The use of pointers...

I don't see where c++ wants to protect the programmer from pointers. If it would do so there would be

You are looking at part of C++'s C legacy. That said, I do not think that C++'s goal here is to protect the programmer from pointers. Rather, it is to provide abstractions to ease tasks that would otherwise need direct use of pointers, and yet allow the programmer to use such a low level construct as necessary.

So you need to know about pointers for the simplest possible c++ program.

Command line arguments can be introduced later. The version of main() that has no parameters does exist, after all.

One doesn't need to know anything about pointers to understand iterators.

And the point isn't to completely get rid of pointers from the teaching. Rather, its to limit there use in cases where better alternatives exist. One doesn't necessarily need to understand a function in order to start learning with int main(). Besides, the signature of main that you mention in your example would not be used by most beginners.

I would liken the supposition that pointers aren't useful, to suggesting that the division operator isn't useful.

Sure if every program you can currently think of can be solved easily by only adding addition and multiplication, then division would seem pretty useless.

All it means is that you haven't ventured very far into computer programming yet. Don't worry about it, if you continue programming the question will answer itself soon enough.

Originally Posted by pheres

I don't see where c++ wants to protect the programmer from pointers. If it would do so there would be

Code:

int main(vector<string> argv);

// instead of

int main(int argc, char** argv);

So you need to know about pointers for the simplest possible c++ program.
And even if the first version of main would exists one wouldn't be able to list all command line arguments without dereferencing the iterator.
Better get pointers right.

http://www.joelonsoftware.com/articl...vaSchools.html

I don't believe I made such a claim that C++ protects programmers from pointers - indeed it doesn't, as you point out (Although I would add that a perfectly valid standards conforming declaration of main permits those parameters to be omitted).

The issue I was addressing was that of complete newcomers to programming, rather than those who have already passed the basics - That it is possible in C++ to learn a large, useful, portion of the language without having to deal directly with pointers - This is largely thanks to the availability of strings, references, STL containers and iterators.

Of course, once a beginner gains confidence with the easier parts of the language, then the floodgates are open for all sorts of low-level techniques which helps them better understand the why's and how's of things working the way they do. (Taking the "top down" approach in C++ doesn't mean that primitive topics are in any way bypassed, as I believe joelonsoftware is suggesting in his JavaSchools article)

ok you are right, int main() can be used at the beginning.

but what is about reading code?
C++ has A LOT of stuff build in it. And even a lot of that stuff isn't required to write a lot of programs (maybe pointers belongs to that stuff).
Reading code makes good programmers, gets new ideas, let's one see how problems can solved better, cleaner, faster. To have fun with nearly all c++ books or tuts one has to read code.
But people who write interesting code to read about do not bother using int main(), or to resign from using all other deep c++ stuff.

C++ has A LOT of stuff build in it. And even a lot of that stuff isn't required to write a lot of programs (maybe pointers belongs to that stuff).

I think you are not reading what we are saying: You have to learn how to use pointers, so you would be more insulated against the Law of Leaky Abstractions (since you want to quote Joel Spolsky ). But you should not start by learning how to use pointers. Read Koenig and Moo's book Accelerated C++ for an example of how to introduce C++ using the standard library, leaving pointers for as late as chapter 10 (and that includes a chapter 0).

Thanks a bunch. I still don't 100% understand but I read in one of those links that they are only really important for more "dynamic programs". So long as I know how to use them (not when), that should be sufficient for now.

IMO "WHEN" to use pointers depends on how you are designing your program. You usually run into pointers when you need interchangability and flexibility. For example, you have a list of functions that all have the same prototype and you want to be able to assign any of those functions to be executed when you press the Z Key on your keyboard. You would use a function pointer. Now what happens when you press the Z Key depends entirely on which of the functions you have assigned to the pointer. Another example would be when you do not want to create a duplicate copy of the data. Lets say you have several integer arrays that have 10 values each and you need to modify them all by adding 1 to each of their values. You can store a list of all the integer arrays by creating an array of 10 pointers to them. Now you can loop through them to process all 10 arrays.

Originally Posted by laserlight

But you should not start by learning how to use pointers. Read Koenig and Moo's book Accelerated C++ for an example of how to introduce C++ using the standard library, leaving pointers for as late as chapter 10 (and that includes a chapter 0).

Thanks for the tip, I will get that book as soon I get money

I guess we are just talking from different situations. You are right, it would be nice if everyone could learn everything step by step from ground to top. But not everyone learns c++ for pleasure. Sometimes you just have to get the thing done asap. for example loading code from a dll you get at run time. this is no big deal so far, but I bet chapter 0 to 9 doesn't explain it. They can't, because you need a silly pointer. Now.

I remember hearing from a friend about a math major where he worked and that when the math major discovered pointers he wrote a program with 13+ levels of indirection.

He kept complaining to his workmates that the compiler kept crashing.

The response was "just because you can think it, doesn't mean the compiler can do it"

Dave

I once wrote a program that would automatically generate a pointer with very large numbers of indirection. I can't remember the exact results, but I think that at least 3000 levels of indirection were supported (this was GCC 2.95). It depended on the surrounding code, so I think that when the compiler eventually crashed it was because it ran out of stack space.

Anyway, I think the C89 standard specifies that a compiler must support at least 8 levels of pointer indirection. I'm pretty sure that C99 says 64, but I'm not sure about C++.

Originally Posted by Dave++

I remember hearing from a friend about a math major where he worked and that when the math major discovered pointers he wrote a program with 13+ levels of indirection.

He kept complaining to his workmates that the compiler kept crashing.

The response was "just because you can think it, doesn't mean the compiler can do it"

Dave

With that many levels of indirection, it's VERY difficult to manage memory. One has to make sure that enough memory is allocated at each level so each access at each level is in bounds. His program probably was just buggy. And if the compiler can't handle more than a certain number of levels, it should give an error when the limit is exceeded.

Originally Posted by dwks

Anyway, I think the C89 standard specifies that a compiler must support at least 8 levels of pointer indirection. I'm pretty sure that C99 says 64, but I'm not sure about C++.

I shudder to think of the use cases that prompted the standard writers to change this requirement. I mean ...

Code:

char ****************************************************************wtf;

You'd be surprised what generated code can look like. I'm sure that sort of thing might be generated by bison or something.

[edit] Actually, this page says (with much elaboration) that C89 supports 12 levels of pointer indirection. At least as far as I read. http://archives.devshed.com/forums/c...e-2322818.html

Strange. I thought I read somewhere that C89 supported 8 levels. I guess not. [/edit]

The final draft of C99 also speaks of 12 levels of indirection.