I've just gotten answers to this question which, at the bottom line, tell me: "Doing X doesn't make sense since it would make you pay for things you might not use."
I find this maxim difficult to follow; my instincts lean more towards seeing what I consider clear semantics, with things defined "in their place". More generally, it's not immediate for me to realize what the hidden costs and secret tariffs would be for a particular design choice?.
Is this covered by (non-reference) books on C++? Is there someplace relevant online to better enlighten myself on following this principle?
In the case you are presenting it is not as general a statement as it seems.
Doing X doesn't make sense since it would make you pay for things you
might not use.
This is merely a statement that if you can, avoid using virtual functions. They add overhead to the function call.
Virtual functions can often be redesigned by using templates and regular function calls. One std:: example is std::vector. In Java for instance a Vector implements interfaces to be usable in algorithms. Accomplished by virtual function calls. std::vector uses iterators.
Despite the question being overly broad and asking for off site material I think it is interesting enough to deserve an answer. Remember that C++ was originally just "C with classes" and it is still possible today to write what is basically C code without using any of the nice abstractions that C++ gives you. For example if you don't want the cost of using exceptions then don't use them, if you don't want the cost of RTTI (virtual functions) then don't use them, if you don't want the overhead of using templates... etc.
As for resources, I'm going to break the rules and recommend Game Programming Patterns which despite the name is a good general purpose guide to writing performant C++.
How can I better learn to “not pay for what you don't use”?
The key to "not paying for what you don't use" is abstractions. When you clearly understand the purpose of a class or a function, you add the data and arguments that are absolutely necessary for the class and the function to work correctly, with as little overhead as possible.
You have to be very vigilant about adding member variables, member functions (virtual as well as non-virtual) to a class. Every member variable adds to the memory requirements of the class. Every member function requires maintenance. Presence of virtual member functions add to the memory requirements of the class as well as a small penalty at run time.
You have to be very vigilant about the arguments to a function. You don't want the user to be burdened with supplying arguments that don't make sense. You also don't want to leave out any arguments by making hidden assumptions.
I realize there are quite a few posts on this subject, but I am having trouble finding the answer to this exact question.
For function calls, which is faster, a pure-virtual interface or a pimpl?
At first glance, it seems to me that the pure-virtual interface would be faster, because the using the pimpl would cost two function calls instead of one...or would some kind of clever compiler trick take over in this case?
edit:
I am trying to decide which of these I should use to abstract away the system-dependent portions of a few objects that may end up having to be spawned quite frequently, and in large numbers.
edit:
I suppose it's worth saying at this point, that the root of my problem was that I mistook the Abstract Factory design pattern for a method of making my code work on multiple platforms, when it's real purpose is for switching implementations for a given interface at runtime.
The two options are not equivalent, they should not be compared on performance as the focus is different. Even if they were equivalent, the performance difference would be minimal to unimportant in most situations. If you are in the rare case that you know that dispatch is being an issue, then you have the tools to measure the difference yourself.
Why do you ask? The question doesn't seem to make sense.
One generally uses virtual functions when one wants polymorphism: when you want them to be overridden in derived classes.
One generally uses pimpl when one wants to remove implementation details from header files.
The two really aren't interchangeable. Off the top of my head, I cannot think of any reasonable situations where you would use one and consider replacing it with the other.
Anyways, that said, for a typical implementation of virtual functions, a function call involves reading the object to find the virtual function table pointer, then reading the virtual function table to find the function pointer, and finally calling the function pointer.
For a class implemented via pimpl, one function call is forced, but it could be absolutely anything 'under the hood'. Despite what you suggest, no second function call is implied by the paradigm.
Finally, don't forget the usual guidelines for optimization apply: you have to actually implement and measure. Trying to "think" up the answer tends to lead to poor results, even from people experienced at this sort of thing.
And, of course, the most important rule of optimization: make sure something matters before you devote a lot of time trying to optimize it. Otherwise, you are going to wind up wasting a lot of time and energy.
I find myself always trying to fit everything into the OOP methodology, when I'm coding in C/C++. But I realize that I don't always have to force everything into this mold. What are some pros/cons for using the OOP methodology versus not? I'm more interested in the pros/cons of NOT using OOP (for example, are there optimization benefits to not using OOP?). Thanks, let me know.
Of course it's very easy to explain a million reasons why OOP is a good thing. These include: design patterns, abstraction, encapsulation, modularity, polymorphism, and inheritance.
When not to use OOP:
Putting square pegs in round holes: Don't wrap everything in classes when they don't need to be. Sometimes there is no need and the extra overhead just makes your code slower and more complex.
Object state can get very complex: There is a really good quote from Joe Armstrong who invented Erlang:
The problem with object-oriented
languages is they’ve got all this
implicit environment that they carry
around with them. You wanted a banana
but what you got was a gorilla holding
the banana and the entire jungle.
Your code is already not OOP: It's not worth porting your code if your old code is not OOP. There is a quote from Richard Stallman in 1995
Adding OOP to Emacs is not clearly an
improvement; I used OOP when working
on the Lisp Machine window systems,
and I disagree with the usual view
that it is a superior way to program.
Portability with C: You may need to export a set of functions to C. Although you can simulate OOP in C by making a struct and a set of functions who's first parameter takes a pointer to that struct, it isn't always natural.
You may find more reasons in this paper entitled Bad Engineering Properties
of Object-Oriented Languages.
Wikipedia's Object Oriented Programming page also discusses some pros and cons.
One school of thought with object-oriented programming is that you should have all of the functions that operate on a class as methods on the class.
Scott Meyers, one of the C++ gurus, actually argues against this in this article:
How Non-Member Functions Improve Encapsulation.
He basically says, unless there's a real compelling reason to, you should keep the function SEPARATE from the class. Otherwise the class can turn into this big bloated unmanageable mess.
Based on experiences in a previous large project, I totally agree with him.
A benefit of non-oop functionality is that it often makes exporting your functionality to different languages easier. For example a simple DLL containing only functions is much easier to use in C#, you can use the P/Invoke to simply call the C++ functions. So in this sense it can be useful for writing extremely time critical algorithms that fit nicely into single/few function calls.
OOP is used a lot in GUI code, computer games, and simulations. Windows should be polymorphic - you can click on them, resize them, and so on. Computer game objects should be polymorphic - they probably have a location, a path to follow, they might have health, and they might have some AI behavior. Simulation objects also have behavior that is similar, but breaks down into classes.
For most things though, OOP is a bit of a waste of time. State usually just causes trouble, unless you have put it safely in the database where it belongs.
I suggest you read Bjarne's Paper about Why C++ is not just an Object-Oriented Programming Language
If we consider, for a moment, not object-orienatation itself but one
of the keystones of object-orientation: encapsulation.
It can be shown that change-propagation probability cannot increase
with distance from the change: if A depends on B and B depends on C,
and we change C, then the probability that A will change
cannot be larger than the proabability that B will
change. If B is a direct dependency on C and A is an indirect
dependency on C, then, more generally, to minimise the potential cost
of any change in a system we must miminimise the potential number of
direct dependencies.
The ISO defines encapsulation as the property that the information
contained in an object is accessible only through interactions at the
interfaces supported by the object.
We use encapsulation to minimise the number of potential dependencies
with the highest change-propagation probability. Basically,
encapsulation mitigates the ripple effect.
Thus one reason not to use encapsulation is when the system is so
small or so unchanging that the cost of potential ripple effects is
negligible. This is also, therefore, a case when OO might not be used
without potentially costly consequences.
Well, there are several alternatives. Non-OOP code in C++ may instead be:
C-style procedural code, or
C++-style generic programming
The only advantages to the first are the simplicity and backwards-compatibility. If you're writing a small trivial app, then messing around with classes is just a waste of time. If you're trying to write a "Hello World", just call printf already. Don't bother wrapping it in a class. And if you're working with an existing C codebase, it's probably not object-oriented, and trying to force it into a different paradigm than it already uses is just a recipe for pain.
For the latter, the situation is different, in that this approach is often superior to "traditional OOP".
Generic programming gives you greater performance (among other things because you often avoid the overhead of vtables, and because with less indirection, the compiler is better able to inline), better type safety (because the exact type is known, rather than hiding it behind an interface), and often cleaner and more concise code as well (STL iterators and algorithms enable much of this, without using a single instance of runtime polymorphism or virtual functions.
OOP is little more than an aging buzzword. A methodology that everyone misunderstood (The version supported by C++ and Java has little to do with what OOP originally meant, as in SmallTalk), and then pretended was the holy grail. There are aspects to it that are useful, certainly, but it is often not the best approach for designing an application.
Rather, express the overall logic by other means, for example generic programming, and when you need a class to encapsulate some simple concept, by all means design it according to OOP principles.
OOP is just a tool among many. The goal is not to write OOP code, but to write good code. Sometimes, the way to do this is by using OOP principles, but often, you can get better code using generic programmming principles, or functional programming.
It is a very project dependent decision. My general feel of OOP is that its useful for organizing large projects that involve multiple components. One area I find that OOP is especially pointless is school assignments. Excepting those specifically designed to teach OOP concepts, or large software design concepts, many of my assignments, specifically those in more algorithmy type classes are best suited to non-OOP design.
So specifically, smaller projects, that are not likely to grow large, and projects that center around a single algorithm seem to be non-OOP candidates in my books. Also, if you can write the specification as a linear set of steps, e.g., with no interactive GUI or state to maintain, this would also be an opportunity.
Of course, if you're required to use an OOP design, or an OOP toolkit, or if you have well defined 'objects' in you're spec, or if you need the features of polymorphism, etc. etc. etc...there are plenty of reasons to use it, the above seem to be indicators of when it would be simple not to.
Just my $0.02.
Having an Ada background, I develop in C in terms of packages containing data and their associated functions. This gives a code very modular with pieces of code that can be taken apart and reused on other projects. I don't feel the need to use OOP.
When I develop in Objective-C, objects are the natural container for data and code. I still develop with more or less the package concept in mind with some new cool features.
I'm used to be an OOP fanboy... Then realized using functions, generics and callbacks can often make a more elegant and change-friendly solution in C++ than classes and virtual functions.
Other big names realized it too: http://harmful.cat-v.org/software/OO_programming/
IMHO, I have a feeling that the OOP concept is not really suits the needs of the Big Data, as OOP assume all the stuff to be kept in memory (concept of Objects and member variables). This always result in memory demanding and heavy applications when OOP is used for example for big images processing. Instead, the simplicity of C maybe used with intensive parallel I/O making apps more efficient and easy to implement. It is the year 2019 I am writing this message...Everything may change in a year! :)
In my mind it comes down to what kind of model suits the problem at hand. It seems to me that OOP is best suited to coding GUI programs, in that the data and functionality for a graphical object is easily bundled together. Other problems- (such as a webserver, as an example off the top of my head), might be more easily modeled with a data centric approach, where there's no strong advantage to having a method and its data near each-other.
tl;dr depends on the problem.
I'd say the greatest benefit of C++ OOP is inheritance and polymorphism (Virtual function etc...) .
This allows for code reuse and extendibility
C++, use OOP - - - C, no, with certain exceptions
In C++ you should use OOP. It's a nice abstraction and it's the tool you are given. You either use it or leave it in the box where it can't help. You don't use the power saw for everything but I would read the manual and have it ready for the right job.
In C, it's a more difficult call. While you can certainly write arbitrarily object-oriented code in C, it's enough of a pain that you immediately find yourself fighting the language in order to use it. You may be more productive dropping the doesn't-fit-so-well design pattern and programming as C was intended to be used.
Furthermore, every time you make an array of function pointers or something in an OOP-in-C design pattern, you sever almost completely all visible links in the inheritance chain, making the code hard to maintain. In real OOP languages, there is an obvious chain of derived classes, often analyzed and documented for you. (mmm, javadoc.) Not so in OOP-in-C, and the tools available won't be able to see it.
So, I would argue in general against OOP in C. For a really complex program, you may well need the abstraction, and then you will have to do it despite needing to fight the language in the process and despite making the program quite hard to follow by anyone other than the original author.
But if you knew the program was going to become that complicated, you shouldn't have written it in C in the first place...
In C, there are some times when I 'emulate' the object oriented approach, by defining some sort of constructor with granular control over things like callbacks, when running several instances of it.
For instance, lets say I have some spiffy event handler library and I know that down the road I'm going to need many allocated copies:
So I would have (in C)
MyEvent *ev1 = new_eventhandler();
set_event_callback_func(ev1, callback_one);
ev1->setfd(fd1);
MyEvent *ev2 = new_eventhandler();
set_event_callback_func(ev2, callback_two);
ev2->setfd(fd2);
destroy_eventhandler(ev1);
destroy_eventhandler(ev2);
Obviously, I would later do something useful with that like handle received events in the two respective callback functions. I'm not going to really elaborate on the method of typing function pointers and structures to hold them, nor what would go on in the 'constructor' because its pretty obvious.
I think, this approach works for more advanced interfaces where its desirable to allow the user to define their own callbacks (and change them on the fly), or when working on complex non-blocking I/O services.
Otherwise, I much prefer a more procedural / functional approach.
Probably an unpopular idea but I think you should stick with non-OOP unless it adds something useful. In most practical problems OOP is useful but if I'm just playing with an idea I start writing non-object code and put functions and data into classes if it becomes useful.
Of course I still use other objects in my code (std::vector et al) and I use namespaces to help organise my functions but why put code into objects until it is useful? Equally don't shy away from free functions in an OO solution.
The question is tricky because OOP encompasses several concepts: object encapsulation, polymorphism, inheritance, etc. It's easy to take those ideas too far. Here's a concrete example:
When C++ first caught on, zillions of string classes sprung into being. Everything you could possibly imagine doing to a string (upcasing, downcasing, trimming, tokenizing, parsing, etc.) was a member function of some string class.
Notice, though, that std::strings from the STL don't have all these methods. STL is object-oriented--the state and implementation details of a string object are well encapsulated, only a small, orthogonal interface is exposed to the world. All the crazy manipulations that people used to include as member functions are now delegated to non-member functions.
This is powerful, because these functions can now work on any string class that exposes the same interface. If you use STL strings for most things and a specialty version tuned to your program's idiosyncracies, you don't have to duplicate member functions. You just have to implement the basic string interface and then you can re-use all those crazy manipulations.
Some people call this hybrid approach generic programming. It's still object-oriented programming, but it moves away from the "everything is a member-function" mentality that a lot of people associate with OOP.