This book collects arguments for using Eiffel.
Even though we in the Eiffel community know that Eiffel is the most advanced object-oriented software development framework on this planet, that fact may not be so obvious to everyone else. Consequently, we felt that it would be helpful to provide some brief documents that spell out the reasons why the people who have chosen Eiffel have been so successful.
You may find the information helpful if you are simply curious about Eiffel, if you are actively looking for a better development framework, or if you are busy trying to convince others to try Eiffel.
If you are both curious about Eiffel and in a hurry, take a couple of minutes to read these facts about Eiffel. If anything here seems too good to be true, please suspend your disbelief. Press on to the more detailed documentation for the rationale, and our success stories for the evidence behind these facts.
Software produced with Eiffel is typically:
Eiffel consists of:
For a more detailed overview see An Invitation to Eiffel .
(After an article in the special Eiffel issue of the Journal of Object-Oriented Programming, May 1996. )
Over its ten-year life Eiffel has evolved into one of the most usable software development environments available today. Other articles discuss its theoretical contributions. This one addresses a more mundane subject: how practical software projects can benefit, today, from the power of Eiffel. In so doing it will largely rely on published assessments from both Eiffel users and book authors. In fact a quotation from one of the best-known books in the object-oriented field -- Object-Oriented Modeling and Design by James Rumbaugh and his colleagues, the text that introduced the OMT O-O analysis method -- provides a good start:
Eiffel is arguably the best commercial object-oriented language available today. [Jim Rumbaugh et al. in Object-Oriented Modeling and Design, Prentice Hall 1988].
First we should define what the word "Eiffel" means. If you are thinking "a programming language" you are not wrong (and the preceding quotation shows that you are in good company). The programming language is indeed the most visible part, but it is only a reflection of something broader: a comprehensive approach to the production of quality software. As Richard Wiener wrote:
Eiffel is more than a language; it is a framework for thinking about, designing and implementing object-oriented software. [Richard Wiener in Software Development using Eiffel: There is life other than C++, Prentice Hall, 1995.]
The Eiffel approach includes a method (a "methodology", as it is sometimes called) based on a number of pervasive ideas such as Design by Contract, seamlessness, reversibility, rigorous architectural rules, systematic use of single and multiple inheritance, static type checking and several others. Besides a method and a language Eiffel also means powerful graphical development environments, such as EiffelStudio, available across a wide number of industry-standard platforms and supporting analysis and design as well as implementation, maintenance and evolution.
The language itself, indeed (which Wiener calls an elegant and powerful language for object-oriented problem solving") is not just a programming language but extends to the phases of system construction that both precede and follow implementation. This is sometimes hard to accept if you have been raised in the view that software development must involve a sequence of separate steps; that one should initially use an analysis method and then at some point switch to a programming language, with perhaps a design method in-between. This view is detrimental to the software process and to the quality of the resulting product, as it does not support the inevitable back-and-forth hesitations that characterize real software development.
Wisdom sometimes blooms late in the season. However careful you may have been at the analysis stage, some great ideas will hit you - or your implementers - past the point at which you thought you had all the specifications right. Why renounce the benefit of such belated but valuable ideas? Eiffel and the associated Business Object Notation approach to analysis and design accommodate them naturally, by providing a single conceptual framework from the beginning to the end of the process.
Here Eiffel does not have much competition. The most bright-eyed Smalltalk or C++ enthusiast would not seriously claim that one can do design, let alone analysis, in his language of choice. And users of any of the popular O-O analysis notations know that at some stage they must stop working on their model and move on to the implementation in some programming language. Eiffel is unique in helping you for all of these tasks, without ever introducing the impedance mismatches that characterize other approaches.
As a reviewer wrote:
As a design language, Eiffel continues to be a better model for object- oriented programming than Ada. It is even better than the new Ada 9X standard. [Richard Riehle in HP Professional, October 1994, A Tour of Eiffel.]
The next few sections give a glimpse of the technical contributions of Eiffel, or more precisely of what other people have written about them. But of course the best technology in the world requires infrastructure and support to succeed.
Eiffel has plenty of these. It has been around for more than a decade. Eiffel is available from several commercial and open-source providers. The number of licenses sold is in the tens of thousands. Reusable library classes are in the thousands.
The platforms covered range from Unix (all of Unix, the famous and the arcane) and Linux to OpenVMS, OS/2, Windows 3. 1, Windows NT, Windows 95/98.
Particularly impressive is the growth of Eiffel usage in education. Eiffel is quickly becoming the language of choice for teaching modern software technology, including, increasingly, introductory programming. A dozen of excellent textbooks are now available from Prentice Hall, Addison-Wesley, Macmillan and others, with about as many announced just for the coming months.
It is not just the professors who like the approach. Here is just one typical comment on student reaction, from an institution (Rochester Institute of Technology) having adopted Eiffel as its first-year introductory language on a massive scale:
We were pleased to discover many of our more skeptical students turning around and admitting that Eiffel was a "fun" language in which to work. [Jim Heliotis in the Proceedings of OOPSLA 95, Experiences teaching objects: A new curriculum for computer science students.]
A Computer World article confirmed the need for Eiffel in training the high-powered software professionals of tomorrow. Quoting Amy Cody-Quinn from Management Recruiters International, the journalist writes
There is a big problem with people who say they know C++ - but they don't really know how to do objects. If they have Eiffel on their resume, then we know they really have the proper understanding of what they are doing. [Leslie Goff in ComputerWorld, Object Edge, December 18, 1995.]
But it would be a mistake to think of Eiffel as an academic tool. A little-known fact is that some of the biggest object-oriented projects ever undertaken (at least the successful ones - other O-O languages have had their share of large-scale failures) are being done in Eiffel. The hot areas at the moment are banking and the financial industry (in particular some very large derivative trading and investment management systems), telecommunications, health care. These are all areas in which all that counts in the end is quality and time to market, so that projects need to select the best technology available. Quoting from an article by Philippe Stephan, the system architect of such a project (Rainbow, a major derivative trading system built with ISE Eiffel):
We evaluated three major object-oriented languages for the project - Smalltalk, C++ and Eiffel - and chose Eiffel. [...] Rainbow currently comprises over 400,000 lines of code, for about 3000 classes. [Current figures are way over these mid-1995 counts. ] The developers feel very productive. This was confirmed when Rainbow's financial backers brought in object professionals to audit the project. The auditors evaluated the project during July 1994 and were impressed with the productivity of the Rainbow development group. [Philippe Stephan in Object Magazine, July-August 1995, Building financial software with object technology.]
The development group in question is remarkable for being made only for a third of software professionals. The others are professionals from other disciplines (such as trading and financial analysis), who, Stephan writes,
can express business concepts in Eiffel because they can focus on design and implementation, rather than struggling with memory management problems and debugging.
The result has received lavish praise from such publications as ComputerWorld and analysts:
Industry experts briefed on Rainbow said they were impressed with the results. CALFP is "progressive" in [...] committing the organization's mission-critical systems development efforts to this architecture, said Richard Crone, senior manager of financial services at KPMG Peat Marwick in Los Angeles. "What's unique here is that [CALFP is] delivering this system end-to-end using object-oriented technologies", said Henry Morris, a research analyst at International Data Corporation (IDC) in Framingham, Mass. [Thomas Hoffmann in ComputerWorld, May 8, 1995, Object- Oriented financial package tames transactions.]
Along with these Eiffel mega-projects, you will also find myriad smaller endeavors. Many consultants, in particular, have found for themselves the key competitive advantage that they can gain from Eiffel's excellence. In ensuring this spread of Eiffel throughout the industry, the benefit of cheap yet complete environments such as EiffelStudio for Linux has been immeasurable.
Also crucial to the development of Eiffel has been the neutral status of its definition, now controlled by a consortium of vendors and users, NICE (the Nonprofit International Consortium for Eiffel). NICE has already produced a library standard and expects to produce soon the language standard that should shortly thereafter enjoy a smooth ride through ANSI and other international standards bodies.
The pace of Eiffel history has been accelerating in the past few months. This has been picked up by many journalists. As Dan Wilder wrote:
With an open specification for both the language and the kernel libraries, and support from multiple vendors, Eiffel now stands poised to take off. [Dan Wilder in Linux Journal, June 1995, Introduction to Eiffel.]
Eiffel - the method, the language, the environment - is based on a small set of goals, addressing the crucial needs of software quality and productivity. Quoting from an article in Byte magazine:
Developers who want an object-oriented language that adheres to the keystone principles of software engineering need look no further than Eiffel. [Peter Varhol in Byte Magazine, February 1996.]
or, as Steve Bilow wrote in a review of Eiffel Software's Melting Ice compiling technology (which he calls "an outstanding marriage between portability and development speed"):
Eiffel was designed precisely for the purpose of enabling software developers to deliver high quality, reliable, efficient, reusable code. [Steve Bilow in The X Journal, The Eiffel alternative, July-August 1995.]
The first goal is reliability. No other approach available today has made the effort to give developers all the tools that they need to produce bug-free software - software that will run without bugs the first time around. Crucial in this effort is the presence of static typing ( real static typing, not "a little bit typed" as in those languages that still keep C-like type conversions); assertions and the whole mechanism of Design by Contract, on which more than one Eiffel developer has said "this has changed my life" by enabling him or her to specify precisely what the software should do, and to track at run time that it does it; disciplined exception handling; automatic garbage collection, which eliminates a source of horrendous bugs in C++-based environments (and a large part of the code, tedious and error-prone); a clean approach to inheritance; the use of dynamic binding as the default policy, meaning the guarantee that all calls will use the right version of each operation; and the simplicity of the language design, which enables Eiffel developers to know all of Eiffel and feel in control.
The role of assertions and Design by Contract is particularly important here. According to the Journal of Object-Oriented Programming:
The contribution of Eiffel is significant: it shows how invariants, preconditions, and postconditions can be incorporated into a practical developer's view of a class. Wider use of Eiffel [...] will encourage a greater use of simple but powerful mathematics during development. [Richard Mitchell et al. in Journal of Object-Oriented Programming, July-August 1995, As-a: a relationship to support code reuse.]
The second goal is reusability. This has become a catchword, but Eiffel is the only approach that has taken this requirement and its consequences all the way to the end. Quoting Roland Racko:
Everything about [Eiffel] is single-mindedly, unambiguously, gloriously focused on reusability - right down to the choice of reserved words and punctuation and right up to the compile time environment. [Roland Racko in Software Development, December 1994, In praise of Eiffel.]
Eiffel benefits here from being a simple ("but not simplistic", writes Racko) and consistent design, not a transposition from older, unrelated technology. Beyond the language and the environment facilities (such as precompilation), the crucial help to reusability is of course the presence of thousands of high-quality library classes, such as, in EiffelBase (a "Linnaean approach to the reconstruction of software fundamentals"), EiffelNet for client-server communication, EiffelStore for relational and O-O database manipulations, EiffelLex and EiffelParse for lexical analysis and parsing, EiffelMath for object-oriented numerical computation, EiffelVision for portable graphics, WEL (the Windows Eiffel Library) for Windows-specific graphics, EiffelWeb to process forms from the Web, and many others. Not even mentioning quality, the result is probably the biggest repository of object-oriented components available elsewhere. The care that has been applied to the production of these libraries also has considerable pedagogical benefits: the way people learn Eiffel is by learning the libraries - first to use them, then to adapt them if necessary, then to write their own software.
Part of the single-mindedness mentioned by acko is the emphasis on abstraction. In contrast with, say, Smalltalk, you do not read the source code of a class when you want to use it. This may be fine for a couple dozen classes, but not for a large, powerful library. Eiffel introduces the notion of short form: an abstract version of the class, keeping only the interface information including assertions. This is an ideal tool for documenting classes but also for discussing designs and presenting them to outsiders - managers or customers - who need to know what is going on without getting bogged down in the details.
Let me mention just one of the unique reusability-supporting features of Eiffel, without which it is, in my experience, impossible to have a long-term reuse effort. Racko again:
The language's designer [...] recognized that no reusable library is ever perfect and, thus, that libraries are always in flux. So he built a kind of version-control system into the language. Specifically, there are language elements to demarcate obsolete code that is, however, still being supported. When these elements are referenced by someone unaware of such code's obsolescence, the compiler will issue a warning at compile time about the impending doom that awaits persons who continue the referencing.
It is this kind of detail that can make or break the success of reuse in a company.
Next comes extendibility. With Eiffel, modifying software is part of the normal process. As Philippe Stephan writes of the external audit of his project:
The auditors rated the responsiveness of the development team as very high.
Chief among the method's support for extendibility is the careful design of the inheritance mechanism. Unlike Smalltalk, which is fatally limited by the absence of multiple inheritance, the Eiffel approach fundamentally relies on multiple inheritance to combine various abstractions into one. As Dan Wilder notes:
Most object-oriented languages do not attempt multiple-inheritance. The literature is full of elaborate explanations why. This is sad. Eiffel demonstrates that multiple inheritance need not be difficult or complex, and it can also yield some quite practical results.
The approach also enforces a strict form of information hiding, which means that a module (a client in Eiffel Design by Contract terminology) that uses another's facilities (its supplier) is protected against many of the changes that can be made later on to these facilities. This is essential in preserving the coherent evolution of a large system - and the sanity of its developers.
Performance is almost as much an obsession in Eiffel as reusability. The software field is still, and will remain for a long time, largely driven by performance considerations. (Do not believe anyone who says that speed does not matter. If we get faster computers, it is to do things faster and especially to do more things - not to use more CPU cycles to run the same old applications at the same old visible speed. )
There is no reason whatsoever to leave the mantle of efficiency to the proponents of machine-oriented languages such as C/C++, or to follow the path of Smalltalk which sacrifices performance to object orientation. With Eiffel you can have the best of both worlds. Thanks to a performance-obsessed language design and ten years of research and competition on compiling algorithms, the speed of Eiffel-generated code (in such modes as what is known as "finalization" in Eiffel Software's implementation) is as good as that of hand-produced C code, or better.
Software producers should stand up to their ideas. That is what we do at Eiffel Software: apart from the run-time engine (a few thousand lines of C), all of our software - thousands of classes, hundreds of thousands of lines - is written in Eiffel, and it runs fast. Typical of the situation is a recent incident with the EiffelLex library: it still had a few C elements, remnants of an earlier design. We rewrote them in Eiffel - for a 30% performance gain.
Why these gains? The answer is simple. The /C++ approach of doing everything by hand, under tight programmer control, works well for small programs. Similarly, a good secretary has no equivalent for keeping one person's records. But in the same way that no humans can match the performance of a computer for managing, say, the records of a bank or a city, no programmer can beat a sophisticated Eiffel compiler for optimizing a large program. Against the automatic application of inlining, static binding, memory management and other optimizations, the human does not stand a chance.
To have one's cake and eat it also means not to have to choose between run-time and compilation-time performance. For programmers used to the contrast between a Smalltalk-like style of rapid turnaround and the interminable edit-compile-link cycle of most compiled environment, the following comments by Dan Wilder (in a separate article) will be a shock:
EiffelStudio uses "melting ice technology", which allows incremental changes to run in an interpreted mode. Only modified classes are recompiled. Changing one class and clicking the Melt button caused only a few seconds of compilation. [...] My test application took 20 seconds to compile from scratch in "melt" mode. [Dan Wilder in Linux Journal, September 1995, Two Eiffel implementations.]
Steve Bilow provides further explanations:
Based on the observation that software development is an iterative process which is usually focused on constructing systems from code modifications, the folks at Eiffel Software have developed something that they call "Melting Ice Technology". Essentially, this means that when you make a [change] and you want to try it out, you simply "melt" it into the system. You don't need to regenerate a bunch of C code, so your changes are integrated into the system proportionally to the amount of code changed. Even in C and C++, `make' still has to relink.
What this also indicates in passing is the technology choice made by Eiffel Software's implementation and all current implementations: using C as the portable implementation vehicle. By going through C, the compilers gain efficiency and portability. This also makes Eiffel one of the most open environments around; in contrast to the self-centered view that predominates in Smalltalk, Eiffel software is born with a sociable attitude, ready to interface with all kinds of other software written in C or other languages. This, needless to say, is a key to the success of realistic applications.
A good way to think about Eiffel - the seamlessness of it, the insistence on getting everything right, the conviction that software should be beautiful in and out, specification and implementation - is this little anecdote stolen from Roman Jakobson's Essays on General Linguistics:
In a far-away country, a missionary was scolding the natives. "You should not go around naked, showing your body like this!". One day a young girl spoke back, pointing at him: "But you, Father, you are also showing a part of your body!". "But of course", the missionary said in a dignified tone; "That is my face". The girl replied: "So you see, Father, it is really the same thing. Only, with us, everything is the face".
So it is with Eiffel too. Everything is the face.
Hundreds of thousands of people have now been exposed to Eiffel through books, through courses, through hearing about other people's successes, and most importantly (for the most fortunate among them) through using it to complete a project on time, within budget, and to the satisfaction of their customers. Shouldn't your next project use Eiffel too?
Often, when we speak about Eiffel to prospective users, we hear them repeat misinformation about the method, the language, or the tools. Most of the time, the stories are familiar to us … and untrue. Here are a few of the myths that we hear most often, as recounted and debunked by Eiffel Trainer Hal Webre in his introductory online presentation to the series entitled "Where Eiffel Fits".
Recently, I was offered the opportunity to speak to a local technology group about Eiffel for Microsoft .Net. The leader of this group is part of a small commercially-oriented software development company. Concerning Eiffel, he said, “All I know about Eiffel is that it’s an academic language.”
We hear that one a lot … and it’s wrong … in two unrelated ways.
First, as you should know by now, Eiffel is a framework for software development. It has a full-lifecycle development method. The Eiffel method is supported by a notation we call the Eiffel programming language. The notation just happens to be designed such that when it contains sufficient detail, it can be compiled into a running software system. Additionally, the method and language are supported by a set of tools including an interactive development and compiler. So to refer to Eiffel only as a "language" is to do injustice to the complete framework of which the language is only one part.
Secondly, I’m not sure what “academic language” means exactly, but if it means only used in an academic setting, or not fit for practical work, then this could not be farther from the truth. It is true that Bertrand Meyer who developed the original Eiffel concepts has academic background and is well-respected in the academic community. It’s also true that many of those Eiffel ideas evolved from work that was done by other academic computer scientists and mathematicians. And it’s true that many colleges and universities use Eiffel to teach the best practices of software development.
But Eiffel is also used successfully in many commercial and government endeavors. If you have any doubts, pay a visit to eiffel.com and check out the success stories and customers testimonials.
Occasionally we’ve heard people say that Eiffel is only suitable for building “toy” systems.
This is similar to the "academic language" argument and is just as false.
In actuality, we see Eiffel being used often in situations in which other technologies fail. If anything it is the other commonly used technologies that tend to break down under stress.
We see Eiffel being used instead of other technologies for systems in which scalability and reliability are essential.
One of our customers is an organization that has developed a payroll system using Eiffel that every week pays over two hundred thousand employees in twenty thousand different institutions … the people in this organization would assure you that Eiffel is indeed industrial grade.
The answer to this one depends a bit upon your frame of reference. Some time ago Microsoft estimated that there were twenty million Visual Basic programmers world wide.
If this is true, then relatively speaking, we have to admit that Eiffel’s market share is considerably smaller than that of Visual Basic.
Despite that, it’s not correct to say that not many people use Eiffel. Eiffel licenses world wide number in the tens of thousands. If you use Eiffel, you are not alone. These license-holders have formed a lasting worldwide quorum. Many world-class organizations are committed now, and will continue to be committed to the Eiffel framework. There is support through your maintenance contract with Eiffel Software. Help and information are available online in the form of the Eiffel Software users’ list and websites like EiffelRoom.
Eiffel Software's dual licensing model gives developers the opportunity to learn Eiffel without a great initial financial commitment.
So, don’t worry about it, plenty of people use Eiffel … and those numbers are increasing constantly.
Through the years some potential Eiffel users have expressed to us a concern that if they embrace Eiffel, that they may not be able to find a sufficient number of qualified developers.
This is of course incorrect.
First, almost all Eiffel people were proficient in some other technology before they became Eiffel people. It turns out that this really works to your advantage. You see, Eiffel people want to stay Eiffel people. So an Eiffel programmer on the job market will be searching for an Eiffel position first, and would probably rather have a root canal than to go back to working in his or her previous technology.
Additionally, it is important also to understand that Eiffel developers are easy to create. Because Eiffel is simple, clean, and elegant, it doesn’t take long to get people going with it. I teach a five-day course that contains fifteen programming exercises. Each time I’ve taught the course, almost every student has finished every exercise. Students leave with a good foundation for how to begin saving time and money for their organization by constructing quality software with Eiffel. These people can be fluent in Eiffel in as little as a couple of months. This can be contrasted with the other extreme ... a well-known Microsoft Windows expert told me a couple of years ago that he estimates it to take 5 to 7 years to become truly fluent in C++/COM programming on Windows. Programmers who are proficient in other technologies often experience Eiffel as a breath of fresh air.
I think the first time I heard this one, it was about 1989.
And of course, I’ve heard it many times in the years since.
And of course, it’s not true.
Eiffel is more complete and functionally superior in most ways to every other commercially viable software development technology … and there are enough people around who recognize this (that quorum of users I mentioned earlier) to ensure that Eiffel will be around for a long time to come.
It’s possible that twenty-five years from now, there will be a significantly better software engineering idea … but certainly there hasn’t been anything that’s come close since Eiffel’s original design in 1985. In most areas, other technologies are playing “catch-up” to Eiffel.
Besides, Eiffel constantly implements refinements and new capabilities with minimal impact on existing software. Void-safe programming is an excellent example of this.
You can get a feel for this issue by watching this video on Microsoft Developers Network Channel9. Here you'll see Emmanuel Stapf, an engineer at Eiffel Software, speak with Mads Torgersen, one of Microsoft's C# language designers. You'll hear how Eiffel stays fresh and continues to set a technological standard worthy of the aspirations of other technologies.
So, don’t worry about it. Eiffel will be here.