5.2 Key Design Concepts
Good design depends on understanding a handful of key concepts. This section discusses the role of complexity, desirable characteristics of designs, and levels of design.
Software's Primary Technical Imperative: Managing Complexity
To understand the importance of managing complexity, it's useful to refer to Fred Brooks's landmark paper, "No Silver Bullets: Essence and Accidents of Software Engineering" (1987).
Accidental and Essential Difficulties
Brooks argues that software development is made difficult because of two different classes of problems—the essential and the accidental. In referring to these two terms, Brooks draws on a philosophical tradition going back to Aristotle. In philosophy, the essential properties are the properties that a thing must have in order to be that thing. A car must have an engine, wheels, and doors to be a car. If it doesn't have any of those essential properties, it isn't really a car.
Accidental properties are the properties a thing just happens to have, properties that don't really bear on whether the thing is what it is. A car could have a V8, a turbocharged 4-cylinder, or some other kind of engine and be a car regardless of that detail. A car could have two doors or four; it could have skinny wheels or mag wheels. All those details are accidental properties. You could also think of accidental properties as incidental, discretionary, optional, and happenstance.
Brooks observes that the major accidental difficulties in software were addressed long ago. For example, accidental difficulties related to clumsy language syntaxes were largely eliminated in the evolution from assembly language to third-generation languages and have declined in significance incrementally since then. Accidental difficulties related to noninteractive computers were resolved when time-share operating systems replaced batch-mode systems. Integrated programming environments further eliminated inefficiencies in programming work arising from tools that worked poorly together.
Brooks argues that progress on software's remaining essential difficulties is bound to be slower. The reason is that, at its essence, software development consists of working out all the details of a highly intricate, interlocking set of concepts. The essential difficulties arise from the necessity of interfacing with the complex, disorderly real world; accurately and completely identifying the dependencies and exception cases; designing solutions that can't be just approximately correct but that must be exactly correct; and so on. Even if we could invent a programming language that used the same terminology as the real-world problem we're trying to solve, programming would still be difficult because of the challenge in determining precisely how the real world works. As software addresses ever-larger real-world problems, the interactions among the real-world entities become increasingly intricate, and that in turn increases the essential difficulty of the software solutions.
The root of all these essential difficulties is complexity—both accidental and essential.
Importance of Managing Complexity
When software-project surveys report causes of project failure, they rarely identify technical reasons as the primary causes of project failure. Projects fail most often because of poor requirements, poor planning, or poor management. But when projects do fail for reasons that are primarily technical, the reason is often uncontrolled complexity. The software is allowed to grow so complex that no one really knows what it does. When a project reaches the point at which no one completely understands the impact that code changes in one area will have on other areas, progress grinds to a halt.
There are two ways of constructing a software design: one way is to make it so simple that there are obviously no deficiencies, and the other is to make it so complicated that there are no obvious deficiencies.
- —C. A. R. Hoare
Managing complexity is the most important technical topic in software development. In my view, it's so important that Software's Primary Technical Imperative has to be managing complexity.
Complexity is not a new feature of software development. Computing pioneer Edsger Dijkstra pointed out that computing is the only profession in which a single mind is obliged to span the distance from a bit to a few hundred megabytes, a ratio of 1 to 109, or nine orders of magnitude (Dijkstra 1989). This gigantic ratio is staggering. Dijkstra put it this way: "Compared to that number of semantic levels, the average mathematical theory is almost flat. By evoking the need for deep conceptual hierarchies, the automatic computer confronts us with a radically new intellectual challenge that has no precedent in our history." Of course software has become even more complex since 1989, and Dijkstra's ratio of 1 to 109could easily be more like 1 to 1015 today.
Dijkstra pointed out that no one's skull is really big enough to contain a modern computer program (Dijkstra 1972), which means that we as software developers shouldn't try to cram whole programs into our skulls at once; we should try to organize our programs in such a way that we can safely focus on one part of it at a time. The goal is to minimize the amount of a program you have to think about at any one time. You might think of this as mental juggling—the more mental balls the program requires you to keep in the air at once, the more likely you'll drop one of the balls, leading to a design or coding error.
- One symptom that you have bogged down in complexity overload is when you find yourself doggedly applying a method that is clearly irrelevant, at least to any outside observer. It is like the mechanically inept person whose car breaks down—so he puts water in the battery and empties the ashtrays.
- —P. J. Plauger
At the software-architecture level, the complexity of a problem is reduced by dividing the system into subsystems. Humans have an easier time comprehending several simple pieces of information than one complicated piece. The goal of all software-design techniques is to break a complicated problem into simple pieces. The more independent the subsystems are, the more you make it safe to focus on one bit of complexity at a time. Carefully defined objects separate concerns so that you can focus on one thing at a time. Packages provide the same benefit at a higher level of aggregation.
Keeping routines short helps reduce your mental workload. Writing programs in terms of the problem domain, rather than in terms of low-level implementation details, and working at the highest level of abstraction reduce the load on your brain.
The bottom line is that programmers who compensate for inherent human limitations write code that's easier for themselves and others to understand and that has fewer errors.
How to Attack Complexity
Overly costly, ineffective designs arise from three sources:
A complex solution to a simple problem
A simple, incorrect solution to a complex problem
An inappropriate, complex solution to a complex problem
As Dijkstra pointed out, modern software is inherently complex, and no matter how hard you try, you'll eventually bump into some level of complexity that's inherent in the real-world problem itself. This suggests a two-prong approach to managing complexity:
Minimize the amount of essential complexity that anyone's brain has to deal with at any one time.
Keep accidental complexity from needlessly proliferating.
Once you understand that all other technical goals in software are secondary to managing complexity, many design considerations become straightforward.
Desirable Characteristics of a Design
A high-quality design has several general characteristics. If you could achieve all these goals, your design would be very good indeed. Some goals contradict other goals, but that's the challenge of design—creating a good set of tradeoffs from competing objectives. Some characteristics of design quality are also characteristics of a good program: reliability, performance, and so on. Others are internal characteristics of the design.
When I am working on a problem I never think about beauty. I think only how to solve the problem. But when I have finished, if the solution is not beautiful, I know it is wrong.
- —R.Buckminster Fuller
Here's a list of internal design characteristics:
Minimal complexity. The primary goal of design should be to minimize complexity for all the reasons just described. Avoid making "clever" designs. Clever designs are usually hard to understand. Instead make "simple" and "easy-to-understand" designs. If your design doesn't let you safely ignore most other parts of the program when you're immersed in one specific part, the design isn't doing its job.
Ease of maintenance. Ease of maintenance means designing for the maintenance programmer. Continually imagine the questions a maintenance programmer would ask about the code you're writing. Think of the maintenance programmer as your audience, and then design the system to be self-explanatory.
Loose coupling. Loose coupling means designing so that you hold connections among different parts of a program to a minimum. Use the principles of good abstractions in class interfaces, encapsulation, and information hiding to design classes with as few interconnections as possible. Minimal connectedness minimizes work during integration, testing, and maintenance.
Extensibility. Extensibility means that you can enhance a system without causing violence to the underlying structure. You can change a piece of a system without affecting other pieces. The most likely changes cause the system the least trauma.
Reusability. Reusability means designing the system so that you can reuse pieces of it in other systems.
High fan-in. High fan-in refers to having a high number of classes that use a given class. High fan-in implies that a system has been designed to make good use of utility classes at the lower levels in the system.
Low-to-medium fan-out. Low-to-medium fan-out means having a given class use a low-to-medium number of other classes. High fan-out (more than about seven) indicates that a class uses a large number of other classes and may therefore be overly complex. Researchers have found that the principle of low fan-out is beneficial whether you're considering the number of routines called from within a routine or the number of classes used within a class (Card and Glass 1990; Basili, Briand, and Melo 1996).
Portability. Portability means designing the system so that you can easily move it to another environment.
Leanness. Leanness means designing the system so that it has no extra parts (Wirth 1995, McConnell 1997). Voltaire said that a book is finished not when nothing more can be added but when nothing more can be taken away. In software, this is especially true because extra code has to be developed, reviewed, tested, and considered when the other code is modified. Future versions of the software must remain backward-compatible with the extra code. The fatal question is "It's easy, so what will we hurt by putting it in?"
Stratification. Stratification means trying to keep the levels of decomposition stratified so that you can view the system at any single level and get a consistent view. Design the system so that you can view it at one level without dipping into other levels.
For example, if you're writing a modern system that has to use a lot of older, poorly designed code, write a layer of the new system that's responsible for interfacing with the old code. Design the layer so that it hides the poor quality of the old code, presenting a consistent set of services to the newer layers. Then have the rest of the system use those classes rather than the old code. The beneficial effects of stratified design in such a case are (1) it compartmentalizes the messiness of the bad code and (2) if you're ever allowed to jettison the old code or refactor it, you won't need to modify any new code except the interface layer.
Standard techniques. The more a system relies on exotic pieces, the more intimidating it will be for someone trying to understand it the first time. Try to give the whole system a familiar feeling by using standardized, common approaches.
Levels of Design
Design is needed at several different levels of detail in a software system. Some design techniques apply at all levels, and some apply at only one or two. Figure 5-2 illustrates the levels.
Figure 5-2. The levels of design in a program. The system (1) is first organized into subsystems (2). The subsystems are further divided into classes (3), and the classes are divided into routines and data (4). The inside of each routine is also designed (5)
Level 1: Software System
The first level is the entire system. Some programmers jump right from the system level into designing classes, but it's usually beneficial to think through higher level combinations of classes, such as subsystems or packages.
In other words—and this is the rock-solid principle on which the whole of the Corporation's Galaxywide success is founded—their fundamental design flaws are completely hidden by their superficial design flaws.
- —Douglas Adams
Level 2: Division into Subsystems or Packages
The main product of design at this level is the identification of all major subsystems. The subsystems can be big: database, user interface, business rules, command interpreter, report engine, and so on. The major design activity at this level is deciding how to partition the program into major subsystems and defining how each subsystem is allowed to use each other subsystem. Division at this level is typically needed on any project that takes longer than a few weeks. Within each subsystem, different methods of design might be used—choosing the approach that best fits each part of the system. In Figure 5-2, design at this level is marked with a 2.
Of particular importance at this level are the rules about how the various subsystems can communicate. If all subsystems can communicate with all other subsystems, you lose the benefit of separating them at all. Make each subsystem meaningful by restricting communications.
Suppose for example that you define a system with six subsystems, as shown in Figure 5-3. When there are no rules, the second law of thermodynamics will come into play and the entropy of the system will increase. One way in which entropy increases is that, without any restrictions on communications among subsystems, communication will occur in an unrestricted way, as in Figure 5-4.
Figure 5-3. An example of a system with six subsystems
Figure 5-4. An example of what happens with no restrictions on intersubsystem communications
As you can see, every subsystem ends up communicating directly with every other subsystem, which raises some important questions:
How many different parts of the system does a developer need to understand at least a little bit to change something in the graphics subsystem?
What happens when you try to use the business rules in another system?
What happens when you want to put a new user interface on the system, perhaps a command-line UI for test purposes?
What happens when you want to put data storage on a remote machine?
You might think of the lines between subsystems as being hoses with water running through them. If you want to reach in and pull out a subsystem, that subsystem is going to have some hoses attached to it. The more hoses you have to disconnect and reconnect, the more wet you're going to get. You want to architect your system so that if you pull out a subsystem to use elsewhere, you won't have many hoses to reconnect and those hoses will reconnect easily.
With forethought, all of these issues can be addressed with little extra work. Allow communication between subsystems only on a "need to know" basis—and it had better be a good reason. If in doubt, it's easier to restrict communication early and relax it later than it is to relax it early and then try to tighten it up after you've coded several hundred intersubsystem calls. Figure 5-5 shows how a few communication guidelines could change the system depicted in Figure 5-4.
Figure 5-5. With a few communication rules, you can simplify subsystem interactions significantly
To keep the connections easy to understand and maintain, err on the side of simple intersubsystem relations. The simplest relationship is to have one subsystem call routines in another. A more involved relationship is to have one subsystem contain classes from another. The most involved relationship is to have classes in one subsystem inherit from classes in another.
A good general rule is that a system-level diagram like Figure 5-5 should be an acyclic graph. In other words, a program shouldn't contain any circular relationships in which Class A uses Class B, Class B uses Class C, and Class C uses Class A.
On large programs and families of programs, design at the subsystem level makes a difference. If you believe that your program is small enough to skip subsystem-level design, at least make the decision to skip that level of design a conscious one.
Common Subsystems. Some kinds of subsystems appear again and again in different systems. Here are some of the usual suspects.
Business rules. Business rules are the laws, regulations, policies, and procedures that you encode into a computer system. If you're writing a payroll system, you might encode rules from the IRS about the number of allowable withholdings and the estimated tax rate. Additional rules for a payroll system might come from a union contract specifying overtime rates, vacation and holiday pay, and so on. If you're writing a program to quote automobile insurance rates, rules might come from government regulations on required liability coverages, actuarial rate tables, or underwriting restrictions
User interface. Create a subsystem to isolate user-interface components so that the user interface can evolve without damaging the rest of the program. In most cases, a user-interface subsystem uses several subordinate subsystems or classes for the GUI interface, command line interface, menu operations, window management, help system, and so forth.
Database access. You can hide the implementation details of accessing a database so that most of the program doesn't need to worry about the messy details of manipulating low-level structures and can deal with the data in terms of how it's used at the business-problem level. Subsystems that hide implementation details provide a valuable level of abstraction that reduces a program's complexity. They centralize database operations in one place and reduce the chance of errors in working with the data. They make it easy to change the database design structure without changing most of the program.
System dependencies. Package operating-system dependencies into a subsystem for the same reason you package hardware dependencies. If you're developing a program for Microsoft Windows, for example, why limit yourself to the Windows environment? Isolate the Windows calls in a Windows-interface subsystem. If you later want to move your program to Mac OS or Linux, all you'll have to change is the interface subsystem. An interface subsystem can be too extensive for you to implement on your own, but such subsystems are readily available in any of several commercial code libraries.
Level 3: Division into Classes
Design at this level includes identifying all classes in the system. For example, a database-interface subsystem might be further partitioned into data access classes and persistence framework classes as well as database metadata. Figure 5-2, Level 3, shows how one of Level 2's subsystems might be divided into classes, and it implies that the other three subsystems shown at Level 2 are also decomposed into classes.
Details of the ways in which each class interacts with the rest of the system are also specified as the classes are specified. In particular, the class's interface is defined. Overall, the major design activity at this level is making sure that all the subsystems have been decomposed to a level of detail fine enough that you can implement their parts as individual classes.
The division of subsystems into classes is typically needed on any project that takes longer than a few days. If the project is large, the division is clearly distinct from the program partitioning of Level 2. If the project is very small, you might move directly from the whole-system view of Level 1 to the classes view of Level 3.
Classes vs. Objects. A key concept in object-oriented design is the differentiation between objects and classes. An object is any specific entity that exists in your program at run time. A class is the static thing you look at in the program listing. An object is the dynamic thing with specific values and attributes you see when you run the program. For example, you could declare a class Person that had attributes of name, age, gender, and so on. At run time you would have the objects nancy, hank, diane, tony, and so on—that is, specific instances of the class. If you're familiar with database terms, it's the same as the distinction between "schema" and "instance." You could think of the class as the cookie cutter and the object as the cookie. This book uses the terms informally and generally refers to classes and objects more or less inter-changeably.
Level 4: Division into Routines
Design at this level includes dividing each class into routines. The class interface defined at Level 3 will define some of the routines. Design at Level 4 will detail the class's private routines. When you examine the details of the routines inside a class, you can see that many routines are simple boxes but a few are composed of hierarchically organized routines, which require still more design.
The act of fully defining the class's routines often results in a better understanding of the class's interface, and that causes corresponding changes to the interface—that is, changes back at Level 3.
This level of decomposition and design is often left up to the individual programmer, and it's needed on any project that takes more than a few hours. It doesn't need to be done formally, but it at least needs to be done mentally.
Level 5: Internal Routine Design
Design at the routine level consists of laying out the detailed functionality of the individual routines. Internal routine design is typically left to the individual programmer working on an individual routine. The design consists of activities such as writing pseudocode, looking up algorithms in reference books, deciding how to organize the paragraphs of code in a routine, and writing programming-language code. This level of design is always done, though sometimes it's done unconsciously and poorly rather than consciously and well. In Figure 5-2, design at this level is marked with a 5.