][17][18] C++^{[citation needed]}, and Delphi^{[citation needed]}. Its dominance was further enhanced by the rising popularity of graphical user interfaces, which rely heavily upon object-oriented programming techniques. An example of a closely related dynamic GUI library and OOP language can be found in the Cocoa frameworks on Mac OS X, written in Objective-C, an object-oriented, dynamic messaging extension to C based on Smalltalk. OOP toolkits also enhanced the popularity of event-driven programming (although this concept is not limited to OOP).

At ETH Zürich, Niklaus Wirth and his colleagues had also been investigating such topics as data abstraction and modular programming (although this had been in common use in the 1960s or earlier). Modula-2 (1978) included both, and their succeeding design, Oberon, included a distinctive approach to object orientation, classes, and such.

Object-oriented features have been added to many previously existing languages, including Ada, BASIC, Fortran, Pascal, and COBOL. Adding these features to languages that were not initially designed for them often led to problems with compatibility and maintainability of code.

More recently, a number of languages have emerged that are primarily object-oriented, but that are also compatible with procedural methodology. Two such languages are Python and Ruby. Probably the most commercially-important recent object-oriented languages are Java, developed by Sun Microsystems, as well as C# and Visual Basic.NET (VB.NET), both designed for Microsoft's .NET platform. Each of these two frameworks shows, in its own way, the benefit of using OOP by creating an abstraction from implementation. VB.NET and C# support cross-language inheritance, allowing classes defined in one language to subclass classes defined in the other language. Developers usually compile Java to bytecode, allowing Java to run on any operating system for which a Java virtual machine is available.

Fundamental features and concepts

A survey by Deborah J. Armstrong of nearly 40 years of computing literature identified a number of fundamental concepts found in the large majority of definitions of OOP.^[19]
Not all of these concepts appear in all object-oriented programming languages. For example, object-oriented programming that uses classes is sometimes called class-based programming, while prototype-based programming does not typically use classes. As a result, a significantly different yet analogous terminology is used to define the concepts of object and instance.

Benjamin C. Pierce and some other researchers view any attempt to distill OOP to a minimal set of features as futile. He nonetheless identifies fundamental features that support the OOP programming style in most object-oriented languages:^[20]

• Dynamic dispatch – when a method is invoked on an object, the object itself determines what code gets executed by looking up the method at run time in a table associated with the object. This feature distinguishes an object from an abstract data type (or module), which has a fixed (static) implementation of the operations for all instances. It is a programming methodology that gives modular component development while at the same time being very efficient.
• Encapsulation (or multi-methods, in which case the state is kept separate)
• Subtype polymorphism
• Object inheritance (or delegation)
• Open recursion – a special variable (syntactically it may be a keyword), usually called this or self, that allows a method body to invoke another method body of the same object. This variable is late-bound; it allows a method defined in one class to invoke another method that is defined later, in some subclass thereof.

Similarly, in his 2003 book, Concepts in programming languages, John C. Mitchell identifies four main features: dynamic dispatch, abstraction, subtype polymorphism, and inheritance.^[21] Michael Lee Scott in Programming Language Pragmatics considers only encapsulation, inheritance, and dynamic dispatch.^[22]

Additional concepts used in object-oriented programming include:

• Classes of objects (object constructors)
• Instances of classes (objects, which have been constructed via a class)
• Methods which act on the attached objects.
• Message passing
• Abstraction

Decoupling

Decoupling refers to careful controls that separate code modules from particular use cases, which increases code re-usability. A common use of decoupling is to mainly decouple the encapsulation (see bridge pattern and adapter pattern) – for example, using a method interface that an encapsulated object must satisfy, as opposed to using the object's class.

Additional features

Encapsulation enforces modularity

Encapsulation refers to the creation of self-contained modules that bind processing functions to the data. These user-defined data types are called "classes", and one instance of a class is an "object". For example, in a payroll system, a class could be Manager, and Pat and Jan could be two instances (two objects) of the Manager class. Encapsulation ensures good code modularity, which keeps routines separate and less prone to conflict with each other.

Inheritance passes "knowledge" down

Classes are created in hierarchies, and inheritance lets the structure and methods in one class pass down the hierarchy. That means less programming is required when adding functions to complex systems. If a step is added at the bottom of a hierarchy, only the processing and data associated with that unique step must be added. Everything else above that step is inherited. The ability to reuse existing objects is considered a major advantage of object technology.

Polymorphism takes any shape

Object-oriented programming lets programmers create procedures for objects whose exact type is not known until runtime. For example, a screen cursor may change its shape from an arrow to a line depending on the program mode. The routine to move the cursor on screen in response to mouse movement can be written for "cursor", and polymorphism lets that cursor take simulating system behaviour.

Formal semantics

Objects are the run-time entities in an object-oriented system. They may represent a person, a place, a bank account, a table of data, or any item that the program has to handle.
There have been several attempts at formalizing the concepts used in object-oriented programming. The following concepts and constructs have been used as interpretations of OOP concepts:

• co algebraic data types^[23]
• abstract data types (which have existential types) allow the definition of modules but these do not support dynamic dispatch
• recursive types
• encapsulated state
• inheritance
• records are basis for understanding objects if function literals can be stored in fields (like in functional programming languages), but the actual calculi need be considerably more complex to incorporate essential features of OOP. Several extensions of System F_<: that deal with mutable objects have been studied;^[24] these allow both subtype polymorphism and parametric polymorphism (generics)

Attempts to find a consensus definition or theory behind objects have not proven very successful (however, see Abadi & Cardelli, A Theory of Objects^[24] for formal definitions of many OOP concepts and constructs), and often diverge widely. For example, some definitions focus on mental activities, and some on program structuring. One of the simpler definitions is that OOP is the act of using "map" data structures or arrays that can contain functions and pointers to other maps, all with some syntactic and scoping sugar on top. Inheritance can be performed by cloning the maps (sometimes called "prototyping").

OOP languages

Simula (1967) is generally accepted as the first language with the primary features of an object-oriented language. It was created for making simulation programs, in which what came to be called objects were the most important information representation. Smalltalk (1972 to 1980) is arguably the canonical example, and the one with which much of the theory of object-oriented programming was developed. Concerning the degree of object orientation, the following distinctions can be made:

• Languages called "pure" OO languages, because everything in them is treated consistently as an object, from primitives such as characters and punctuation, all the way up to whole classes, prototypes, blocks, modules, etc. They were designed specifically to facilitate, even enforce, OO methods. Examples: Eiffel, Emerald,^[25] JADE, Obix, Ruby, Scala, Smalltalk, Self.
• Languages designed mainly for OO programming, but with some procedural elements. Examples: Delphi/Object Pascal, C++, Java, C#, VB.NET.
• Languages that are historically procedural languages, but have been extended with some OO features. Examples: Pascal, Visual Basic (derived from BASIC), MATLAB, Fortran, Perl, COBOL 2002, PHP, ABAP, Ada 95.
• Languages with most of the features of objects (classes, methods, inheritance), but in a distinctly original form. Examples: Oberon (Oberon-1 or Oberon-2).
• Languages with abstract data type support, but not all features of object-orientation, sometimes called object-based languages. Examples: Modula-2, Pliant, CLU.
• Chameleon languages that support multiple paradigms, including OO. Tcl stands out among these for TclOO, a hybrid object system that supports both prototype-based programming and class-based OO.

OOP in dynamic languages

In recent years, object-oriented programming has become especially popular in dynamic programming languages. Python, Ruby and Groovy are dynamic languages built on OOP principles, while Perl and PHP have been adding object-oriented features since Perl 5 and PHP 4, and ColdFusion since version 6.

The Document Object Model of HTML, XHTML, and XML documents on the Internet has bindings to the popular JavaScript/ECMAScript language. JavaScript is perhaps the best known prototype-based programming language, which employs cloning from prototypes rather than inheriting from a class (contrast to class-based programming). Another scripting language that takes this approach is Lua. Before ActionScript 2.0 (a partial superset of the ECMA-262 R3, otherwise known as ECMAScript) only a prototype-based object model was supported.

Design patterns

Challenges of object-oriented design are addressed by several methodologies. Most common is known as the design patterns codified by Gamma et al.. More broadly, the term "design patterns" can be used to refer to any general, repeatable solution to a commonly occurring problem in software design. Some of these commonly occurring problems have implications and solutions particular to object-oriented development.

Inheritance and behavioral subtyping

It is intuitive to assume that inheritance creates a semantic "is a" relationship, and thus to infer that objects instantiated from subclasses can always be safely used instead of those instantiated from the superclass. This intuition is unfortunately false in most OOP languages, in particular in all those that allow mutable objects. Subtype polymorphism as enforced by the type checker in OOP languages (with mutable objects) cannot guarantee behavioral subtyping in any context. Behavioral subtyping is undecidable in general, so it cannot be implemented by a program (compiler). Class or object hierarchies must be carefully designed, considering possible incorrect uses that cannot be detected syntactically. This issue is known as the Liskov substitution principle.

Gang of Four design patterns

Design Patterns: Elements of Reusable Object-Oriented Software is an influential book published in 1995 by Erich Gamma, Richard Helm, Ralph Johnson, and John Vlissides, often referred to humorously as the "Gang of Four". Along with exploring the capabilities and pitfalls of object-oriented programming, it describes 23 common programming problems and patterns for solving them. As of April 2007, the book was in its 36th printing.

The book describes the following patterns:

• Creational patterns (5): Factory method pattern, Abstract factory pattern, Singleton pattern, Builder pattern, Prototype pattern
• Structural patterns (7): Adapter pattern, Bridge pattern, Composite pattern, Decorator pattern, Facade pattern, Flyweight pattern, Proxy pattern
• Behavioral patterns (11): Chain-of-responsibility pattern, Command pattern, Interpreter pattern, Iterator pattern, Mediator pattern, Memento pattern, Observer pattern, State pattern, Strategy pattern, Template method pattern, Visitor pattern

Object-orientation and databases

Both object-oriented programming and relational database management systems (RDBMSs) are extremely common in software today^[update]. Since relational databases don't store objects directly (though some RDBMSs have object-oriented features to approximate this), there is a general need to bridge the two worlds. The problem of bridging object-oriented programming accesses and data patterns with relational databases is known as object-relational impedance mismatch. There are a number of approaches to cope with this problem, but no general solution without downsides.^[26] One of the most common approaches is object-relational mapping, as found in IDE languages such as Visual FoxPro and libraries such as Java Data Objects and Ruby on Rails' ActiveRecord.
There are also object databases that can be used to replace RDBMSs, but these have not been as technically and commercially successful as RDBMSs.

Real-world modeling and relationships

OOP can be used to associate real-world objects and processes with digital counterparts. However, not everyone agrees that OOP facilitates direct real-world mapping (see Criticism section) or that real-world mapping is even a worthy goal; Bertrand Meyer argues in Object-Oriented Software Construction^[27] that a program is not a model of the world but a model of some part of the world; "Reality is a cousin twice removed". At the same time, some principal limitations of OOP had been noted.^[28] For example, the circle-ellipse problem is difficult to handle using OOP's concept of inheritance.

However, Niklaus Wirth (who popularized the adage now known as Wirth's law: "Software is getting slower more rapidly than hardware becomes faster") said of OOP in his paper, "Good Ideas through the Looking Glass", "This paradigm closely reflects the structure of systems 'in the real world', and it is therefore well suited to model complex systems with complex behaviours" (contrast KISS principle).

Steve Yegge and others noted that natural languages lack the OOP approach of strictly prioritizing things (objects/nouns) before actions (methods/verbs).^[29] This problem may cause OOP to suffer more convoluted solutions than procedural programming.^[30]

OOP and control flow

OOP was developed to increase the reusability and maintainability of source code.^[31] Transparent representation of the control flow had no priority and was meant to be handled by a compiler. With the increasing relevance of parallel hardware and multithreaded coding, developing transparent control flow becomes more important, something hard to achieve with OOP.^{[32][33][34][35]}

Responsibility- vs. data-driven design

Responsibility-driven design defines classes in terms of a contract, that is, a class should be defined around a responsibility and the information that it shares. This is contrasted by Wirfs-Brock and Wilkerson with data-driven design, where classes are defined around the data-structures that must be held. The authors hold that responsibility-driven design is preferable.

Criticism

The OOP paradigm has been criticised for a number of reasons, including not meeting its stated goals of reusability and modularity,^[36][37] and for overemphasizing one aspect of software design and modeling (data/objects) at the expense of other important aspects (computation/algorithms).^[38][39]
Luca Cardelli has claimed that OOP code is "intrinsically less efficient" than procedural code, that OOP can take longer to compile, and that OOP languages have "extremely poor modularity properties with respect to class extension and modification", and tend to be extremely complex.^[36]
The latter point is reiterated by Joe Armstrong, the principal inventor of Erlang, who is quoted as saying:^[37]

The problem with object-oriented languages is they've got all this implicit environment that they carry around with them. You wanted a banana but what you got was a gorilla holding the banana and the entire jungle.

A study by Potok et al. has shown no significant difference in productivity between OOP and procedural approaches.^[40]

Christopher J. Date stated that critical comparison of OOP to other technologies, relational in particular, is difficult because of lack of an agreed-upon and rigorous definition of OOP;^[41] however, Date and Darwen have proposed a theoretical foundation on OOP that uses OOP as a kind of customizable type system to support RDBMS.^[42]

In an article Lawrence Krubner claimed that compared to other languages (lisps, functional languages, etc) OOP languages have no unique strengths, and inflict a heavy burden of unneeded complexity.^[43]

Alexander Stepanov compares object orientation unfavourably to multimethods:^[38]

I find OOP technically unsound. It attempts to decompose the world in terms of interfaces that vary on a single type. To deal with the real problems you need multisorted algebras — families of interfaces that span multiple types. I find OOP philosophically unsound. It claims that everything is an object. Even if it is true it is not very interesting — saying that everything is an object is saying nothing at all.

Paul Graham has suggested that OOP's popularity within large companies is due to "large (and frequently changing) groups of mediocre programmers." According to Graham, the discipline imposed by OOP prevents any one programmer from "doing too much damage."^[44]
Steve Yegge noted that, as opposed to functional programming:^[45]

Object Oriented Programming puts the Nouns first and foremost. Why would you go to such lengths to put one part of speech on a pedestal? Why should one kind of concept take precedence over another? It's not as if OOP has suddenly made verbs less important in the way we actually think. It's a strangely skewed perspective.

Rich Hickey, creator of Clojure, described object systems as overly simplistic models of the real world. He emphasized the inability of OOP to model time properly, which is getting increasingly problematic as software systems become more concurrent.^[39]

Eric S. Raymond, a Unix programmer and open-source software advocate, has been critical of claims that present object-oriented programming as the "One True Solution," and has written that object-oriented programming languages tend to encourage thickly-layered programs that destroy transparency.^[46] Raymond contrasts this to the approach taken with Unix and the C programming language.^[47]

XML

XML Filename extension .xml Internet media type application/xml text/xml[1] Uniform Type Identifier (UTI) public.xml UTI conformation public.text Developed by World Wide Web Consortium Type of format Markup language Extended from SGML Extended to Numerous languages, including XHTML RSS Atom KML Standard 1.0 (Fifth Edition) (November 26, 2008; 6 years ago (2008-11-26)) 1.1 (Second Edition) (August 16, 2006; 8 years ago (2006-08-16)) Open format? Yes

Extensible Markup Language (XML) Status Published Year started 1996 Editors Tim Bray Jean Paoli C. M. Sperberg-McQueen Eve Maler François Yergeau John Cowan Related standards XML Schema Domain Data serialization Abbreviation XML Website XML 1.0

Extensible Markup Language (XML) is a markup language that defines a set of rules for encoding documents in a format which is both human-readable and machine-readable. It is defined by the W3C's XML 1.0 Specification^[2] and by several other related specifications,^[3] all of which are free open standards.^[4]

The design goals of XML emphasize simplicity, generality and usability across the Internet.^[5] It is a textual data format with strong support via Unicode for different human languages. Although the design of XML focuses on documents, it is widely used for the representation of arbitrary data structures^[6] such as those used in web services.

Several schema systems exist to aid in the definition of XML-based languages, while many application programming interfaces (APIs) have been developed to aid the processing of XML data.

Applications of XML

As of 2009^[update], hundreds of document formats using XML syntax have been developed,^[7] including RSS, Atom, SOAP, and XHTML. XML-based formats have become the default for many office-productivity tools, including Microsoft Office (Office Open XML), OpenOffice.org and LibreOffice (OpenDocument), and Apple's iWork. XML has also been employed as the base language for communication protocols, such as XMPP. Applications for the Microsoft .NET Framework use XML files for configuration. Apple has an implementation of a registry based on XML.^[8]

XML has come into common use for the interchange of data over the Internet. IETF RFC 7303 gives rules for the construction of Internet Media Types for use when sending XML. It also defines the media types application/xml and text/xml, which say only that the data are in XML, and nothing about its semantics. The use of text/xml has been criticized as a potential source of encoding problems and it has been suggested that it should be deprecated.^[9]

RFC 7303 also recommends that XML-based languages be given media types ending in +xml; for example image/svg+xml for SVG.
Further guidelines for the use of XML in a networked context may be found in RFC 3470, also known as IETF BCP 70 a document which covers many aspects of designing and deploying an XML-based language.

Key terminology

The material in this section is based on the XML Specification. This is not an exhaustive list of all the constructs that appear in XML; it provides an introduction to the key constructs most often encountered in day-to-day use.

(Unicode) character

By definition, an XML document is a string of characters. Almost every legal Unicode character may appear in an XML document.

Processor and application

The processor analyzes the markup and passes structured information to an application. The specification places requirements on what an XML processor must do and not do, but the application is outside its scope. The processor (as the specification calls it) is often referred to colloquially as an XML parser.

Markup and content

The characters making up an XML document are divided into markup and content, which may be distinguished by the application of simple syntactic rules. Generally, strings that constitute markup either begin with the character < and end with a >, or they begin with the character & and end with a ;. Strings of characters that are not markup are content. However, in a CDATA section, the delimiters <![CDATA[ and ]]> are classified as markup, while the text between them is classified as content. In addition, whitespace before and after the outermost element is classified as markup.

Tag

A markup construct that begins with < and ends with >. Tags come in three flavors:

• start-tags; for example: <section>
• end-tags; for example: </section>
• empty-element tags; for example: <line-break />

Element

A logical document component which either begins with a start-tag and ends with a matching end-tag or consists only of an empty-element tag. The characters between the start- and end-tags, if any, are the element's content, and may contain markup, including other elements, which are called child elements. An example of an element is <Greeting>Hello, world.</Greeting> (see hello world). Another is <line-break />.

Attribute

A markup construct consisting of a name/value pair that exists within a start-tag or empty-element tag. In the example (below) the element img has two attributes, src and alt:

<img src="madonna.jpg" alt='Foligno Madonna, by Raphael' />

Another example would be

<step number="3">Connect A to B.</step>

where the name of the attribute is "number" and the value is "3".

An XML attribute can only have a single value and each attribute can appear at most once on each element. In the common situation where a list of multiple values is desired, this must be done by encoding the list into a well-formed XML attribute^{[note 1]} with some format beyond what XML defines itself. Usually this is either a comma or semi-colon delimited list or, if the individual values are known not to contain spaces,^{[note 2]} a space-delimited list can be used.

<div class="inner greeting-box" >Hello!</div>

where the attribute "class" has both the value "inner greeting-box" and also indicates the two CSS class names "inner" and "greeting-box".

XML declaration

XML documents may begin by declaring some information about themselves, as in the following example:

<?xml version="1.0" encoding="UTF-8"?>

Characters and escaping

XML documents consist entirely of characters from the Unicode repertoire. Except for a small number of specifically excluded control characters, any character defined by Unicode may appear within the content of an XML document.

XML includes facilities for identifying the encoding of the Unicode characters that make up the document, and for expressing characters that, for one reason or another, cannot be used directly.

Valid characters

Unicode code points in the following ranges are valid in XML 1.0 documents:^[10]

• U+0009, U+000A, U+000D: these are the only C0 controls accepted in XML 1.0;
• U+0020–U+D7FF, U+E000–U+FFFD: this excludes some (not all) non-characters in the BMP (all surrogates, U+FFFE and U+FFFF are forbidden);
• U+10000–U+10FFFF: this includes all code points in supplementary planes, including non-characters.

XML 1.1^[11] extends the set of allowed characters to include all the above, plus the remaining characters in the range U+0001–U+001F. At the same time, however, it restricts the use of C0 and C1 control characters other than U+0009, U+000A, U+000D, and U+0085 by requiring them to be written in escaped form (for example U+0001 must be written as &#x01; or its equivalent). In the case of C1 characters, this restriction is a backwards incompatibility; it was introduced to allow common encoding errors to be detected.

The code point U+0000 is the only character that is not permitted in any XML 1.0 or 1.1 document.

Encoding detection

The Unicode character set can be encoded into bytes for storage or transmission in a variety of different ways, called "encodings". Unicode itself defines encodings that cover the entire repertoire; well-known ones include UTF-8 and UTF-16.^[12] There are many other text encodings that predate Unicode, such as ASCII and ISO/IEC 8859; their character repertoires in almost every case are subsets of the Unicode character set.

XML allows the use of any of the Unicode-defined encodings, and any other encodings whose characters also appear in Unicode. XML also provides a mechanism whereby an XML processor can reliably, without any prior knowledge, determine which encoding is being used.^[13] Encodings other than UTF-8 and UTF-16 will not necessarily be recognized by every XML parser.

Escaping

XML provides escape facilities for including characters which are problematic to include directly. For example:

• The characters "<" and "&" are key syntax markers and may never appear in content outside a CDATA section.^[14]
• Some character encodings support only a subset of Unicode. For example, it is legal to encode an XML document in ASCII, but ASCII lacks code points for Unicode characters such as "é".
• It might not be possible to type the character on the author's machine.
• Some characters have glyphs that cannot be visually distinguished from other characters: examples are
  • non-breaking space ( ) " "

    compare space ( ) " "

  • Cyrillic Capital Letter A (А) "А"

    compare Latin Capital Letter A (A) "A"

There are five predefined entities:

• < represents "<"
• > represents ">"
• & represents "&"
• ' represents '
• " represents "

All permitted Unicode characters may be represented with a numeric character reference. Consider the Chinese character "中", whose numeric code in Unicode is hexadecimal 4E2D, or decimal 20,013. A user whose keyboard offers no method for entering this character could still insert it in an XML document encoded either as 中 or 中. Similarly, the string "I <3 Jörg" could be encoded for inclusion in an XML document as "I <3 Jörg".
"�" is not permitted, however, because the null character is one of the control characters excluded from XML, even when using a numeric character reference.^[15] An alternative encoding mechanism such as Base64 is needed to represent such characters.

Comments may appear anywhere in a document outside other markup. Comments cannot appear before the XML declaration. Comments start with "<!--" and end with "-->". For compatibility with SGML, the string "--" (double-hyphen) is not allowed inside comments;^[16] this means comments cannot be nested. The ampersand has no special significance within comments, so entity and character references are not recognized as such, and there is no way to represent characters outside the character set of the document encoding.

An example of a valid comment: "<!--no need to escape <code> & such in comments-->"

International use

XML 1.0 (Fifth Edition) and XML 1.1 support the direct use of almost any Unicode character in element names, attributes, comments, character data, and processing instructions (other than the ones that have special symbolic meaning in XML itself, such as the less-than sign, "<"). The following is a well-formed XML document including Chinese, Armenian and Cyrillic characters:

<?xml version="1.0" encoding="UTF-8"?>
<俄语 լեզու="ռուսերեն">данные</俄语>

Well-formedness and error-handling

The XML specification defines an XML document as a well-formed text – meaning that it satisfies a list of syntax rules provided in the specification. Some key points in the fairly lengthy list include:

• The document contains only properly encoded legal Unicode characters
• None of the special syntax characters such as < and & appear except when performing their markup-delineation roles
• The begin, end, and empty-element tags that delimit the elements are correctly nested, with none missing and none overlapping
• The element tags are case-sensitive; the beginning and end tags must match exactly.
• Tag names cannot contain any of the characters !"#$%&'()*+,/;<=>?@[\]^`{|}~, nor a space character, and cannot start with -, ., or a numeric digit.
• A single "root" element contains all the other elements.

The definition of an XML document excludes texts that contain violations of well-formedness rules; they are simply not XML. An XML processor that encounters such a violation is required to report such errors and to cease normal processing. This policy, occasionally referred to as "draconian error handling," stands in notable contrast to the behavior of programs that process HTML, which are designed to produce a reasonable result even in the presence of severe markup errors.^[17] XML's policy in this area has been criticized as a violation of Postel's law ("Be conservative in what you send; be liberal in what you accept").^[18]

The XML specification defines a valid XML document as a well-formed XML document which also conforms to the rules of a Document Type Definition (DTD).

Schemas and validation

In addition to being well-formed, an XML document may be valid. This means that it contains a reference to a Document Type Definition (DTD), and that its elements and attributes are declared in that DTD and follow the grammatical rules for them that the DTD specifies.

XML processors are classified as validating or non-validating depending on whether or not they check XML documents for validity. A processor that discovers a validity error must be able to report it, but may continue normal processing.

A DTD is an example of a schema or grammar. Since the initial publication of XML 1.0, there has been substantial work in the area of schema languages for XML. Such schema languages typically constrain the set of elements that may be used in a document, which attributes may be applied to them, the order in which they may appear, and the allowable parent/child relationships.

Document Type Definition

The oldest schema language for XML is the Document Type Definition (DTD), inherited from SGML.
DTDs have the following benefits:

• DTD support is ubiquitous due to its inclusion in the XML 1.0 standard.
• DTDs are terse compared to element-based schema languages and consequently present more information in a single screen.
• DTDs allow the declaration of standard public entity sets for publishing characters.
• DTDs define a document type rather than the types used by a namespace, thus grouping all constraints for a document in a single collection.

DTDs have the following limitations:

• They have no explicit support for newer features of XML, most importantly namespaces.
• They lack expressiveness. XML DTDs are simpler than SGML DTDs and there are certain structures that cannot be expressed with regular grammars. DTDs only support rudimentary datatypes.
• They lack readability. DTD designers typically make heavy use of parameter entities (which behave essentially as textual macros), which make it easier to define complex grammars, but at the expense of clarity.
• They use a syntax based on regular expression syntax, inherited from SGML, to describe the schema. Typical XML APIs such as SAX do not attempt to offer applications a structured representation of the syntax, so it is less accessible to programmers than an element-based syntax may be.

Two peculiar features that distinguish DTDs from other schema types are the syntactic support for embedding a DTD within XML documents and for defining entities, which are arbitrary fragments of text and/or markup that the XML processor inserts in the DTD itself and in the XML document wherever they are referenced, like character escapes.

DTD technology is still used in many applications because of its ubiquity.

XML Schema

A newer schema language, described by the W3C as the successor of DTDs, is XML Schema, often referred to by the initialism for XML Schema instances, XSD (XML Schema Definition). XSDs are far more powerful than DTDs in describing XML languages. They use a rich datatyping system and allow for more detailed constraints on an XML document's logical structure. XSDs also use an XML-based format, which makes it possible to use ordinary XML tools to help process them.xs:schema element that defines a schema:

<?xml version="1.0" encoding="ISO-8859-1" ?>
<xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema">
</xs:schema>

RELAX NG

RELAX NG was initially specified by OASIS and is now also an ISO/IEC International Standard (as part of DSDL). RELAX NG schemas may be written in either an XML based syntax or a more compact non-XML syntax; the two syntaxes are isomorphic and James Clark's conversion tool - 'Trang', can convert between them without loss of information. RELAX NG has a simpler definition and validation framework than XML Schema, making it easier to use and implement. It also has the ability to use datatype framework plug-ins; a RELAX NG schema author, for example, can require values in an XML document to conform to definitions in XML Schema Datatypes.

Schematron

Schematron is a language for making assertions about the presence or absence of patterns in an XML document. It typically uses XPath expressions.

ISO DSDL and other schema languages

The ISO DSDL (Document Schema Description Languages) standard brings together a comprehensive set of small schema languages, each targeted at specific problems. DSDL includes RELAX NG full and compact syntax, Schematron assertion language, and languages for defining datatypes, character repertoire constraints, renaming and entity expansion, and namespace-based routing of document fragments to different validators. DSDL schema languages do not have the vendor support of XML Schemas yet, and are to some extent a grassroots reaction of industrial publishers to the lack of utility of XML Schemas for publishing.

Some schema languages not only describe the structure of a particular XML format but also offer limited facilities to influence processing of individual XML files that conform to this format. DTDs and XSDs both have this ability; they can for instance provide the infoset augmentation facility and attribute defaults. RELAX NG and Schematron intentionally do not provide these.

Related specifications

A cluster of specifications closely related to XML have been developed, starting soon after the initial publication of XML 1.0. It is frequently the case that the term "XML" is used to refer to XML
together with one or more of these other technologies which have come to be seen as part of the XML core.

• XML Namespaces enable the same document to contain XML elements and attributes taken from different vocabularies, without any naming collisions occurring. Although XML Namespaces are not part of the XML specification itself, virtually all XML software also supports XML Namespaces.
• XML Base defines the xml:base attribute, which may be used to set the base for resolution of relative URI references within the scope of a single XML element.
• The XML Information Set or XML infoset describes an abstract data model for XML documents in terms of information items. The infoset is commonly used in the specifications of XML languages, for convenience in describing constraints on the XML constructs those languages allow.
• xml:id Version 1.0 asserts that an attribute named xml:id functions as an "ID attribute" in the sense used in a DTD.
• XPath defines a syntax named XPath expressions which identifies one or more of the internal components (elements, attributes, and so on) included in an XML document. XPath is widely used in other core-XML specifications and in programming libraries for accessing XML-encoded data.
• XSLT is a language with an XML-based syntax that is used to transform XML documents into other XML documents, HTML, or other, unstructured formats such as plain text or RTF. XSLT is very tightly coupled with XPath, which it uses to address components of the input XML document, mainly elements and attributes.
• XSL Formatting Objects, or XSL-FO, is a markup language for XML document formatting which is most often used to generate PDFs.
• XQuery is an XML-oriented query language strongly rooted in XPath and XML Schema. It provides methods to access, manipulate and return XML, and is mainly conceived as a query language for XML databases.
• XML Signature defines syntax and processing rules for creating digital signatures on XML content.
• XML Encryption defines syntax and processing rules for encrypting XML content.

Some other specifications conceived as part of the "XML Core" have failed to find wide adoption, including XInclude, XLink, and XPointer.

Programming interfaces

The design goals of XML include, "It shall be easy to write programs which process XML documents."^[5] Despite this, the XML specification contains almost no information about how programmers might go about doing such processing. The XML Infoset specification provides a vocabulary to refer to the constructs within an XML document, but also does not provide any guidance on how to access this information. A variety of APIs for accessing XML have been developed and used, and some have been standardized.

Existing APIs for XML processing tend to fall into these categories:

• Stream-oriented APIs accessible from a programming language, for example SAX and StAX.
• Tree-traversal APIs accessible from a programming language, for example DOM.
• XML data binding, which provides an automated translation between an XML document and programming-language objects.
• Declarative transformation languages such as XSLT and XQuery.

Stream-oriented facilities require less memory and, for certain tasks which are based on a linear traversal of an XML document, are faster and simpler than other alternatives. Tree-traversal and data-binding APIs typically require the use of much more memory, but are often found more convenient for use by programmers; some include declarative retrieval of document components via the use of XPath expressions.

XSLT is designed for declarative description of XML document transformations, and has been widely implemented both in server-side packages and Web browsers. XQuery overlaps XSLT in its functionality, but is designed more for searching of large XML databases.

Simple API for XML

Simple API for XML (SAX) is a lexical, event-driven interface in which a document is read serially and its contents are reported as callbacks to various methods on a handler object of the user's design. SAX is fast and efficient to implement, but difficult to use for extracting information at random from the XML, since it tends to burden the application author with keeping track of what part of the document is being processed. It is better suited to situations in which certain types of information are always handled the same way, no matter where they occur in the document.

Pull parsing

Pull parsing^[19] treats the document as a series of items which are read in sequence using the Iterator design pattern. This allows for writing of recursive-descent parsers in which the structure of the code performing the parsing mirrors the structure of the XML being parsed, and intermediate parsed results can be used and accessed as local variables within the methods performing the parsing, or passed down (as method parameters) into lower-level methods, or returned (as method return values) to higher-level methods. Examples of pull parsers include StAX in the Java programming language, XMLPullParser in Smalltalk, XMLReader in PHP, ElementTree.iterparse in Python, System.Xml.XmlReader in the .NET Framework, and the DOM traversal API (NodeIterator and TreeWalker).

A pull parser creates an iterator that sequentially visits the various elements, attributes, and data in an XML document. Code which uses this iterator can test the current item (to tell, for example, whether it is a start or end element, or text), and inspect its attributes (local name, namespace, values of XML attributes, value of text, etc.), and can also move the iterator to the next item. The code can thus extract information from the document as it traverses it. The recursive-descent approach tends to lend itself to keeping data as typed local variables in the code doing the parsing, while SAX, for instance, typically requires a parser to manually maintain intermediate data within a stack of elements which are parent elements of the element being parsed. Pull-parsing code can be more straightforward to understand and maintain than SAX parsing code.

Document Object Model

The Document Object Model (DOM) is an interface-oriented application programming interface that allows for navigation of the entire document as if it were a tree of node objects representing the document's contents. A DOM document can be created by a parser, or can be generated manually by users (with limitations). Data types in DOM nodes are abstract; implementations provide their own programming language-specific bindings. DOM implementations tend to be memory intensive, as they generally require the entire document to be loaded into memory and constructed as a tree of objects before access is allowed.

Data binding

Another form of XML processing API is XML data binding, where XML data are made available as a hierarchy of custom, strongly typed classes, in contrast to the generic objects created by a Document Object Model parser. This approach simplifies code development, and in many cases allows problems to be identified at compile time rather than run-time. Example data binding systems include the Java Architecture for XML Binding (JAXB) and XML Serialization in .NET.^[20]

XML as data type

XML has appeared as a first-class data type in other languages. The ECMAScript for XML (E4X) extension to the ECMAScript/JavaScript language explicitly defines two specific objects (XML and XMLList) for JavaScript, which support XML document nodes and XML node lists as distinct objects and use a dot-notation specifying parent-child relationships.^[21] E4X is supported by the Mozilla 2.5+ browsers (though now deprecated) and Adobe Actionscript, but has not been adopted more universally. Similar notations are used in Microsoft's LINQ implementation for Microsoft .NET 3.5 and above, and in Scala (which uses the Java VM). The open-source xmlsh application, which provides a Linux-like shell with special features for XML manipulation, similarly treats XML as a data type, using the <[ ]> notation.^[22] The Resource Description Framework defines a data type rdf:XMLLiteral to hold wrapped, canonical XML.^[23]

History

XML is an application profile of SGML (ISO 8879).^[24]

The versatility of SGML for dynamic information display was understood by early digital media publishers in the late 1980s prior to the rise of the Internet.^[25][26] By the mid-1990s some practitioners of SGML had gained experience with the then-new World Wide Web, and believed that SGML offered solutions to some of the problems the Web was likely to face as it grew. Dan Connolly added SGML to the list of W3C's activities when he joined the staff in 1995; work began in mid-1996 when Sun Microsystems engineer Jon Bosak developed a charter and recruited collaborators. Bosak was well connected in the small community of people who had experience both in SGML and the Web.^[27]

XML was compiled by a working group of eleven members,^[28] supported by a (roughly) 150-member Interest Group. Technical debate took place on the Interest Group mailing list and issues were resolved by consensus or, when that failed, majority vote of the Working Group. A record of design decisions and their rationales was compiled by Michael Sperberg-McQueen on December 4, 1997.^[29] James Clark served as Technical Lead of the Working Group, notably contributing the empty-element "<empty />" syntax and the name "XML". Other names that had been put forward for consideration included "MAGMA" (Minimal Architecture for Generalized Markup Applications), "SLIM" (Structured Language for Internet Markup) and "MGML" (Minimal Generalized Markup Language). The co-editors of the specification were originally Tim Bray and Michael Sperberg-McQueen. Halfway through the project Bray accepted a consulting engagement with Netscape, provoking vociferous protests from Microsoft. Bray was temporarily asked to resign the editorship. This led to intense dispute in the Working Group, eventually solved by the appointment of Microsoft's Jean Paoli as a third co-editor.

The XML Working Group never met face-to-face; the design was accomplished using a combination of email and weekly teleconferences. The major design decisions were reached in a short burst of intense work between August and November 1996,^[30] when the first Working Draft of an XML specification was published.^[31] Further design work continued through 1997, and XML 1.0 became a W3C Recommendation on February 10, 1998.

Sources

XML is a profile of an ISO standard SGML, and most of XML comes from SGML unchanged. From SGML comes the separation of logical and physical structures (elements and entities), the availability of grammar-based validation (DTDs), the separation of data and metadata (elements and attributes), mixed content, the separation of processing from representation (processing instructions), and the default angle-bracket syntax. Removed were the SGML declaration (XML has a fixed delimiter set and adopts Unicode as the document character set).

Other sources of technology for XML were the Text Encoding Initiative (TEI), which defined a profile of SGML for use as a "transfer syntax"; and HTML, in which elements were synchronous with their resource, document character sets were separate from resource encoding, the xml:lang attribute was invented, and (like HTTP) metadata accompanied the resource rather than being needed at the declaration of a link. The Extended Reference Concrete Syntax (ERCS) project of the SPREAD (Standardization Project Regarding East Asian Documents) project of the ISO-related China/Japan/Korea Document Processing expert group was the basis of XML 1.0's naming rules; SPREAD also introduced hexadecimal numeric character references and the concept of references to make available all Unicode characters. To support ERCS, XML and HTML better, the SGML standard IS 8879 was revised in 1996 and 1998 with WebSGML Adaptations. The XML header followed that of ISO HyTime.

Ideas that developed during discussion which were novel in XML included the algorithm for encoding detection and the encoding header, the processing instruction target, the xml:space attribute, and the new close delimiter for empty-element tags. The notion of well-formedness as opposed to validity (which enables parsing without a schema) was first formalized in XML, although it had been implemented successfully in the Electronic Book Technology "Dynatext" software;^[32] the software from the University of Waterloo New Oxford English Dictionary Project; the RISP LISP SGML text processor at Uniscope, Tokyo; the US Army Missile Command IADS hypertext system; Mentor Graphics Context; Interleaf and Xerox Publishing System.

Versions

There are two current versions of XML. The first (XML 1.0) was initially defined in 1998. It has undergone minor revisions since then, without being given a new version number, and is currently in its fifth edition, as published on November 26, 2008. It is widely implemented and still recommended for general use.

The second (XML 1.1) was initially published on February 4, 2004, the same day as XML 1.0 Third Edition,^[33] and is currently in its second edition, as published on August 16, 2006. It contains features (some contentious) that are intended to make XML easier to use in certain cases.^[34] The main changes are to enable the use of line-ending characters used on EBCDIC platforms, and the use of scripts and characters absent from Unicode 3.2. XML 1.1 is not very widely implemented and is recommended for use only by those who need its unique features.^[35]

Prior to its fifth edition release, XML 1.0 differed from XML 1.1 in having stricter requirements for characters available for use in element and attribute names and unique identifiers: in the first four editions of XML 1.0 the characters were exclusively enumerated using a specific version of the Unicode standard (Unicode 2.0 to Unicode 3.2.) The fifth edition substitutes the mechanism of XML 1.1, which is more future-proof but reduces redundancy. The approach taken in the fifth edition of XML 1.0 and in all editions of XML 1.1 is that only certain characters are forbidden in names, and everything else is allowed, in order to accommodate the use of suitable name characters in future versions of Unicode. In the fifth edition, XML names may contain characters in the Balinese, Cham, or Phoenician scripts among many others which have been added to Unicode since Unicode 3.2.^[34]

Almost any Unicode code point can be used in the character data and attribute values of an XML 1.0 or 1.1 document, even if the character corresponding to the code point is not defined in the current version of Unicode. In character data and attribute values, XML 1.1 allows the use of more control characters than XML 1.0, but, for "robustness", most of the control characters introduced in XML 1.1 must be expressed as numeric character references (and #x7F through #x9F, which had been allowed in XML 1.0, are in XML 1.1 even required to be expressed as numeric character references^[36]). Among the supported control characters in XML 1.1 are two line break codes that must be treated as whitespace. Whitespace characters are the only control codes that can be written directly.

There has been discussion of an XML 2.0, although no organization has announced plans for work on such a project. XML-SW (SW for skunkworks), written by one of the original developers of XML,^[37] contains some proposals for what an XML 2.0 might look like: elimination of DTDs from syntax, integration of namespaces, XML Base and XML Information Set (infoset) into the base standard.

The World Wide Web Consortium also has an XML Binary Characterization Working Group doing preliminary research into use cases and properties for a binary encoding of the XML infoset. The working group is not chartered to produce any official standards. Since XML is by definition text-based, ITU-T and ISO are using the name Fast Infoset for their own binary infoset to avoid confusion (see ITU-T Rec. X.891 | ISO/IEC 24824-1).

Criticism

XML and its extensions have regularly been criticized for verbosity and complexity.^[38] Mapping the basic tree model of XML to type systems of programming languages or databases can be difficult, especially when XML is used for exchanging highly structured data between applications, which was not its primary design goal. Other criticisms attempt to refute the claim that XML is a self-describing language^[39] (though the XML specification itself makes no such claim). JSON, YAML, and S-Expressions are frequently proposed as alternatives (see Comparison of data serialization formats);^[40] which focus on representing highly structured data rather than documents, which may contain both highly structured and relatively unstructured content.