by Brad Appleton <brad@bradapp.net>
http://www.bradapp.net/
last
modified 02/14/2000
Fundamental to any science or engineering discipline is a common vocabulary for expressing its concepts, and a language for relating them together. The goal of patterns within the software community is to create a body of literature to help software developers resolve recurring problems encountered throughout all of software development. Patterns help create a shared language for communicating insight and experience about these problems and their solutions. Formally codifying these solutions and their relationships lets us successfully capture the body of knowledge which defines our understanding of good architectures that meet the needs of their users. Forming a common pattern language for conveying the structures and mechanisms of our architectures allows us to intelligibly reason about them. The primary focus is not so much on technology as it is on creating a culture to document and support sound engineering architecture and design.
The current use of the term "pattern" is derived from the writings of the architect Christopher Alexander who has written several books on the topic as it relates to urban planning and building architecture:
Although these books are ostensibly about architecture and urban planning, they are applicable to many other disciplines, including software development. In [Notes], Alexander argues that current architectural methods result in products that fail to meet the real demands and requirements of its users, society and its individuals, and are unsuccessful in fulfilling the quintessential purpose of all design and engineering endeavors: to improve the human condition. Alexander wanted to create structures that are good for people and have a positive influence on them by improving their comfort and their quality of life. He concluded that architects must constantly strive to produce work products that better fit and adapt to the needs of all their inhabitants and users and their respective communities. In [APL] Alexander describes some "timeless" design ideas to try and realize these goals. In [TTWoB] Alexander proposes a paradigm for architecture based on three concepts: the quality, the gate, and the way:
In 1987, Ward Cunningham and Kent Beck were working with Smalltalk and designing user interfaces. They decided to use some of Alexander's ideas to develop a small five pattern language for guiding novice Smalltalk programmers. They wrote up the results and presented them at OOPSLA'87 in Orlando in the paper "Using Pattern Languages for Object-Oriented Programs".
Soon afterward, Jim Coplien (more affectionately referred to as "Cope") began compiling a catalog of C++ idioms (which are one kind of pattern) and later published them as a book in 1991, Advanced C++ Programming Styles and Idioms.
From 1990 to 1992, various members of the Gang of Four had met one another and had done some work compiling a catalog of patterns. Discussions of patterns abounded at OOPSLA'91 at a workshop given by Bruce Andersen (which was repeated in 1992). Many pattern notables participated in these workshops, including Jim Coplien, Doug Lea, Desmond D'Souza, Norm Kerth, Wolfgang Pree, and others.
In August 1993, Kent Beck and Grady Booch sponsored a mountain retreat in Colorado, the first meeting of what is now known as the Hillside Group. Another patterns workshop was held at OOPSLA'93 and then in April of 1994, the Hillside Group met again (this time with Richard Gabriel added to the fold) to plan the first PLoP conference.
Shortly thereafter, the [GoF] Design Patterns book was published, and the rest, is history.
A pattern is the abstraction from a concrete form which keeps recurring in specific non-arbitrary contexts.
The above authors point out that, within the software patterns community, the notion of a pattern is "geared toward solving problems in design." More specifically, the concrete form which recurs is that of a solution to a recurring problem. But a pattern is more than just a battle-proven solution to a recurring problem. The problem occurs within a certain context, and in the presence of numerous competing concerns. The proposed solution involves some kind of structure which balances these concerns, or "forces", in the manner most appropriate for the given context. Using the pattern form, the description of the solution tries to capture the essential insight which it embodies, so that others may learn from it, and make use of it in similar situations. The pattern is also given a name, which serves as a conceptual handle, to facilitate discussing the pattern and the jewel of information it represents. So a definition which more closely reflects its use within the patterns community is:
A pattern is a named nugget of instructive information that captures the essential structure and insight of a successful family of proven solutions to a recurring problem that arises within a certain context and system of forces.
A slightly more compact definition which might be easier to remember is:
A pattern is a named nugget of insight that conveys the essence of a proven solution to a recurring problem within a certain context amidst competing concerns.
Patterns are usually concerned with some kind of architecture or organization of constituent parts to produce a greater whole. Richard Gabriel, author of Patterns of Software: Tales From the Software Community, provides a clear and concise definition of the term pattern in the Patterns Definitions section of the Patterns Home Page:
Each pattern is a three-part rule, which expresses a relation between a certain context, a certain system of forces which occurs repeatedly in that context, and a certain software configuration which allows these forces to resolve themselves.
As Gabriel explains, Alexander describes it a bit more colorfully in [TTWoB]:
Each pattern is a three-part rule, which expresses a relation between a certain context, a problem, and a solution.As an element in the world, each pattern is a relationship between a certain context, a certain system of forces which occurs repeatedly in that context, and a certain spatial configuration which allows these forces to resolve themselves.
As an element of language, a pattern is an instruction, which shows how this spatial configuration can be used, over and over again, to resolve the given system of forces, wherever the context makes it relevant.
The pattern is, in short, at the same time a thing, which happens in the world, and the rule which tells us how to create that thing, and when we must create it. It is both a process and a thing; both a description of a thing which is alive, and a description of the process which will generate that thing (p. 247).
In Software Patterns, esteemed "patternite" Jim Coplien writes:
I like to relate this definition to dress patterns. I could tell you how to make a dress by specifying the route of a scissors through a piece of cloth in terms of angles and lengths of cut. Or, I could give you a pattern. Reading the specification, you would have no idea what was being built or if you had built the right thing when you were finished. The pattern foreshadows the product: it is the rule for making the thing, but it is also, in many respects, the thing itself.
So we see that a pattern involves a general description of a recurring solution to a recurring problem replete with various goals and constraints. But a pattern does more than just identify a solution, it also explains why the solution is needed!
These patterns in our minds are, more or less, mental images of the patterns in the world: they are abstract representations of the very morphological rules which define the patterns in the world. However, in one respect they are very different. The patterns in the world merely exist. But the same patterns in our minds are dynamic. They have force. They are generative. They tell us what to do; they tell us how we shall, or may, generate them; and they tell us too, that under certain circumstances, we must create them. Each pattern is a rule which describes what you have to do to generate the entity which it defines. (pp. 181-182)
Generative patterns tell us how to create something and can be observed in the resulting system architectures they helped shape. Non-generative patterns describe recurring phenomena without necessarily saying how to reproduce them. We should strive to document generative patterns because they not only show us the characteristics of good systems, they teach us how to build them!
For Alexander, however, this "instructive" element is only one facet of what he calls generativity. He wants patterns, and especially pattern languages, to be capable of generating whole, living structures. Part of the desire to create architectures that emulate life lies in the remarkably unique ability of living things to evolve and adapt to their ever-changing environments (not only for the sake of individual survival, but also for survival of the species). Alexander wants to impart these same qualities into his architectures. Similarly, in software, good software architectures are all about being adaptable and resilient to change. So another aspect of generativity is about striving to create "living" architectures which are capable of dynamically adapting to fulfill changing needs and demands.
But there is still more to "generativity" than teaching and adaptability. The successive application of several patterns, each encapsulating its own problem and forces, unfolds a larger solution which emerges indirectly as a result of the smaller solutions. It is the generation of such emergent behavior that appears to be what is meant by generativity. In this fashion, patterns and pattern languages should guide their users to generate whole architectures that have the quality. This particular aspect of Alexander's paradigm seems a bit too mystical for some people's tastes.
Such "patterns-hype" ultimately causes disappointment, resentment, and even disdain when the hype proves different than the reality. This harms the credibility and legitimacy of those in the patterns community making genuine efforts to document "true" patterns. This greatly upsets many "patternites" and, as a result, there is a strong ethic within this community to avoid and dispel hype about patterns and patterns-related work. One might call this the hype-no-cratic oath: First, do no hype!.
If, as Alexander writes, a pattern must be both a process and a thing, then a pattern must describe not only the process that creates that thing, but also the thing created by that process. For this reason, many contend that patterns ultimately deal with some kind of visually discernible structure: you should be able to draw a picture of the kind of structure that results from using the pattern in practice.
A "pattern in waiting" that is not yet known to pass the patternity tests mentioned above, or some of the ones mentioned below, is often called a proto-pattern. Brief descriptions of such proto-patterns are sometimes called patlets (pronounced "pat-lets").
Documenting good patterns can be an extremely difficult task. To quote Cope once again (this time from the Patterns Definitions page), good patterns do the following:
Bear in mind that just because something isn't a pattern doesn't mean it's not any good. If something is a pattern, then (hopefully) it is good. This should by no means imply that everything good should therefore be a pattern. There are many useful things in the world besides patterns; We must not let the surrounding patterns hype lull us into equating patterns with "sugar and spice and everything nice." This simply is not true.
Clearly, the very word "pattern" suggests recurrence; if something doesn't recur, it can't possibly be a pattern. But recurrence is not the sole characteristic of importance. We also need to show that a pattern is fit for use, and that it is a useful fit. Recurrence is purely quantitative characteristic, fitness and usefulness are qualitative characteristics. We can show recurrence simply by using the rule of three; We show fitness by explaining how the pattern is successful; and we show usefulness by explaining why it is successful and beneficial.
A pattern is where theory and practice meet to reinforce and complement one another, by showing that the structure it describes is useful, useable, and used!
A pattern must be useful because this shows how having the pattern in our minds may be transformed into an instance of the pattern in the real world, as something thing that adds value to our lives as developers and practitioners. A pattern must also be useable because this shows how a pattern described in literary form may be transformed into a pattern that we have in our minds. And a pattern must be used because this is how patterns that exist in the real world first became documented as patterns in literary form.
This yields a continuously repeating cycle from pattern writers, to pattern readers, to pattern users: writers documenting patterns in literary form make them usable to pattern readers, who can then remember them in their minds, which makes them useful to practitioners and developers, who can use them in the real world, and enhance the user's quality of life.
In [Notes], Alexander writes at length about "forces" and "misfits" (or "misfit variables"), and about the struggle to achieve good fit between a good design structure and its context:
Every design problem begins with an effort to achieve fitness between two entities: the form in question and its context. The form is the solution to the problem; the context defines the problem. ... The context and the form are complementary. ... Once we have the diagram of the forces in the literal sense ... this will in essence also describe the form as a complementary diagram of forces.These misfits are the forces which must shape it [the design], and there is no mistaking them. ... We may summarize the state of each potential misfit by a [boolean] variable. Each [misfit] variable stands for one possible kind of misfit between the form and context. These [misfit variables] describe a state of affairs that is neither in the form alone nor in the context alone, but a relation between the two. The state of this relation, fit or misfit, describes one aspect of the whole ensemble. It is a condition of harmony and good fit in the ensemble that none of the possible misfits should actually occur.
These "conditions of harmony" of forces between a form and its context are what Alexander later came to call "patterns." The misfits, and the situations in which misfits dominate the "form" correspond to what many now call anti-patterns. Anti-patterns can be valuable because it is often just as important to see and understand bad solutions as it is to see and understand good ones. This makes for some potentially useful signposts to help us recognize why a particular alternative might seem sensible at first, but turns out not to be the best route to follow. It helps guide us away from the misfits and toward the appropriate patterns that might better solve our problem.
More recently on the anti-patterns front, Brown, Malveau, McCormick and Mowbray have written a book entitled Anti-Patterns, which is about "refactoring software, architectures, and projects in crisis." The anti-patterns and refactored solutions described in the book are very similar in nature to many of the "pitfalls" described by Bruce Webster in Pitfalls of Object-Oriented Development and the "lessons learned" described by Tom Love in Object Lessons.
These books attempt to do more than just identify and diagnose various "anti-solutions"; they try to go a step further by recommending a course of action for correction, recovery, and prevention. The most useful anti-patterns are those that do more than just point out Dilbert-like examples of bad practice (or of good practice gone bad). Anti-patterns that stop short at this point are about the death and destruction of good design and good process. Recognizing when this occurs is certainly valuable, but the most useful anti-patterns are about recovery, refactoring, realignment, and reconstitution of the good from the bad; they show us how to turn lemons into lemonade and how to avoid subsequent lemons, leading us away from the misfits of anti-patterns, and back towards the patterns that resolve them.
A design pattern names, abstracts, and identifies the key aspects of a common design structure that make it useful for creating a reusable object-oriented design. The design pattern identifies the participating classes and their instances, their roles and collaborations, and the distribution of responsibilities. Each design pattern focuses on a particular object-oriented design problem or issue. It describes when it applies, whether or not in can be applied in view of other design constraints, and the consequences and trade-offs of its use. Since we must eventually implement our designs, a design pattern also provides sample ... code to illustrate an implementation. Although design patterns describe object-oriented designs, they are based on practical solutions that have been implemented in mainstream object-oriented programming languages ....
The above description is slanted toward object-oriented design, but with only minor changes, it could be readily adjusted to describe software design patterns in general (simply remove the words "object-oriented" and replace "class" with "module" or "subsystem"). Since the [GoF] book was the first (and currently the most popular) of the software patterns books, the term "design pattern" is often used to refer to any pattern which directly addresses issues of software architecture, design, or programming implementation. Many choose to make an important distinction between these three conceptual levels by categorizing them into architectural patterns, design patterns, and idioms (idioms are sometimes called coding patterns). The authors of Patterns of Software Architecture [POSA] define these three types of patterns as follows:
The difference between these three kinds of patterns are in their corresponding levels of abstraction and detail. Architectural patterns are high-level strategies that concern large-scale components and the global properties and mechanisms of a system. They have wide-sweeping implications which affect the overall skeletal structure and organization of a software system. Design patterns are medium-scale tactics that flesh out some of the structure and behavior of entities and their relationships. They do not influence overall system structure, but instead define micro-architectures of subsystems and components. Idioms are paradigm-specific and language-specific programming techniques that fill in low-level internal or external details of a component's structure or behavior.
In "Understanding and Using Patterns in Software Development", Riehle and Zullighoven make similar distinctions, but seem to partition the different kinds of patterns among analysis, design, and implementation. They define the terms conceptual patterns, design patterns, and programming patterns as follows:
Using these definitions, conceptual patterns are based upon metaphors in a restricted application domain. Design patterns complement, or elaborate upon conceptual patterns by delving into the implementation of elements from the conceptual space. And programming patterns descend further into implementation details using a specific implementation language.
When comparing and contrasting these two sets of definitions, it appears that programming patterns are equivalent to idioms. For the other types of patterns described above, the first set of authors choose to delineate them by their architectural scope whereas the latter set of authors choose to delineate them by whether they employ language from the problem space or the solution space.
Although it is not strictly required, good patterns often begin with an Abstract that provides a short summary or overview. This gives readers a clear picture of the pattern and quickly informs them of its relevance to any problems they may wish to solve (sometimes such a description is called a thumbnail sketch of the pattern, or a pattern thumbnail). A pattern should identify its target audience and make clear what it assumes of the reader.
The aim is that, if well written, each pattern describes a whole that is greater than the sum of its parts, due to skillful choreography of its elements working together to satisfy all its varying demands.
Rules aren't commonly supported by a rationale, nor put in context. A rule may be part of the solution in a pattern description, but a rule solution is neither sufficient nor necessary. Patterns aren't designed to be executed or analyzed by computers, as one might imagine to be true for rules: patterns are to be executed by architects with insight, taste, experience, and a sense of aesthetics.
A pattern is the process that generates a solution, but it may generate any one of a vast number of variant solutions (conceivably without repeating the same solution twice). The human element of patterns is what chiefly contributes to their variability and adaptability, and usually requires a greater degree of creativity in their application and combination. So, just as the processes of architecture and design are creative endeavors, so too is the application of patterns. In the same paper quoted above, Cope goes on to say:
If design is codified in patterns, does the need for creativity go away? Can we replace high-priced expensive designers with less sophisticated programmers who are guided by patterns? The answer is that creativity is still needed to shape the patterns to a given context. Just as a dressmaker tailors a pattern to an individual customer, and perhaps to a specific event where the dress is to be worn, so designers must be creative when using patterns. Patterns channel creativity; they neither replace nor constrain it.
Algorithms and data structures are usually concerned almost exclusively with optimizing space or time or some other aspect of computational complexity and resource consumption. They are about finding the most compact and efficient means to perform some important computation or store and recall its results. Such algorithms and data structures are rarely concerned with compromises and tradeoffs regarding other concerns that other concerns that have little to do with things like performance or memory, and more to do with things like maintainability and adaptability and (re)usability of the architecture.
There are a great many books of algorithms and data-structures that provide source code and quantitative analysis for structures like AVL trees or splay-trees. But in many of these same textbooks there is little mention of how to implement these structures in ways that emphasize maintainability and adaptability and reuse; and when they do there is a whole new set of issues to worry about (like maintainability and reusability and encapsulation). Looking at the computational time and space aspects alone simply doesn't address the fuller set of forces that affect the architects and implementors as well as the users.
This is why algorithms and data structures tend to be more fine-grained than patterns: Because they mostly address issues of computational complexity, and not so much the underlying issues of the people who are both using and building the software. Patterns fundamentally address people issues (like maintainability) more so than simple hardware and software efficiency/memory issues.
Of course software developers need to be concerned both with finding appropriate architectures and with finding appropriate solutions to computational problems. So there will always be a need for patterns as well as for algorithms and data structures (and their use together).
A software framework is a reusable mini-architecture that provides the generic structure and behavior for a family of software abstractions, along with a context of memes/metaphors which specifies their collaboration and use within a given domain.
The framework accomplishes this by hardcoding the context into a kind of "virtual machine" (or "virtual engine"), while making the abstractions open-ended by designing them with specific plug-points (also called hot spots). These plug-points (typically implemented using callbacks, polymorphism, or delegation) enable the framework to be adapted and extended to fit varying needs, and to be successfully composed with other frameworks. A framework is usually not a complete application: it often lacks the necessary application-specific functionality. Instead, an application may be constructed from one or more frameworks by inserting this missing functionality into the plug-and-play "outlets" provided by the frameworks. Thus, a framework supplies the infrastructure and mechanisms that execute a policy for interaction between abstract components with open implementations.
A definition of an object-oriented software framework is given in [GoF]:
A framework is a set of cooperating classes that make up a reusable design for a specific class of software. A framework provides architectural guidance by partitioning the design into abstract classes and defining their responsibilities and collaborations. A developer customizes a framework to a particular application by subclassing and composing instances of framework classes.... [a framework] dictates the architecture of your application. It will define the overall structure, its partitioning into classes and objects, the key responsibilities thereof, how the classes and objects collaborate, and the thread of control. A framework predefines these design parameters so that you, the application designer/implementer, can concentrate on the specifics of your application. The framework captures the design decisions that are common to its application domain. Frameworks thus emphasize design reuse over code reuse, though a framework will usually include concrete subclasses you can put to work immediately.
The difference between a framework and an ordinary programming library is that a framework employs an inverted flow of control between itself and its clients. When using a framework, one usually just implements a few callback functions or specializes a few classes, and then invokes a single method or procedure. At this point, the framework does the rest of the work for you, invoking any necessary client callbacks or methods at the appropriate time and place. For this reason, frameworks are often said to abide by the Hollywood Principle ("Don't call us, we'll call you.") or the Greyhound Principle ("Leave the driving to us.").
Design patterns may be employed both in the design and the documentation of a framework. A single framework typically encompasses several design patterns. In fact, a framework can be viewed as the implementation of a system of design patterns. Despite the fact that they are related in this manner, it is important to recognize that frameworks and design patterns are two distinctly separate beasts: a framework is executable software, whereas design patterns represent knowledge and experience about software. In this respect, frameworks are of a physical nature, while patterns are of a logical nature: frameworks are the physical realization of one or more software pattern solutions; patterns are the instructions for how to implement those solutions.
The [GoF] book describes the major differences between design patterns and frameworks as follows:
Alexander proposes the existence of an objective quality of aesthetic beauty that is universally recognizable. He claims there are certain timeless attributes and properties which are considered beautiful and aesthetically pleasing to all people in all cultures (not just "in the eye of the beholder"). It is these fundamental properties which combine to generate the QWAN, and which make a structure feel "whole" and "alive".
Alexander conducted some experiments using configurations of colored beads, and also using the design of carpets and tapestries. Those results, combined with his own architectural experience with the users of his buildings, suggested the presence of this "objective beauty" was closely tied to the presence of symmetries and subsymmetries that balance the use of contrasting space, light, and color to form fields of visual centers. Feelings of beauty and order would increase when these visual centers unfolded recursively at multiple hierarchical levels of granular scale throughout a design (much like fractals). However, if the symmetries are too pure (too perfect or exact), it seems to be less desirable than if slight irregularities and imperfections exist. Apparently some imperfections are not only palatable, but can also be utilitarian (just as doped imperfections in precious stones can actually enhance their crystalline structure).
In The Laws of Architecture from a Physicist's Perspective, artist and mathematician Nikos A. Salingaros (who has worked closely with Alexander over the past ten years) discusses the rules of beauty and order in past times and, as a result, proposes the following three laws of architecture:
Salingaros goes on to say that these laws don't govern beauty in architecture, they govern life in architecture: "They are extracted from physics and mathematics by looking at how nature is put together. I discovered these laws by observing how fundamental particles come together to form a structure." He says that when structures conform to these architectural laws, those who enter them feel a kind of resonant harmony, an almost emotional connection to the structure which feels peaceful and nourishing. This connection between life and architecture is due to the thermodynamics of living forms, but the idea of building structures which possess these intrinsic qualities of living biological organisms comes from Alexander.
In Reengineering the Application Development Process, Michael Beedle tries to succinctly describe the QWAN as something that:
is created when the attributes in the design makes that design "live". That is, designs that are flexible, extensible, adaptable, reusable and have other qualities of living things; except of course self-reproduction and metabolism.
According to this definition, the QWAN is attained when the pattern or pattern language generates a "live" architectural solution; one which emulates the capability of living things to dynamically adapt to fulfill changing needs and demands. This is closely related to Alexander's idea of generativity. In "Christopher Alexander: an Introduction for Object-Oriented Designers" , Doug Lea attempts to describe the QWAN as follows:
... "the quality without a name", the possession of which is the ultimate goal of any design product. It is impossible to briefly summarize this. Alexander presents a number of partial synonyms: freedom, life, wholeness, comfortability, harmony, But no single term or example fully conveys meaning or captures the force of Alexander's writings on the reader, especially surrounding the human impact of design, the feelings and aesthetics of designers and users, the need for commitment by developers to obtain and preserve wholeness, and its basis in the objective equilibrium of form.
So another crucial aspect of the QWAN is the effect it has upon the architecture's inhabitants and designers that makes them feel alive, whole, and comfortable. It is this kind of "habitability" that improves user comfort and quality of life (what TQM circles might refer to as "total customer satisfaction", and what Tom Peters means by "to thrill and delight" the customer in the pursuit of "Wow!").
Of course this "ultimate goal" of achieving the QWAN is very elusive. This particular aspect of patterns seems to borrow concepts from Zen Buddhism, Taoism, and Platonic ideals. There are those who feel the whole idea of the QWAN is just a bit too whimsical and metaphysical; that it is not scientific or tangible enough to have a place in true engineering disciplines (especially if it is something that cannot be measured or quantified). The notion of beauty as something that is universally objective can be a trifle hard for some to swallow.
But in many respects, an individual's sense of the QWAN is also about cognitive judgement. Every master designer develops their own highly honed intuition which is borne from extensive experience. Although this "intuition" may be subjective, it can be uncannily accurate and give the designer an almost instinctive sense of what will work and what wont (even before the measures are brought to bear to try and verify it). This stems from the designer being intimate with the design and internalizing it at a visceral level, almost to the point of becoming an inhabitant whose sensory network is "plugged-in" to the system. If a pattern can impart to its readers and users, this same "plugged-in" feeling of being connected to the design and deeply comprehending it, then in theory it will impart to the reader the same cognitive feeling of its aptness that the designer experienced. If a pattern succeeds in this attempt, then all who see and use it will supposedly experience the resonant feeling of beauty and harmony that the QWAN is supposed to evoke.
A pattern language defines a collection of patterns and the rules to combine them into an architectural style. Pattern languages describe software frameworks or families of related systems.
Cope gives a slightly different definition in "Software Design Patterns: Common Questions and Answers":
A pattern language is a structured collection of patterns that build on each other to transform needs and constraints into an architecture.
Cope then goes on to say that:
Good pattern languages guide the designer toward useful architectures and away from architectures whose literary analogies are gibberish or unartful writing. Good architectures are durable, functional, and aesthetically pleasing, and a good combination of patterns can balance the forces on a system to strive towards these three goals. A good pattern language gives designers freedom to express themselves and to tailor the solution to the particular needs of the context where the patterns are applied.
Unlike a mere pattern compilation or catalog, a pattern language includes rules and guidelines which explain how and when to apply its patterns to solve a problem which is larger than any individual pattern can solve. These rules and guidelines suggest the order and granularity for applying each pattern in the language. A pattern language may be regarded as a lexicon of patterns plus a grammar that defines how to weave them together into valid sentences (or artful tapestries if you will). Ideally, good pattern languages are generative, capable of generating all the possible sentences from a rich and expressive pattern vocabulary.
A pattern language forms a gestalt in which each of its patterns collaborate to solve a more fundamental problem that is not explicitly addressed by any individual pattern. This helps a pattern language to achieve an organic order, which Alexander describes in [Oregon] as "the kind of order that is achieved when there is a perfect balance between the needs of the parts and the needs of the whole." So in a sense, a pattern language is like an ecosystem of patterns, all of which are inherently related at some level. This "ecological quality" of pattern languages contributes to their "wholeness" and their ability to help us generate "live" architectures" possessing the QWAN.
Throughout The Timeless Way of Building, Alexander remarks on what a pattern language truly embodies:
Michael Beedle, author of Reengineering the Application Development Process, likens the effects of using pattern languages to the generation of emergent behaviors: spontaneously recurring patterns of dense local interaction between entities, resulting in dynamic, self-organizing systems that are adaptive, open, and capable of multi-scale effects. In other words, pattern languages provide a dynamic process for the orderly resolution of problems within their domain which indirectly leads to the resolution of a much broader problem. The patterns and rules in a pattern language combine to form an architectural style. In this manner, pattern languages guide system analysts, architects, designers, and implementors to produce workable systems that solve common organizational and development problems at all levels of scale and diversity.
Large lump development hinges on a view of the environment which is static and discontinuous; piecemeal growth hinges on a view of the environment which is dynamic and continuous.... According to the large lump point of view, each act of design or construction is an isolated event which creates an isolated building -- "perfect" at the time of construction, and then abandoned by its builders and designers forever. According to the piecemeal point of view, every environment is changing and growing all the time, in order to keep its use in balance; and the quality of the environment is a kind of semi-stable equilibrium in the flux of time.... Large lump development is based on the idea of replacement. Piecemeal growth is based on the idea of repair.
Those familiar with software development lifecycles might see some similarities between large lump development and the waterfall model, and between piecemeal growth, and the spiral models which involve prototyping and incremental/evolutionary development (or what Booch calls "round-trip gestalt design").
The relationship between patterns and piecemeal growth is that pattern languages are intended to grow and evolve whole architectures through this process of piecemeal growth. The various sequences in which a pattern language instructs the user in the application of the patterns therein should unfold a complete architecture which gradually emerges from the successive application of individual patterns in the appropriate order.
A pattern catalog adds a modicum of structure and organization to a pattern collection, but doesn't usually go very far beyond showing only the most outwardly visible structure and relationships (if in fact it shows any of them). A pattern system adds deep structure, rich pattern interaction, and uniformity to a pattern catalog. Both pattern systems and pattern languages form coherent sets of tightly interwoven patterns for describing and solving problems in a particular domain. But a pattern language adds robustness, comprehensiveness, and wholeness to a pattern system. The primary difference is that, ideally, pattern languages are computationally complete, showing all possible combinations of patterns and their variations to produce complete architectures. In practice however, the difference between pattern systems and pattern languages can be extremely difficult to ascertain.
While a pattern system may be a cohesive collection of patterns about a very broad topic, a pattern language has to be about more than just a "broad topic". A pattern language ultimately corresponds to a single-minded collective that forms a kind of "mega pattern" or "super pattern" in that the entire language possesses an underlying, shared problem and its context, forces, solution, resulting context, and rationale (which each pattern in the language addresses at some level or another). This coherence of purpose is what gives the pattern language a sense of closure. A pattern system does not necessarily have all of these pattern elements. It may focus on an equally broad or narrow topic, but it may not necessarily have a clear mission or agenda (and may result in many relationships between patterns being harder to find), or it leaves several important gaps unfilled in the problem space (and hence may not attain overall resolution or closure).
But perhaps the most important difference between pattern languages and pattern systems is that pattern languages are not created all at once. They evolve from pattern systems through the process of piecemeal growth (and a pattern system may, in turn, evolve from a pattern catalog in a similar manner). So just as pattern languages help to incrementally grow whole architectures, pattern systems may serve to incrementally grow into whole pattern languages.
How do you know a pattern when you come across one? The answer is you don't always know. You may jot down the beginnings of some things you think are patterns, but it may turn out that they aren't patterns at all, or they are only pieces of patterns, or simply good principles or rules of thumb that may form part of the rationale of a particular pattern. It is important to remember that a solution in which no forces are present is not a pattern.
The best way to learn how to recognize and document useful patterns is by learning from others who have done it well! Pick up several books and articles which describe patterns (don't choose just one) and try and see if you can recognize all the necessary pattern elements and desirable qualities that were mentioned earlier in this paper. When you see one that appeals to you, ask yourself why it is good. If you see one you don't like, try and figure out exactly what it is about the pattern that leaves you unsatisfied. Read as much as you can, and try to learn from the masters. Numerous resources for learning more about patterns are given near the end of this paper. Most importantly, be introspective about everything you read! Examine how it is meaningful to you and how it will help you accomplish future goals.
After you have been exposed to a wealth of pattern literature, choose one of the various pattern formats and see if you can flesh out some of your earlier ideas for things you thought might be patterns. If you are trying to compose a language of patterns, start by examining the forces and context of each pattern and try to identify any simple underlying principles that seem to help organize the patterns together into useful configurations.
Further assistance for those who are courageous enough to undertake writing a pattern, or a pattern language, may be found in Ward Cunningham's "Tips for Writing Pattern Languages", and in Gerard Meszaros' and Jim Doble's "A Pattern Language for Pattern Writing". Both of these papers are indispensable resources for writing patterns and pattern languages.
How do the experts decide what makes a good pattern? The PLoP conferences have several criteria which they feel submitted pattern papers should meet. These are as follows (summarized from Buschmann et. al. in Pattern-Oriented Software Architecture):
There is speculation that patterns will one day replace computer programmers. Such speculation has occurred many times in the past regarding various other "new" technologies. The languages and tools that will be used to solve software problems in the future may advance far beyond what is presently available (much like programming languages and development environments have evolved since the days of programming machine-code on punch cards), but they will still require developers with similarly evolved skills and wisdom to use them effectively. As long as software developers keep abreast of emerging software concepts and technologies, there will always be a need for them to skillfully and creatively construct useful software solutions using these new tools and languages.
So while the ability to codify patterns as generic software components may be important, even more important is the knowledge of how/when to apply and combine patterns, in conjunction with the ability to use a shared vocabulary of pattern names to communicate the nuggets of insight they represent. Because patterns capture knowledge that is primarily intended for humans, it is the social impact of patterns which largely shapes their technological impact.
What about Christopher Alexander? His writings on patterns and pattern languages were composed more than two decades before they became popular within the software community. During these past few decades, Alexander has been laboring on a new book entitled The Nature of Order which significantly advances his earlier ideas about patterns, architecture, and beauty. This new work will span four volumes, some of which are tentatively planned for publication by the end of 1997. In describing the book, one of its editors, Nikos A. Salingaros writes:
Alexander develops a comprehensive theory of how matter comes together to form coherent structures. Paralleling, but not copying, recent results from complexity theory, he argues that the same laws apply to all structures in the universe; from atoms, to crystals, to living forms, to galaxies.
This book encompasses all the concepts and theories discussed in Alexander's earlier works and extends them to develop a much broader paradigm for creating architectures. It discusses a new concept of wholeness that emerges from fields of centers (which are constructed from small pattern languages) with universal recursive properties, and from dynamic processes called sequences (which relate the order in which to visit centers and apply patterns) that employ structure-preserving transformations. Alexander presents his new architectural paradigm with concepts like:
Many in the patterns community are awaiting publication of this book with enormous anticipation as it promises to be a work of monumental significance for all walks of architecture.
Patterns represent distilled experience which, through their assimilation, convey expert insight and knowledge to inexpert developers. They help forge the foundation of a shared architectural vision, and collective of styles. If we want software development to evolve into a mature engineering discipline, then these proven "best practices" and "lessons learned" must be aggressively and formally documented, compiled, scrutinized, and widely disseminated as patterns (and anti-patterns). Once a solution has been expressed in pattern form, it may then be applied and reapplied to other contexts, and facilitate widespread reuse across the entire spectrum of software engineering artifacts such as: analyses, architectures, designs, implementations, algorithms and data structures, tests, plans, and organization structures.
Patterns are not a "silver bullet"! They are extremely valuable tools for capturing and communicating acquired knowledge and experience to improve software quality and productivity by addressing fundamental issues in the development of software. By employing these tools we are better suited to meet challenges like "communication of architectural knowledge among developers; accommodating a new design paradigm or architectural style; resolving nonfunctional forces such as reusability, portability, and extensibility; and avoiding development traps and pitfalls that have traditionally been learned only by experience." (quoted from the PLoP'96 Call for Papers.)
Perhaps the final remarks from Pattern-Oriented Software Architecture best describe the significance of patterns for software:
Patterns expose knowledge about software construction that has been gained by many experts over many years. All work on patterns should therefore focus on making this precious resource widely available. Every software developer should be able to use patterns effectively when building software systems. When this is achieved, we will be able to celebrate the human intelligence that patterns reflect, both in each individual pattern and in all patterns in their entirety.
Other more general patterns resources on the web are:
|
|
|