Do I need to learn C before C ++?

[jtbell]

If you talk to the professor and he specifically says that he will teach C++ beginning with the assumption that the students have zero C knowledge, then possibly I'll relent.

For several years (roughly 1995-2003 IIRC), I taught a two-semester introductory college computer science course that used C++. I did not assume that students knew any C beforehand, or indeed any other programming language. Many other colleges and universities had similar courses. During this period, C++ was also the preferred language for AP computer science courses in US high schools.

Around 2005, the College Board changed the AP computer science course from C++ to Java, and many colleges and universities (including us) followed suit with their intro CS courses.

Even though there is less demand for them now, there are still some introductory C++ textbooks for such courses, for example:

https://www.amazon.com/dp/0133378713/?tag=pfamazon01-20

https://www.amazon.com/dp/0763771562/?tag=pfamazon01-20

I used this book which is no longer in print:

A Computer Science Tapestry: Exploring Computer Science with C++ (Astrachan)

 

[phinds]

Sorry, it's Bjarne Stroustrup
https://en.wikipedia.org/wiki/C++
Who is Dennis Ritchie? I have his book.
Wait..
Ah there he is: https://en.wikipedia.org/wiki/Dennis_Ritchie
Ahh, died in 2011. At the same time as Steven Jobs as I recall.
RIP Dennis Ritchie...

Stroustrup created C++ on top of C. That's why I keep saying that C++ is just C plus OOP stuff (and as you seem conflicted about, but are clearly aware of, overloading only comes with the OOP stuff).

 

[Mark44]

Sorry, it's Bjarne Stroustrup
https://en.wikipedia.org/wiki/C++
Who is Dennis Ritchie? I have his book.

Dennis Ritchie is one-half of the acronym "K & R" which is very well-known to all long-time C programmers. Brian Kernighan (the 'K' of K & R) and Dennis Ritchie (the 'R' of K & R) wrote "The C Programming Language" back when they worked at Bell Labs.

Wait..
Ah there he is: https://en.wikipedia.org/wiki/Dennis_Ritchie
Ahh, died in 2011. At the same time as Steven Jobs as I recall.
RIP Dennis Ritchie...

 

[jtbell]

Stroustrup created C++ on top of C.

True. In fact, IIRC Stroustrup's working name for the language while he was developing it was "C with Classes."

That's why I keep saying that C++ is just C plus OOP stuff

At one time this was sort of true, to some approximation. I remember reading, while I was learning C++ around 1995, that there was some C code that most C++ compilers did not accept. I think there may have been an FAQ about in the Usenet newsgroup comp.lang.c++. My understanding is that after the first C++ standard came out in 1998(?) C and C++ started to diverge further from each other. Later versions of C added features which were not added to C++, and vice versa.

 

[Stephanus]

Isn't it interesting. I bet all the mentors/staffs here are very capable and fluently coding in C or C++. But it seems that we argue what is C relative to C++.
But I am not astonished, because only in C do the programming language has its fanatics.
I have (had?, don't know where the book now) a book. It's "Born to Code in C". The author is Herbert Schildt.
There is a semaphore programming in his books. And if I'm not mistaken, there is a simple compiler, or there isn't.
But C is an interesting P/L. Never do I find it in Pascal, assembly.
Much less Clipper

 

[Stephanus]

And for the OP question.
"Do I need to learn C before C++"
I think my answer is this.
If you don't know about OOP, then even if you compile your code by C++ compiler, then it's C, As @phinds says.
And if you don't use function overloading, and class definition, then it's C.
If I recall correctly, in C++ we can even put methods in union and struct, not just class. I don't know about enum. Can we do that in Enum?

 

[Stephanus]

Stroustrup created C++ on top of C. That's why I keep saying that C++ is just C plus OOP stuff (and as you seem conflicted about, but are clearly aware of, overloading only comes with the OOP stuff).

To use operator overloading, you need OOP.
Function overloading does not need OOP.
And OOP does not depend whether the language can do function overloading or not.
C++ can have multiple inheritances, but C++ can't have two constructors without function overloading. Pascal/Delphi can have more than one constructors per class. But in my 20 years of Delphi programming, I never use two constructors.

 

[Mark44]

But I am not astonished, because only in C do the programming language has its fanatics.
I have (had?, don't know where the book now) a book. It's "Born to Code in C". The author is Herbert Schildt.
There is a semaphore programming in his books. And if I'm not mistaken, there is a simple compiler, or there isn't.

I'm pretty sure Herb Schildt didn't include a compiler with any of his books.

And if you don't use function overloading, and class definition, then it's C.
If I recall correctly, in C++ we can even put methods in union and struct, not just class. I don't know about enum. Can we do that in Enum?

I'm going to say that you definitely can't put functions/methods in an enumeration (why would you want to?), and I don't think you can put them in a union. If you can, I've never seen an example of it.

You can have C structs with member functions -- by using function pointers.

 

[Stephanus]

I'm pretty sure Herb Schildt didn't include a compiler with any of his books.I'm going to say that you definitely can't put functions/methods in an enumeration (why would you want to?), and I don't think you can put them in a union. If you can, I've never seen an example of it.

You can have C structs with member functions -- by using function pointers.

It's been 20 years since I learn C++. I'm using C now. But I remembered something like "member function in union"
http://stackoverflow.com/questions/5275100/union-in-c-are-they-feasible

template<typename T>
unionFoo{
public:
  Foo(){}
  Foo(const T& value): _val(value){}

  constchar* data()const{
    return _tab;
  }

  std::size_t size()const{
    returnsizeof(T);
  }

  charoperator[](unsignedint index)const{
      return _tab[index];
  }

private:
  T _val;
  char _tab[sizeof(T)];
}

Yes, member function in enum really doesn't make any sense.

 

[Stephanus]

...
You can have C structs with member functions -- by using function pointers.

No, it's not that we have a function pointer element in a struct. But it's a real method.
Remember, that class without methods works like struct. Just member variable.
But in struct, you can't derive it, you can't inherit it. And I'm pretty sure there are no virtual/dynamic method in struct. Frankly, I never used method in struct before. I just read in the C++ book, that it can do that. I only use class and played around with operator overloading.

 

[meBigGuy]

You are making exactly the point I have been propounding all along. It seems to me that you are agreeing with me while disagreeing with me.

No, I'm not reacting to you at all. The fact that C++ is built on top of C is not at issue (at least not for me). It's whether you learn the C basics before learning the higher level constructs. I'm a strong proponent for learning the basics first, but as has been mentioned there are courses that explicitly state that no C knowledge is required.

In any C++ course a person who already knows C will always have an easier time of it than a person who does not.

 

[Mark44]

No, it's not that we have a function pointer element in a struct. But it's a real method.

The member functions in a C++ class are typically implemented as pointers, same as what I was talking about.

Remember, that class without methods works like struct. Just member variable.
But in struct, you can't derive it, you can't inherit it.

You can simulate derivation by having a struct member that is itself a struct.

And I'm pretty sure there are no virtual/dynamic method in struct.

Of course, you can't get all of the OOP features of C++ in C, because C is roughly a subset of C++.

Frankly, I never used method in struct before. I just read in the C++ book, that it can do that. I only use class and played around with operator overloading.

 

[Stephanus]

The member functions in a C++ class are typically implemented as pointers, same as what I was talking about.

You mean in Test1 or Test2?

class Test1 {
   public: 
     int MyVar;
     int MyFunction(int b) { return(b*10); }
};

class Test2 {
   public:
     int MyVar;
     int (*MyFunction) (int);
};

int main()
{
  Test1 Test1Var[1000]; 
  Test2 Test2Var[1000];
}

Now, I'd say that Test2Var takes up more memory then Test1Var, IF we implement the member function as pointer.

 

[Mark44]

You mean in Test1 or Test2?

I mean in both.

class Test1 {
   public:
     int MyVar;
     int MyFunction(int b) { return(b*10); }
};

class Test2 {
   public:
     int MyVar;
     int (*MyFunction) (int);
};

int main()
{
  Test1 Test1Var[1000];
  Test2 Test2Var[1000];
}

Now, I'd say that Test2Var takes up more memory then Test1Var, IF we implement the member function as pointer.

I wouldn't say that. I would bet that both classes are exactly the same size.

 

[Stephanus]

I mean in both.

I wouldn't say that. I would bet that both classes are exactly the same size.

I think Test2Var takes up more space.
We just can't sizeof(a class). If we sizeof it, it would return 4, the size of the pointer. But...
Test1::MyFunction is a method
Test2->MyFunction is a variable, which is a pointer of a method.
In a class, you can new 1000 objects of that class. Only the variables are created. But there is only 1 method.
In Test2, since MyFunction is a variable it's included in every objects that are created.
We define a class, we create an object.

 

[Mark44]

I think Test2Var takes up more space.

I compiled your code (VS 2013), and a Test2 instance does take up more space, at least the MSFT does it, which is probably an implementation detail. The last time I checked, which was about 20 years ago, I believe that they implemented class methods as pointers to code, but they don't seem to do that any more.

We just can't sizeof(a class). If we sizeof it, it would return 4, the size of the pointer.

But you can take sizeof an instance of a class.

 

[Stephanus]

I compiled your code (VS 2013), and a Test2 instance does take up more space, at least the MSFT does it, which is probably an implementation detail.

How do you know? You increase the size of the array say.. 1000000 elements, and see the remaining memory in your computer?
And Mark, I forgot something. IF you define Test1::MyFunction as virtual not as static, I think it Test1 Test1Var[10000], will take some memory, because the compiler needs to store its address. But that's how it works in Pascal. I think the same goes to C++, too.

I The last time I checked, which was about 20 years ago, I believe that they implemented class methods as pointers to code, but they don't seem to do that any more.

Really?

But you can take sizeof an instance of a class.

In Pascal, we can't sizeof the instance of the class. How you do that in C?
MyClass *MyClassVar;
x = SizeOf(MyClassVar) will give you 4.
x = SizeOf(MyClassVar*), like this?

 

[Stephanus]

Correction.

... IF you define Test1::MyFunction as virtual not as static, I think it Test1 Test1Var[10000], will take some memory, because the compiler needs to store its address. But that's how it works in Pascal. I think the same goes to C++, too.

No. It doesn't work that way.
The compiler doesn't have to store virtual method pointer in the class. The compiler stores it in the class definition. For each class (not object) that we define, the compiler reserved some memory for their VMT, virtual method table. And some static function pointer.
Test1Var[100000] still takes a little memory. The compiler does not store the address of MyFunc, even if it is a virtual method.
I guess so.

 

[Mark44]

I compiled your code (VS 2013), and a Test2 instance does take up more space, at least the MSFT does it, which is probably an implementation detail.

How do you know? You increase the size of the array say.. 1000000 elements, and see the remaining memory in your computer?
And Mark, I forgot something. IF you define Test1::MyFunction as virtual not as static, I think it Test1 Test1Var[10000], will take some memory, because the compiler needs to store its address. But that's how it works in Pascal. I think the same goes to C++, too.

I didn't bother creating an array of objects of each type, just one of each. My program used the sizeof operator to determine how many bytes were used by each of the two objects

I The last time I checked, which was about 20 years ago, I believe that they implemented class methods as pointers to code, but they don't seem to do that any more.

Really?

That's the way I remember it.

But you can take sizeof an instance of a class.

In Pascal, we can't sizeof the instance of the class. How you do that in C?
MyClass *MyClassVar;
x = SizeOf(MyClassVar) will give you 4.
x = SizeOf(MyClassVar*), like this?

Yes, except that the operator is sizeof in C/C++

 

[Stephanus]

I didn't bother creating an array of objects of each type, just one of each. My program used the sizeof operator to determine how many bytes were used by each of the two objects.

I think the result of A=sizeof(ClassType*) could be misleading. There are additional variables involved.

But that is in the declaration for each class definition not each class instance that the compiler generate.
And it's in Pascal. Perhaps C compiler generates some table, too for each of their class definition declared.

Yes, except that the operator is sizeof in C/C++

Come on, I've been doing Pascal 90% than C. And in your later years as a progammer, you wouldn't use sizeof anymore. You might have known the size of the variables you once define them.
I think learning C in C++ compiler is Ok. As long as the beginner doesn't go straight to OOP.
I'm impressed with @jtbell 's approach.

I didn't do pointers (or OOP [added by me]) until near the end of the course...

... I remember having trouble ... to motivate introducing pointers. "OK, here's what I'd like to do, but none of the tools we've learned so far can accomplish it. So here's something new: pointers. (or OOP [added by me])"

 

[Stephanus]

And in your later years as a progammer, you wouldn't use sizeof anymore. You might have known the size of the variables you once define them.

Oops, sorry.

#define ArrSize 100
int a[ArrSize];
int *p;
p=malloc(ArrSize*sizeof(int)); // much better than
p=malloc(ArrSize*4);

It's a good discussion we have. Add more insights for me.

 

[jtbell]

The original question posed by this thread has been answered. The thread is now veering around in different directions related to C versus C++ with some Pascal thrown in. (I admit to being guilty of contributing to this.) This thread is now closed. If someone wants to continue one of these veered-off discussions in a new thread with a focused title, please feel free to do so.