Author Hal Fulton has updated his modern classic, The Ruby Way. The publication of the second edition, due the third week of October to coincide with RubyConf 2006, marks the launch of Addison Wesley's Professional Ruby Series. In this InfoQ exclusive excerpt, Hal answers the question which sets the tone for his entire book:
What is “the Ruby Way”?
Let us prepare to grapple with the ineffable itself, and see if we may not eff it after all.
— Douglas Adams, Dirk Gently’s Holistic Detective Agency
What do we mean by the Ruby Way? My belief is that there are two related aspects: One is the philosophy of the design of Ruby; the other is the philosophy of its usage. It is natural that design and use should be interrelated, whether in software or hardware; why else should there be such a field as ergonomics? If I build a device and put a handle on it, it is because I expect someone to grab that handle.
Ruby has a nameless quality that makes it what it is. We see that quality present in the design of the syntax and semantics of the language, but it is also present in the programs written for that interpreter. Yet as soon as we make this distinction, we blur it.
Clearly Ruby is not just a tool for creating software, but it is a piece of software in its own right. Why should the workings of Ruby programs follow laws different from those that guide the workings of the interpreter? After all, Ruby is highly dynamic and extensible. There might be reasons that the two levels should differ here and there, probably for accommodating to the inconvenience of the real world. But in general, the thought processes can and should be the same. Ruby could be implemented in Ruby, in true Hofstadter-like fashion, though it is not at the time of this writing.
We don’t often think of the etymology of the word way; but there are two important senses in which it is used. On the one hand, it means a method or technique, but it can also mean a road or path. Obviously these two meanings are interrelated, and I think when I say “the Ruby Way,” I mean both of them.
So what we are talking about is a thought process, but it is also a path that we follow. Even the greatest software guru cannot claim to have reached perfection but only to follow the path. And there may be more than one path, but here I can only talk about one.
The conventional wisdom says that form follows function. And the conventional wisdom is, of course, conventionally correct. But Frank Lloyd Wright (speaking in his own field) once said: “Form follows function—that has been misunderstood. Form and function should be one, joined in a spiritual union.”
What did Wright mean? I would say that this truth is not something you learn from a book, but from experience.
However, I would argue that Wright expressed this truth elsewhere in pieces easier to digest. He was a great proponent of simplicity, saying once, “An architect’s most useful tools are an eraser at the drafting board and a wrecking bar at the site.”
So one of Ruby’s virtues is simplicity. Shall I quote other thinkers on the subject? According to Antoine de St. Exupery, “Perfection is achieved, not when there is nothing left to add, but when there is nothing left to take away.”
But Ruby is a complex language. How can I say that it is simple?
If we understood the universe better, we might find a “law of conservation of complexity”—a fact of reality that disturbs our lives like entropy so that we cannot avoid it but can only redistribute it.
And that is the key. We can’t avoid complexity, but we can push it around. We can bury it out of sight. This is the old “black box” principle at work; a black box performs a complex task, but it possesses simplicity on the outside.
If you haven’t already lost patience with my quotations, a word from Albert Einstein is appropriate here: “Everything should be as simple as possible, but no simpler.”
So in Ruby we see simplicity embodied from the programmer’s view (if not from the view of those maintaining the interpreter). Yet we also see the capacity for compromise. In the real world, we must bend a little. For example, every entity in a Ruby program should be a true object, but certain values such as integers are stored as immediate values. In a trade-off familiar to computer science students for decades, we have traded elegance of design for practicality of implementation. In effect, we have traded one kind of simplicity for another.
What Larry Wall said about Perl holds true: “When you say something in a small language, it comes out big. When you say something in a big language, it comes out small.” The same is true for English. The reason that biologist Ernst Haeckel could say “Ontogeny recapitulates phylogeny” in only three words was that he had these powerful words with highly specific meanings at his disposal. We allow inner complexity of the language because it enables us to shift the complexity away from the individual utterance.
I would state this guideline this way: Don’t write 200 lines of code when 10 will do.
I’m taking it for granted that brevity is generally a good thing. A short program fragment will take up less space in the programmer’s brain; it will be easier to grasp as a single entity. As a happy side effect, fewer bugs will be injected while the code is being written.
Of course, we must remember Einstein’s warning about simplicity. If we put brevity too high on our list of priorities, we will end up with code that is hopelessly obfuscated. Information theory teaches us that compressed data is statistically similar to random noise; if you have looked at C or APL or regular expression notation — especially badly written — you have experienced this truth firsthand. “Simple, but not too simple”; that is the key. Embrace brevity, but do not sacrifice readability.
It is a truism that both brevity and readability are good. But there is an underly- ing reason for this, one so fundamental that we sometimes forget it. The reason is that computers exist for humans, not humans for computers.
In the old days, it was almost the opposite. Computers cost millions of dollars and ate electricity at the rate of many kilowatts. People acted as though the computer were a minor deity and the programmers were humble supplicants. An hour of the computer’s time was more expensive than an hour of a person’s time.
When computers became smaller and cheaper, high-level languages also became more popular. These were inefficient from the computer’s point of view but efficient from the human perspective. Ruby is simply a later development in this line of thought. Some, in fact, have called it a VHLL (Very High-Level Language); though this term is not well-defined, I think its use is justified here.
The computer is supposed to be the servant, not the master, and, as Matz has said, a smart servant should do a complex task with a few short commands. This has been true through all the history of computer science. We started with machine languages and progressed to assembly language and then to high-level languages.
What we are talking about here is a shift from a machine-centered paradigm to a human-centered one. In my opinion, Ruby is an excellent example of human-centric programming.
I’ll shift gears a little. There was a wonderful little book from the 1980s called The Tao of Programming (by Geoffrey James). Nearly every line is quotable, but I’ll repeat only this: “A program should follow the ‘Law of Least Astonishment.’ What is this law? It is simply that the program should always respond to the user in the way that astonishes him least.” (Of course, in the case of a language interpreter, the user is the programmer.)
I don’t know whether James coined this term, but his book was my first introduction to the phrase. This is a principle that is well known and often cited in the Ruby community, though it is usually called the Principle of Least Surprise or POLS. (I myself stubbornly prefer the acronym LOLA.)
Whatever you call it, this rule is a valid one, and it has been a guideline through- out the ongoing development of the Ruby language. It is also a useful guideline for those who develop libraries or user interfaces.
The only problem, of course, is that different people are surprised by different things; there is no universal agreement on how an object or method “ought” to behave. We can strive for consistency and strive to justify our design decisions, and each person can train his own intuition.
For the record, Matz has said that “least surprise” should refer to him as the designer. The more you think like him, the less Ruby will surprise you. And I assure you, imitating Matz is not a bad idea for most of us.
No matter how logically constructed a system may be, your intuition needs to be trained. Each programming language is a world unto itself, with its own set of assumptions, and human languages are the same. When I took German, I learned that all nouns were capitalized, but the word deutsch was not. I complained to my professor; after all, this was the name of the language, wasn’t it? He smiled and said, “Don’t fight it.”
What he taught me was to let German be German. By extension, that is good advice for anyone coming to Ruby from some other language. Let Ruby be Ruby. Don’t expect it to be Perl, because it isn’t; don’t expect it to be LISP or Smalltalk, either. On the other hand, Ruby has common elements with all three of these. Start by following your expectations, but when they are violated, don’t fight it. (Unless Matz agrees it’s a needed change.)
Every programmer today knows the orthogonality principle (which would better be termed the orthogonal completeness principle). Suppose we have an imaginary pair of axes with a set of comparable language entities on one and a set of attributes or capabilities on the other. When we talk of “orthogonality,” we usually mean that the space defined by these axes is as “full” as we can logically make it.
Part of the Ruby Way is to strive for this orthogonality. An array is in some ways similar to a hash; so the operations on each of them should be similar. The limit is reached when we enter the areas where they are different.
Matz has said that “naturalness” is to be valued over orthogonality. But to fully understand what is natural and what is not may take some thinking and some coding. Ruby strives to be friendly to the programmer. For example, there are aliases or synonyms for many method names; size and lenght will both return the number of entries in an array. The variant spelling indexes and indices both refer to the same method. Some consider this sort of thing to be an annoyance or anti-feature, but I consider it a good design.
Ruby strives for consistency and regularity. There is nothing mysterious about this; in every aspect of life, we yearn for things to be regular and parallel. What makes it a little more tricky is learning when to violate this principle.
For instance, Ruby has the habit of appending a question mark (?) to the name of a predicatelike method. This is well and good; it clarifies the code and makes the namespace a little more manageable. But what is more controversial is the similar use of the exclamation point in marking methods that are “destructive” or “dangerous” in the sense that they modify their receivers. The controversy arises because not all of the destructive methods are marked in this way. Shouldn’t we be consistent?
No, in fact we should not. Some of the methods by their very nature change their receiver (such as the Array methods replace and concat). Some of them are “writer” methods allowing assignment to a class attribute; we should not append an exclamation point to the attribute name or the equal sign. Some methods arguably change the state of the receiver, such as read; this occurs too frequently to be marked in this way. If every destructive method name ended in a !, our programs soon would look like sales brochures for a multilevel marketing firm.
Do you notice a kind of tension between opposing forces, a tendency for all rules to be violated? Let me state this as Fulton’s Second Law: Every rule has an exception, except Fulton’s Second Law. (Yes, there is a joke there, a very small one.) What we see in Ruby is not a “foolish consistency” nor a rigid adherence to a set of simple rules. In fact, perhaps part of the Ruby Way is that it is not a rigid and inflexible approach. In language design, as Matz once said, you should “follow your heart.” Yet another aspect of the Ruby philosophy is: Do not fear change at runtime; do not fear what is dynamic. The world is dynamic; why should a programming language be static? Ruby is one of the most dynamic languages in existence.
I would also argue that another aspect is: Do not be a slave to performance issues. When performance is unacceptable, the issue must be addressed, but it should normally not be the first thing you think about. Prefer elegance over efficiency where efficiency is less than critical. Then again, if you are writing a library that may be used in unforeseen ways, performance may be critical from the start.
When I look at Ruby, I perceive a balance between different design goals, a complex interaction reminiscent of the n-body problem in physics. I can imagine it might be modeled as an Alexander Calder mobile. It is perhaps this interaction itself, the harmony, that embodies Ruby’s philosophy rather than just the individual parts. Programmers know that their craft is not just science and technology but art. I hesitate to say that there is a spiritual aspect to computer science, but just between you and me, there certainly is. (If you have not read Robert Pirsig’s Zen and the Art of Motorcycle Maintenance, I recommend that you do so.)
Ruby arose from the human urge to create things that are useful and beautiful. Programs written in Ruby should spring from that same God-given source. That, to me, is the essence of the Ruby Way.
About the Author
Hal Fulton has two degrees in computer science from the University of Mississippi. He taught computer science for four years at the community college level before moving to Austin, Texas, for a series of contracts (mainly at IBM Austin). He has worked for more than 15 years with various forms of UNIX, including AIX, Solaris, and Linux. He was first exposed to Ruby in 1999, and in 2001 he began work on the first edition of this book, which was the second Ruby book in the English language. He has attended six Ruby conferences and has given presentations at four of those, including the first European Ruby Conference in Karlsruhe, Germany. He currently works at Broadwing Communications in Austin, Texas, working on a large data warehouse and related telecom applications. He works daily with C++, Oracle, and of course, Ruby.
Hal is still active daily on the Ruby mailing list and IRC channel, and has several Ruby projects in progress. He is a member of the ACM and the IEEE Computer Society. In his personal life, he enjoys music, reading, writing, art, and photography. He is a member of the Mars Society and is a space enthusiast who would love to go into space before he dies. He lives in Austin, Texas.