Standard Deviation: An Interview with C++ Author Nicolai Josuttis
Like this article? We recommend
Like this article? We recommend
Like this article? We recommend
Danny Kalev: The first edition of The C++ Standard Library: A Tutorial and Reference was published in 1999, a year after the ratification of the first international C++ standard known as C++98. Can you highlight the main differences between the first edition and the forthcoming second edition of your book?
Nicolai Josuttis: You know, I didn’t follow the standardization process of C++11. At the end of 2008 I looked first into the new standard by comparing the C++98/03 versions of classes, such as pair and vector, with their new versions. I was shocked. I had trouble understanding what I found: “What the hell does && mean in declarations?” So if you ask me about the difference, my first answer is: Everything is different! The way you write simple programs and the way you define complicated classes have changed dramatically. C++11's pair<>, for instance, doubled the number of lines.
Nevertheless, the changes go in the right direction. C++11 consequently focuses on the power of C++ -- performance. However, it still has the drawback of making it even harder for programmers to design good classes. Well, it might not be harder, if you know all you have to know; but it’s harder to know all you have to know, now. To some extent, C++11 is a new language, and my new edition simply reflects this change, covering both the new C++ programming style and new classes.
Danny: What inspired you to write a second edition of your successful book? Was it readers' demand? The new C++11 standard? Or something else?
Nicolai: A combination of both: there were persistent requests to update the book to cover C++11.
Danny: What is the best way to master the C++11 Standard Library for a programmer who has been using C++ but needs to catch up with the new changes in C++11? Which aspects of the C++11 Standard Library are the most important for the typical C++ programmer?
Nicolai: Buy my book ;-) Which aspects are most important depends on the programmer and the problem domain. Of course, concurrency and multithreading support is something a lot of programmers will be interested in. But others might care for random number generation and distribution (as described in section 26.5 of the C++11 standard). For the ordinary programmer, the most important changes probably lie in the core language -- some of which change the way you use the Standard Library dramatically.
Danny: Templates are the heart and soul of the Standard Library -- containers, algorithms, and iterators are built upon templates. Are the principles of traditional OOP such as inheritance, virtual functions and polymorphism no longer relevant in C++ today?
Nicolai: A couple of possible answers come to my mind here. First, templates are also a way to program polymorphism (at compile time instead of runtime). The second possible answer is that C++ is no pure object-oriented language. Neither is it a pure generic language. The power of C++ comes from the ability to combine both. And the third possible answer is simply: “yes”.
Danny: Let's talk about core C++. What are the advantages, if any, of lambda expressions? Do you predict that they will become widely-used in typical programmers' code, or will they remain restricted primarily to C++ libraries? Are there any potential disadvantages that lambda might have?
Nicolai: I admit I never use algorithms and function objects when I program in C++. But to be fair I have to say that these days I don’t program a lot. Nevertheless I know that I am not alone. In my opinion, with lambdas I have for the first time the ability to write code that is both fast and readable using algorithms. That’s awesome. Nevertheless lambdas have drawbacks as well -- they provide local behavior (as do function objects), but they don’t maintain a state between multiple calls. In this respect, function objects are still a better choice.
Danny: Rvalue references are the most significant change in core C++11 [you can read more about this feature here]. Have you had a chance to use them in real-world applications and assess their usefulness? What are their main advantages?
Nicolai: Sorry, but my projects usually don’t use C++ these days. Again I have to admit that my major knowledge of C++11 is based on what I learned and tried out while writing the second edition. However, I tried a lot of stuff and discussed features with all the experts in the community. So, it’s not a book by a novice; it’s a book by an “experienced novice.” ;-)
For this reason, I am not sure that rvalue references are the most important new feature. I agree that it is a key feature to further emphasize the major advantage of C++, i.e., performance. However, I know how hard it was for me to motivate and explain move semantics and rvalue reference in the book. So for me, the most significant changes in C++ are: auto, range-based for loops, and initializer lists. These three features simplify at least my small example programs to such an extent that I wouldn’t want to miss these features anymore.
Danny: Generally speaking C++11 is a compromise. Certain features (e.g., an automated Garbage Collector, thread pools, a socket library) didn't make it into the standard -- they yielded to pressing deadlines. Do you believe that the current C++11 standard is a good standard after all? Does it live up to the expectations that you had back in 2003?
Nicolai: After writing this book I can say that I am a big fan of C++11. In combination, all these little simplifications and improvements are huge. However, again features and components that made it into the language are sometimes hard to use. And error messages caused by template instantiations are still a big problem [they're hardly decipherable, DK]]. In my opinion, we [the standards committee] don’t care enough for simplicity. But this is not a complaint about the people in charge of the standardization. They do an incredibly hard and good job. With C++, we see the typical result of all software under maintenance. A lot of problems result from its heterogeneity combined with the need to remain backward compatible.
Danny: What does C++11 still lack? Which features and libraries would you like to see in the next C++ standard?
Nicolai: A file system library is definitely missing. Also the support for different character sets like UTF8, UTF16, and Unicode is still poor (while I have an example to convert UTF8 into UTF16 in the book, a lot of other stuff of this kind is still missing). I DON’T miss concepts. In fact, I had decided not to present concepts in the second edition before the concepts concept was removed from C++11 [you can read more about the history of this controversial feature and its removal from the standard here].
Danny: Tell us a bit about yourself. You have been an active member of the C++ standards committee for many years. In which aspects of the C++11 standardization endeavor have you been involved? Which libraries have you co-designed?
Nicolai: I am not one of the leading experts. Almost all people in the committee know C++ better than me. I am the one who asks and cares for the right quality. For both C++98 and C++11 I jumped in late and have since focused on making things more consistent, cleaning up flaws, and clarifying behavior. The only thing that comes from me is the first version of the class template array<>, which however was the result of a theft of a class c_array that Bjarne Stroustrup had described in The C++ Programming Language.
Danny: Recently, there seems to be renewed interest in C++ ("a C++ renaissance" as some dub it). Is the programming world sobering up from the unrealistic expectations from other programming languages or has C++ simply gotten better? On a more personal note, what's your favorite programming language, if any? Which types of projects have you been involved in recently?
Nicolai: Programming is not the core task in my current jobs and contracts. As an expert for SOA and system integration my main concern these days is to support projects and companies to establish techniques and processes that realize behavior ("business processes") distributed over multiple systems inside and between multiple companies [read more about SAO here]. Thus, my topic is the maintenance of system landscapes rather than system development.
C++ is not my favorite programming language. In fact, I have no favorite language. My last fight about programming languages ended 15 years ago. The question is, which problem do we have to solve, and in which context? Depending on the context I mostly use the old UNIX shell scripts including sed and awk, Java, C++, and Excel to "program" behavior (although Excel might not be considered a programming language, for specific problems it IMO definitely has all the power of a programming language).
If performance really matters, then I don’t see any alternative to C++. However, you pay a big price for using it. Due to the complexity of C++, the support for the ordinary programmer is incredibly bad compared to other languages, and that’s a major drawback. But all that helps to sell my books. You can say that C++ creates jobs, and that’s important in these times.