Overview of File Translation

C++ programs, like C programs, consist of one or more files. Each of these files is translated in the following conceptual order (the actual order follows the “as if” rule: translation must occur as if these steps had been followed):

  1. Lexical tokenizing. Character mapping and trigraph processing, line splicing, and tokenizing are performed in this translation phase.

  2. Preprocessing. This translation phase brings in ancillary source files referenced by #include directives, handles “stringizing” and “charizing” directives, and performs token pasting and macro expansion (see Preprocessor Directives in the Preprocessor Reference for more information). The result of the preprocessing phase is a sequence of tokens that, taken together, define a “translation unit.”

    Preprocessor directives always begin with the number-sign (#) character (that is, the first nonwhite-space character on the line must be a number sign). Only one preprocessor directive can appear on a given line. For example:

    #include <iostream.h>  // Include text of iostream.h in 
                       //  translation unit.
#define NDEBUG         // Define NDEBUG (NDEBUG contains empty 
                       //  text string).
  3. Code generation. This translation phase uses the tokens generated in the preprocessing phase to generate object code.

    During this phase, syntactic and semantic checking of the source code is performed.

See Phases of Translation in the Preprocessor Reference for more information.

The C++ preprocessor is a strict superset of the ANSI C preprocessor, but the C++ preprocessor differs in a few instances. The following list describes several differences between the ANSI C and the C++ preprocessors: