Slashdot Mirror
Objective-C Overtakes C++, But C Is Number One
mikejuk writes "Although the TIOBE Index has its shortcomings, the finding that Objective-C has overtaken C++ is reiterated in the open source Transparent Language Popularity Index. The reason is, of course, that Objective-C is the language you have to use to create iOS applications — and as iPads and iPhones have risen in popularity, so has Objective-C. If you look at the raw charts then you can see that C++ has been in decline since about 2005 and Objective-C has shot up to overtake it with amazing growth. But the two charts are on different scales: if you plot both on the same chart, you can see that rather than rocketing up, Objective-C has just crawled its way past, and it is as much to do with the decline of C++. It simply hasn't reached the popularity of C++ in its heyday before 2005. However the real story is that C, a raw machine independent assembler-like language, with no pretense to be object oriented or sophisticated, has beaten all three of the object oriented heavy weights — Java, C++ and Objective C. Yes C is number one (and a close second in the transparent index)."
Objects simply allow for an efficient programming structure for large software. That's the main reason. The real debate is about how far this object orientation should go. There are people like me that are of the opinion: use objects when necessary for structure. Others on the other hand will wrap anything in layers of objects. Dynamic allocation isn't strictly necessary for object support (see C++ to know why). It's just that most object oriented languages now also want to use polymorphism, at that point dynamic allocation is necessary cause it's near impossible for the compiler to predict what'll happen. But it's a rather pointless debate in the long run. To each his own as they say.
It's like the static vs dynamic linking debate that you sometimes hear. There's no real valid answer to that one either, it's a best guess on what'll lead to the best performance. With dynamic linking you don't need to load all the libraries at the start, on the other hand with static linking you don't need to call up the linker each time a library is loaded, and so on... My main advice: stay out of it. There's no real valid answer to these sort of things.
So strange... I find Ayn Rand completely guilty of the very same romantic notions that got the founders of Communism (she so despised) into so much hot water. Perhaps its true what they say about choosing your enemies well. Both presumed that the underlying greatness and magnificence of the human spirit either as a society or as a specific productive individual would prove the guiding light for humanity. In fact humanity has shown precious few guiding lights and for the most part, we are little descended from our primate ancestors. This isn't to say that we aren't capable of transcendence, simply that you can't depend on that to build a social or philosophical framework.
Design the system that demands human transcendence, inspires greatness, and puts strict limits to personal power and responsibly accounts for the grosser of human foibles and frailties, and you'll have a winner. We had that system in the form of checks and balances, until the "Randian" among us began to systematically dismantle those very defenses against our poorer natures, beginning in the 80s. Up until then, we had the time and means to look at the future we wanted as a society, not just a few social (read financial) elites, and strive towards that future wisely and with due consideration. Now we're in a kettle of fish. Those elite have proven to be every bit as ignorant, self obsessed/serving and foolish as everyone else and they've squandered the future on extra McMansions, expensive cars and yachts, and the virtual hijacking of our society.
C is a great language. You can't any closer to bare metal without slugging assembly around, and as we move to more and more intelligent particles infiltrating everything from household appliances to ubiquitous sensors in the roads we drive on, you better believe that C will bring consciousness to the dross matter that surrounds us. I can only hope, that we can put aside our prejudices (not only racial, but societal), and begin to replace belief systems with educated inquiry, and treat the future with our intelligence rather than our primate predilections. It is the only hope I can see for a future worth living in.
What the fuck? Why the fuck would I subclass a button just to make it blue? That's just data, and damned trivial data at that. If your button object doesn't already have some mechanism for dealing with that data, it sucks and I'm using a different object, not yours.
Instead of having to code buttons from scratch, you sub class them...
No no no no NO. Goddamnit NO. Fucking Java. Motherfucking Javascript. They've ruined a generation of programmers.
Subclassing is the LAST thing you should be doing. The very last. First you should be using the customization features built in to the object, and using them directly on an instance of that object. Set the blue color on the Button class and be done with it. If that's not sufficient, use object composition. Most of the time, your object is NOT a Button. It's a something that needs to have a button. Only as a last possible resort do you subclass Button, and you'd damn well better be writing an object that still is-a Button. If you're not, you've done it WRONG.
Anyone who complains about STL portability issues clearly hasn't written a line of C++ code in the past decade. STL is the single most portable and consistent library behind the C standard library. And don't even get me started about efficiency psuedo-arguments. std::sort outperforms qsort by a huge margin on most platforms.
And--I know I'm going to be stoned for this--Linus =/= God.
So, you think that is object-orientation? Oh boy.
From wikipedia: "Object-oriented programming (OOP) is a programming paradigm using "objects" - data structures consisting of data fields and methods together with their interactions - to design applications and computer programs."
The GP's method certainly qualifies. Just because it doesn't include all the sugary syntax or features that are included in your favorite so-called "OOP language" doesn't mean that you can't do object-oriented programming in C.
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
An OO language breaks the organization of THINGS in a very natural way for western thinkers.
Yes. And that right there is a subtle trap.
The first problem is that the "tree of subclasses" organisation, while on the surface seeming natural, is not in fact an accurate description of real taxonomies found either in nature or in large software projects. Especially so if the "software" includes business data. It turns out there are an awful lot of platypuses in the real world, things which simply don't fit neatly into the tree.
For example, a classic "toy" example often used in the OO analysis world is a database of employees. Lets see, we have managers, and we have workers. Great, we can subclass those! We'll have an abstract Person class, then personWorker and personManager who are subclasses of Person. Instantiate Jack Smith as an instance of class personWorker. Problem sol - um. Wait. Jack just got promoted from a worker to a manager. Crap. Can our OO system of choice handle dynamically changing an object's class during its lifetime? No, it enforces strict classing, so it can't. Oops. No problem, we'll delete Jack and recreate... oh. His entire work history was attached to that object, linked by opaque reference and not by name or staff ID, and now it's all gone forever. Double crap. Oh well. He's left the company anyway, and now he's come back as a private contractor. We'll just make him a new personContractor. Easy. Yeah, wait, now we're dealing with him also over in the billing system as a personSupplier. But wait, there's more, he just bought some stuff from us, so he's also a personCustomer! Now he's three classes at once! The universe has gone crazy!
Most real OO systems "solve" this problem by either not doing inheritance here at all - therefore completely invalidating the "OO is about inheritance" line - or duplicating the data in multiple objects - thereby invalidating the "OO is about modelling the business domain directly" line. But at this point we're really starting to lose most of the advantages of OO entirely.
But there's a second, even more subtle problem: although OO usually uses "class" as a synonym for "type", it turns out that subclasses are NOT at all the same thing mathematically as a subtype. (Because you can override the behaviour of a class, meaning its behaviour is now not a strict subset of its superclass, but can also be a superset.) In fact there's no really sensible definition of "subtype" at all - Liskov substitutability requires that you define a context within which you want to limit your idea of "equality", and over the scope and lifetime of a sufficiently large software system, that context is going to change radically. So there goes all your type safety. Add in runtime reflection (which was a fundamental principle of Smalltalk, the first OO system, but seems to be an optional add-on recently tossed haphazardly back into the modern variety) and things get even more confused.
And finally, even the idea of typing can become a third subtle trap. Even if you could (which you can't in the real world) restrict your software system to a neat tree of subclasses corresponding exactly to strict subtypes in a glorious Platonic universe - if you look at your code carefully, you find out that your class/type structure, no matter how strict and clever you make it, doesn't actually tell you anything about the behaviour of your objects. It only tells you the calling signature. That you've defined an addOne method in all your IncrementableByOne class structures doesn't mean that any of those subclasses actually have to implement int addOne(int X) as returning X plus one - just that they receive and emit an integer. So after all your compile-time declarations, you've gained a whole lot of not much at all, and you have to implement a whole testing harness apparatus to do by hand what your compiler initially promised it will do.
tl;dr: Just like (insert a political philosophy you dislike), OO is a big idea, a seductive idea, but not actually a correct idea. And attempting to apply thoughtlessly will cause pain.
You are not a brain: http://books.google.com/books?id=2oV61CeDx-YC
I think that falls under "I hope to the great FSM that I never have to maintain your code."
-1 overrated isn't the same thing as "I disagree".
Linus is a pretty smart kernel programmer & architect who programs almost exclusively in C. It's no surprise that he doesn't even want to grok C++.