C++ for C programmers, part 1 of 2

May 2010

Original article on blog.brush.co.nz

As Stroustrup said, there’s no such programming language as C/C++. C and C++ are two different beasts.

I’m not going to argue here about which is better. For small or embedded projects, the complexities of C++ can get in the way. For larger projects, some of the features of C++ are very useful.

I’m simply going to list the features C++ adds to C, from an ex-C programmer’s point of view (yes, I was one of those people who believed that C++ was just “C with classes”). The idea is to follow the links to further information if you want to know more about any of them.

This is part 1 of 2, the non-OO features:

Default arguments to functions: defining int read(int bufsize =512) means that if you call read() without arguments it’s like you called read(512).
Somewhat related is function overloading. You can define a function twice, with two different argument lists. The compiler will use the “closest match” when you call it. For example, defining int print(int n) {...} and int print(const char* s) {...} means you can print() both ints and char strings.
The compiler uses ambiguity resolution to resolve all these overloaded functions.
You can declare local variables wherever in the code you need them, not just at the start of a function (also a C99 feature). This means variables can be declared closer to where you use them. The variable is scoped to the {...} block you put it in.
References, as in int& x. This isn’t some weird Microsoft-only syntax for smart pointers. I started to think of these as pointers without all the *‘s, but they’re quite different. An alias is another name for the referred-to object, usually used for pass-by-reference — and unlike pointers, you can call say swap(x, y) instead of swap(&x, &y). A const reference means you’re not allowed to change the object it refers to.
Const-correctness in general. This is not just a C++ thing, but C++ folks tend to be stricter about it (usually for good reason).
A namespace just groups a bunch of names together (functions or variables). C should have had them all along, to avoid everyone having to use their own brand of mylib_myfunc() pseudo-namespaces. See also the using keyword in its various forms.
The built-in boolean type, bool, with true and false keywords (also in C99 via stdbool.h).
The inline keyword and associated quirks, including implicit inlining and auto-inlining.
Most people don’t realise it, but (not f or n == 5) is both valid Python and valid C++, thanks to ISO646-style operator keywords.
A much larger standard library than C. The Standard Template Library (STL) is definitely worth knowing, especially the container datatypes like vector (growable array) and map (associative array), and the STL algorithms such as find() and sort(). C++’s iostreams are just plain weird (I mean, using the bit-shift operators to print stuff out?), but they’re here to stay. Of course, you can still use all of C’s standard library, and if you #include <cstdio> instead of <stdio.h>, you’ll get the names in the std namespace, as in std::printf().
Linking C++ code to C code has several gotchas.

Please send your feedback, and let me know if I’ve missed any non-OO features. The following week’s entry contains the second part describing the object-oriented features C++ has added.

Comments

Fabien 3 May 2010, 23:50

You forgot the most important:

Don’t use malloc, free, new[], delete[], or even delete.

Do use std::string, std::vector<>, and smart pointers.

Loup Vaillant 4 May 2010, 00:37

Talking about gotchas, I prefer the FQA over the FAQ. Quite depressing, but more entertaining.

Jason Baker 4 May 2010, 00:47

You might want to consider calling this part “1 of 2”. I was trying to figure out exactly what part one half meant. :-)

José Carlos Penfield da … 4 May 2010, 02:18

Hi !

The tutorial is excellent and is now part of my bookmarks, under the heading of useful tips.

However, i can’t let a comment pass uncommented. :)

The advice of not using new, new[] and their associates delete and delete[] is rather questionable.

C and C++ share the same philosophy of supposing a programmer aware of the risks. The languages usually trade extreme security for extreme performance.

And while malloc()/calloc() and free() should be generally avoided in terms of orthodox C++ programming, sometimes their use is recommended. For instance, when you got to pass a pointer to a C library.

Kevin Bowling 4 May 2010, 07:08

Looking forward to part 2.

caf 10 Aug 2010, 17:44

You might want to mention that `inline’ is also a co-feature with C99.

coder 26 Aug 2010, 05:46

Thanks for posting this, I have this book, but a lot of things have changed since this book has been published. Thanks again for posting this! :-)

Coder 23 Nov 2010, 06:35

Thanks for posting this information, very helpful when comparing C and C++, thanks again! :-)

Geoffrey Hunter 23 Oct 2012, 11:20

I would be interested to know why malloc is ‘acceptable’ in C yet frowned upon in C++.

Ben 23 Oct 2012, 11:45

@Geoffrey: in embedded systems, you often don’t use either malloc or free (heap allocation can be dangerous). But to answer your question more generally, here’s a StackOverflow answer on that — basically new is usually better because it’s type safe and calls your constructors.

Sayeed Sezan 28 Nov 2012, 08:36

Your post so is informative. I like C and C++ programming. thanks