This document defines and formalizes the presentation and redaction of all source files. A strict adherence to these conventions is required to ensure a readable and understandable code.
There are very few restrictions on file names and locations. The following rules are expected:
- Files must end with the appropriate suffix:
- cpp for ANSI C++ code
- c for ANSI C code
- h for included declarations
- Significant, English, file names. Length is not capped but should be kept within reasonable bounds.
- No spaces or uppercase letters. If the name is composed of separate, discrete, words the underline character is used instead of a space as the separator.
A quick check of the source directory will show how files are organized within the source tree. The sources are organized in two levels.
The first level specifies which part of the platform the source belongs to:
The second level specifies the types of data:
NOTE: The directory include only contains public headers, if you have private headers put them in src.
The third level specifies which logical part the file belongs to:
This applies to basic source as well as include files, who are in a separate include tree and never along their source code files. Thus, an include declaration would be:
All code files must begin with a specific header. The model of that header is defined in headers.txt in the documentation directory. It is recommended that people include that file immediately when starting a new file.
- The first letter specifies s for signed integers or u for unsigned integers. It is not possible to specify a dont care sign integer.
- The next three letters are int.
- The type ends with the number of bits used to represent the integral type, or nothing to specify an integer that is at least 32 bits wide.
Examples: uint8 (unsigned 8 bit integer), sint64 (signed 64 bit integer), sint (signed integer of at least 32 bits)
- float (32 bits floating point number, use is discouraged)
- double (64 bits floating point number)
- bool (an equivalent to the hypothetical uint1 type)
- void (empty type)
- char (ASCII character, not to be confused with uint8/sint8)
- ucchar (Unicode character, on 16 bits)
- string (ASCII string)
- ucstring (Unicode character string, inherited from std::string)
As a general rule, the classic and sloppy use of short / int / long is proscribed, as is using a char to represent a 8-bit number, unless this is required to call an external API. a char should use only when you want to represent a ASCII character.
All elements will be in their own namespace, which is short, in all uppercase, and consist of a two letter code (NL for NeL), followed by the directory name. Thus an element of the NeL 3D library will be declared in the NL3D namespace.
The following conventions are applied to names within these namespaces:
- All names must include one or more english words which have relevance to the entity being named. If multiple words are used for a more descriptive naming, all words begin with an uppercase letter, the underscore character is not used to separate words (e.g. Word, FullWord, ComplexQualifiedWord)
- Class methods start with a lowercase character, e.g. load(), store(), giveMoney()
- Public attributes of a class start with an uppercase, e.g. Layer, Set, ResourceType
- Local variables and parameters start with a lowercase letter, e.g. size, x, i
- Structures and Classes start with the uppercase C letter, then the name starting with an uppercase letter as well, e.g. CTexture, CMyClass
- Interfaces start with the uppercase I letter, then the name starting with an uppercase letter as well
- Iterators start with the prefix 'It' (uppercase I, lowercase t), then the name starting with an uppercase letter, e.g. ItTexture
- Other types, enums and the like are prefixed with an uppercase T, then the name.
- Globals function and values are prefixed by their namespace values, then follow the standard conventions, e.g. NLMISC::createAttribut()
- Preprocessor constants are in full uppercase, and consist of the namespace, an underscore character, then a name, with underscores as separators, e.g. NLNET_MACRO, NL_OS_UNIX
Note that there is no difference between variables and constants.
A strong suggestion is to set your tabulation marks to 4 characters, rather than the usual 8. Indentation is always done using tabulations, never with spaces. Make sure your text editor does not replace tabulations with spaces.
If possible, function, members and variable lists should align the first character of all their names.
Source code is written using a flat style. In this style, the opening and closing curly braces are placed on separate lines (never with code), and at the same indentation as the preceding defining element, while the content of the block enclosed between the braces is indented once from the source. Your code source should look like this:
Comments may use either the C++ style comments //, or the more classic C commenting method. If possible the source must be heavily commented and as many comments as possible must conform to the Doxygen commenting methods so that they can be extracted and reused as software documentation.
Comment "separators" will be made with a line of repeated star '*' characters. You should not use more than 50 stars to separate.
The practice of comment blocks, comments where each line begins and end with a star character, to create a "text block" is proscribed.
Final Coding Recommendations
Even if most of these are obvious recommendations, it is useful to restate these basic principles:
- The use of #define must be restricted. Constants values and macro functions can be expressed directly in C++ using constants and inline functions, and do not require pre-processor definitions.
- The 'using namespace' declaration is proscribed in include files.
- Include files paths, and all other file names should use the '/' character as the directory separator, for portability purposes.
- Use C++ classes, like string instead of old school 'char *' C.
- Use the const keyword as much as possible.