Hi all.
I come from a the Objective-C and Cocoa world where there are lots of conventions and many people will say it makes your code beautiful! Now programming in C++ I cannot find a good document like this one for C++.
Standard C++ probably does not have something like above but I hope I can stick to some other SDK or APIs (like Microsoft's(?),etc) conventions.
I hope you can provide me with some links.
Thanks.
Ignacio.
While many people will suggest more or less strict Hungarian notation variants (scary!), for naming suggestions I'd suggest you take a look at Google C++ Coding Guidelines. This may well be not the most popular naming conventions, but at least it's fairly complete. Apart from sound naming conventions, there's some useful guidelines there, however much of it should be taken with a grain of salt (exception ban for example, and the tendency to keep away from modern C++ coding style).
Although personally I like the extreme low-tech convention style of STL and Boost ;).
not nearly as concise as the link you provided: but the following chapter 14 - 24 may help :) hehe
To address the broader issue of style guides I suggest picking on and stickign to it. the Google style guide is very detail/comprhensive, so it a good choice.
A common convention just for naming is:
camelCase for all names, (except constants)
Begining with a capital if it's a class.
lowercase if not.
Constants in ALL_CAPS with underscore seperating works.
functions can be written the same as veriables, camelCase (you can tell it's a function cos it requiere the () after it)
and member veribles prefixed with an m and thuse would be written mCamelCase
It's really up to you or alternitivly the people you work for/with or the university you study at.
For a detail look at a variety of styguides / rationals behind htem, check out the "Pragmatic Programmer" by Andrew Hunt and David Thomas
Do whatever you want as long as its minimal, consistent, and doesn't break any rules.
Personally, I find the Boost style easiest; it matches the standard library (giving a uniform look to code) and is simple. I personally tack on m and p prefixes to members and parameters, respectively, giving:
#ifndef NAMESPACE_NAMES_THEN_PRIMARY_CLASS_OR_FUNCTION_THEN_HPP
#define NAMESPACE_NAMES_THEN_PRIMARY_CLASS_OR_FUNCTION_THEN_HPP
#include <boost/headers/go/first>
#include <boost/in_alphabetical/order>
#include <then_standard_headers>
#include <in_alphabetical_order>
#include "then/any/detail/headers"
#include "in/alphabetical/order"
#include "then/any/remaining/headers/in"
// (you'll never guess)
#include "alphabetical/order/duh"
#define NAMESPACE_NAMES_THEN_MACRO_NAME(pMacroNames) ARE_ALL_CAPS
namespace lowercase_identifers
{
class separated_by_underscores
{
public:
void because_underscores_are() const
{
volatile int mostLikeSpaces = 0; // but local names are condensed
while (!mostLikeSpaces)
single_statements(); // don't need braces
for (size_t i = 0; i < 100; ++i)
{
but_multiple(i);
statements_do();
}
}
const complex_type& value() const
{
return mValue; // no conflict with value here
}
void value(const complex_type& pValue)
{
mValue = pValue ; // or here
}
protected:
// the more public it is, the more important it is,
// so order: public on top, then protected then private
template <typename Template, typename Parameters>
void are_upper_camel_case()
{
// gman was here
}
private:
complex_type mValue;
};
}
#endif
That. (And like I've said in comments, do not adopt the Google Style Guide for your code, unless it's for something as inconsequential as naming convention.)
It really doesn't matter. Just make sure you name your variables and functions descriptively. Also be consistent.
Nowt worse than seeing code like this:
int anInt; // Great name for a variable there ...
int myVar = Func( anInt ); // And on this line a great name for a function and myVar
// lacks the consistency already, poorly, laid out!
Edit: As pointed out by my commenter that consistency needs to be maintained across an entire team. As such it doesn't matter WHAT method you chose, as long as that consistency is maintained. There is no right or wrong method, however. Every team I've worked in has had different ideas and I've adapted to those.
There are probably as many naming conventions as there are individuals, the debate being as endless (and sterile) as to which brace style to use and so forth.
So I'll have 2 advices:
The rest is up to you.
There are many different sytles/conventions that people use when coding C++. For example, some people prefer separating words using capitals (myVar or MyVar), or using underscores (my_var). Typically, variables that use underscores are in all lowercase (from my experience).
There is also a coding style called hungarian, which I believe is used by microsoft. I personally believe that it is a waste of time, but it may prove useful. This is were variable names are given short prefixes such as i, or f to hint the variables type. For example: int iVarname, char* strVarname.
It is accepted that you end a struct/class name with _t, to differentiate it from a variable name. E.g.:
class cat_t {
...
};
cat_t myCat;
It is also generally accepted to add a affix to indicate pointers, such as pVariable or variable_p.
In all, there really isn't any single standard, but many. The choices you make about naming your variables doesn't matter, so long as it is understandable, and above all, consistent. Consistency, consistency, CONSISTENCY! (try typing that thrice!)
And if all else fails, google it.
We follow the guidelines listed on this page: C++ Programming Style Guidelines
I'd also recommend you read The Elements of C++ Style by Misfeldt et al, which is quite an excellent book on this topic.
I actually often use Java style: PascalCase for type names, camelCase for functions and variables, CAPITAL_WORDS for preprocessor macros. I prefer that over the Boost/STL conventions because you don't have to suffix types with _type. E.g.
Size size();
instead of
size_type size(); // I don't like suffixes
This has the additional benefit that the StackOverflow code formatter recognizes Size as a type name ;-)
consistency and readability (self-documenting code) are important. some clues (such as case) can and should be used to avoid collisions, and to indicate whether an instance is required.
one of the best practices i got into was the use of code formatters (astyle and uncrustify are 2 examples). code formatters can destroy your code formatting - configure the formatter, and let it do its job. seriously, forget about manual formatting and get into the practice of using them. they will save a ton of time.
as mentioned, be very descriptive with naming. also, be very specific with scoping (class types/data/namespaces/anonymous namespaces). in general, i really like much of java's common written form - that is a good reference and similar to c++.
as for specific appearance/naming, this is a small sample similar to what i use (variables/arguments are lowerCamel and this only demonstrates a portion of the language's features):
/** MYC_BEGIN_FILE_ID::FD_Directory_nanotimer_FN_nanotimer_hpp_::MYC_BEGIN_FILE_DIR::Directory/nanotimer::MYC_BEGIN_FILE_FILE::nanotimer.hpp::Copyright... */
#ifndef FD_Directory_nanotimer_FN_nanotimer_hpp_
#define FD_Directory_nanotimer_FN_nanotimer_hpp_
/* typical commentary omitted -- comments detail notations/conventions. also, no defines/macros other than header guards */
namespace NamespaceName {
/* types prefixed with 't_' */
class t_nanotimer : public t_base_timer {
/* private types */
class t_thing {
/*...*/
};
public:
/* public types */
typedef uint64_t t_nanosecond;
/* factory initializers -- UpperCamel */
t_nanotimer* WithFloat(const float& arg);
/* public/protected class interface -- UpperCamel */
static float Uptime();
protected:
/* static class data -- UpperCamel -- accessors, if needed, use Get/Set prefix */
static const t_spoke Spoke;
public:
/* enums in interface are labeled as static class data */
enum { Granularity = 4 };
public:
/* construction/destruction -- always use proper initialization list */
explicit t_nanotimer(t_init);
explicit t_nanotimer(const float& arg);
virtual ~t_nanotimer();
/*
public and protected instance methods -- lowercaseCamel()
- booleans prefer is/has
- accessors use the form: getVariable() setVariable().
const-correctness is important
*/
const void* address() const;
virtual uint64_t hashCode() const;
protected:
/* interfaces/implementation of base pure virtuals (assume this was pure virtual in t_base_timer) */
virtual bool hasExpired() const;
private:
/* private methods and private static data */
void invalidate();
private:
/*
instance variables
- i tend to use underscore suffix, but d_ (for example) is another good alternative
- note redundancy in visibility
*/
t_thing ivar_;
private:
/* prohibited stuff */
explicit t_nanotimer();
explicit t_nanotimer(const int&);
};
} /* << NamespaceName */
/* i often add a multiple include else block here, preferring package-style inclusions */
#endif /* MYC_END_FILE::FD_Directory_nanotimer_FN_nanotimer_hpp_ */
For what it is worth, Bjarne Stroustrup, the original author of C++ has his own favorite style, described here: http://www2.research.att.com/~bs/bs_faq2.html