The Context Object

Introduction
Header 'wave/context.hpp' synopsis
Public Typedefs
Template parameters
Member functions

Introduction

The boost::wave::context<> object is the main user visible object of the Wave library. It exists to generate the pair of iterators, which while dereferenced return the preprocessed tokens. Additionally it is used to control other aspects of the preprocessing, such as

 include search paths, which define, where to search for files to be included with #include <...> and #include "..." directives
 which macros to predefine and which of the predefined macros to undefine
 several other options as for instance to control, whether to enable several extensions to the C++ Standard (as for instance variadics and placemarkers) or not.

Header wave/cpp_context.hpp synopsis

namespace boost {
namespace wave {

struct this_type {};

template <
typename Iterator, typename LexIterator,
typename InputPolicy, typename ContextPolicy,
typename Derived = this_type
>
class context : public InputPolicy
{
typedef <unspecified> token_sequence_type;

public:

typedef pp_iterator<context> iterator_type;
typedef Token token_type;
typedef typename token_type::position_type position_type;
typedef std::list<token_type> token_sequence_type;

// constructor
context(Iterator const &first_,
Iterator const &last_,
char const *fname = "<Unknown>",
ContextPolicy const& ctx_policy = ContextPolicy())

// iterator interface
iterator_type begin() const;
iterator_type begin(Iterator const &first,
Iterator const &last) const;
iterator_type end() const;

// maintain include paths
bool add_include_path(char const *path_);
bool add_sysinclude_path(char const *path_);
void set_sysinclude_delimiter();

size_t get_iteration_depth() const;

// maintain defined macros
template <typename String>
bool add_macro_definition(String macrostring,
bool is_predefined = false);
template <typename String>
bool add_macro_definition(StringT const &name,
position_type const& pos, bool has_params,
std::vector<token_type> &parameters,
token_sequence_type &definition,
bool is_predefined = false)

template <typename String>
bool is_defined_macro(String const &name) const;
template <typename String>
bool remove_macro_definition(String const &name,
bool even_predefined = false);
void reset_macro_definitions();
template <typename String>
bool get_macro_definition(
String const &name, bool &is_function_style,
bool
&is_predefined, position_type &pos,
std::vector<token_type> &parameters,
token_sequence_type &definition) const;

typedef <unspecified> macromap_type;
typedef typename macromap_type::name_iterator name_iterator;
typedef typename macromap_type::const_name_iterator const_name_iterator;

name_iterator macro_names_begin();
name_iterator macro_names_end();
const_name_iterator macro_names_begin() const;
const_name_iterator macro_names_end() const;

// other options
void set_language(language_support enable);
language_support get_language() const;

void set_max_include_nesting_depth(std::size_t new_depth);
size_t get_max_include_nesting_depth() const;

// get the Wave version information
static std::string get_version();
static std::string get_version_string();
};

} // namespace wave
} // namespace boost

Template parameters

The boost::wave::context object has three template parameters to specify the concrete behaviour of its operation. The following table describes these with more detail.

Template parameters required for the boost::wave::context class
Iterator

The type of the underlying iterator, through which the input stream is accessed.
This should be at least an forward_iterator type iterator.

LexIterator

The type of the lexer type to be used by the Wave library to identify tokens in the input stream.

InputPolicy

The type of the input policy class, which allows to customize the behaviour of the Wave library and the type of the iterators to use, when it comes to including and opening an included file.

ContextPolicy

The type of the context policy class, which allows to customize different aspects of the behaviour of the preprocessing. The requirements of this policy and the available customization points provided by this policy are described here.    

Derived

The type of the type being derived from the context type (if any). This template parameter is optional and defaults to 'this_type', which means that the context type will be used assuming no derived type exists.

For further information about the lexer type to use, please refer to the The Lexer Interface .

If the template parameter InputPolicy is omitted, the template boost::wave::iteration_context_policies::load_file_to_string is used. For further information about the input policy, please refer to the topic The Input Policy.

If the template parameter ContextPolicy is omitted, the boost::wave::context_policies::eat_whitespace policy type is used, implementing certain basic preprocessing hooks needed for normal (default) execution. For further information about the context policy, please refer to the topic The Context Policy.

Public Typedefs

The boost::wave::context template defines the following public typedefs, which may be useful while using this class:

Public typedefs defined by the boost::wave::context class
iterator_type

The Iterator template parameter provided, while the context class was instantiated.

lex_type

The LexIterator template parameter provided, while the context class was instantiated.

token_type

The token type, which is returned by the context generated iterators. This type is taken from the LexIterator template parameter provided, whicle the context class was instantiated.

input_policy_type

The InputPolicy template parameter provided, while the context class was instantiated.

context_policy_type

The ContextPolicy template parameter provided, while the context class was instantiated.

position_type

The type of the position information contained in every returned token, which describes the point, at which the given token was recognised.

Member functions

Constructor

    context(Iterator const &first, 
Iterator const &last, char const *filename,
ContextPolicy const& ctx_policy);

Constructs a context object on top of the input stream given by the pair of auxilliary iterators [first, last). The iterators should be at least forward_iterator type iterators. The filename parameter is to be supplied for informational purposes only. This string is used for indicating the token positions inside the input stream, it is not validated against the file system. If the filename parameter is not given it defaults to "<Unknown>". If the ctx_policy parameter isn't supplied it defaults to a default constructed ContextPolicy object.

Additionally the macro symbol table is filled with the predefined macros and the current reference directory is set to the path of the given filename. If this filename does not reference valid file system item, the current reference directory is set to the current system directory. (The current reference directory is the file system path, which is used as the target directory during the processing of #include "..." directives).

Iterator interface

The pair of iterators returned by the context::begin and context::end functions is the main interface for accessing the preprocessed tokens from the preprocessor engine. While iterating over the given iterator range [begin, end) there are returned the preprocessed C++ tokens, which are generated on the fly from the underlying input stream. The returned iterators are conceptually of forward_iterator type.

begin

    iterator_type begin();
iterator_type begin(Iterator const &first,
Iterator const &last, );

Initializes and returns the starting iterator for the preprocessed token stream. The iterator pair given by the parameters  [first, last) specifies the input sequence to preprocess. The first version of this function uses the input stream as specified by the constructor. 

end

    iterator_type end() const;

Initializes and returns the end of stream iterator to compare with for detecting the end of the preprocessed token stream.

Maintain include paths

The Wave library maintains two separate search paths for include files. A search path for user include files and a search path for system include files. Any directories specified with the add_include_path() function before the function set_sysinclude_delimiter() is called are searched only for the case of #include "..." directives, they are not searched for #include <file> directives. I.e. these directories are added to the user include search path.

If additional directories are specified with the add_include_path() function after a call to the function set_sysinclude_delimiter(), these directories are searched for all #include directives. I.e. these directories are added to the system include search path.

In addition, a call to the function set_sysinclude_delimiter() inhibits the use of the current reference directory as the first search directory for #include "..." directives. Therefore, the current reference directory is searched only, if it is requested explicitly with a call to the function add_include_path(".").

Callig both functions, the set_sysinclude_delimiter() and add_include_path(".") allows you to control precisely, which directories are searched before the current one and which are searched after.

These functions are modelled after the command line behaviour implemented by the popular gcc compiler.

add_include_path

    bool add_include_path(char const *path);

Adds the given file system path to the user include search paths. After a call to the set_sysinclude_delimiter() this function adds the given file system path to the system include search paths. Note though, that the given path is validated against the file system.

If the given path string does not form a name of a valid file system directory item, the function returns false. If the given path was successfully added to the include search paths in question, the function returns true.

add_sysinclude_path

    bool add_sysinclude_path(char const *path);

Adds the given file system path to the system include search paths. This function operates on the system include search path regardless of the mode of operation of the add_include_path(). Note though, that the given path is validated against the file system.

If the given path string does not form a name of a valid file system directory item, the function returns false. If the given path was successfully added to the system include search paths, the function returns true.

set_sysinclude_delimiter

    void set_sysinclude_delimiter();

Switches the mode, how the add_include_path() function operates. By default the given file system path is added to the user include search paths. After calling this function a subsequent call to the add_include_path() adds the given file system path to the system include search paths. Additionally it inhibits the the use of the current reference directory as the first search directory for #include "..." directives.

get_iteration_depth

    size_t get_iteration_depth() const;

Returns the actual include iteration depth, i.e. the current number of include levels to be popped from the include iteration context, before the main (topmost) iteration context is reached.

Maintain defined macros

add_macro_definition

    bool add_macro_definition(
std::string macrostring, bool is_predefined);

Adds a new macro definition to the macro symbol table. The parameter macrostring should contain the macro to define in the command line format, i.e. something like MACRO(x)=definition. The following table describes this format in more detail. The parameter is_predefined should be true while defining predefined macros, i.e. macros, which are not undefinable with an #undef directive from inside the preprocessed input stream. If this parameter is not given, it defaults to false.

Summary of possible formats for defining macros
MACRO define MACRO as 1
MACRO= define MACRO as nothing (empty)
MACRO=definition define MACRO as definition
MACRO(x) define MACRO(x) as 1
MACRO(x)= define MACRO(x) as nothing (empty)
MACRO(x)=definition define MACRO(x) as definition

The function returns false, if the macro to define already was defined and the new definition is equivalent to the existing one, it returns true, if the new macro was successfully added to the macro symbol table.

If the given macro definition resembles a redefinition and the new macro is not identical to the already defined macro (in the sense defined by the C++ Standard), the function throws a corresponding preprocess_exception.

is_defined_macro

    template <typename String>
bool is_defined_macro(String const &name) const;

Returns, if a macro with the given name is defined, i.e. if it is contained in the macro symbol table.

get_macro_definition

    template <typename String>
bool get_macro_definition(
String const &name, bool &is_function_style,
bool
&is_predefined, position_type &pos,
std::vector<token_type> &parameters,
token_sequence_type &definition) const;

Allows to retrieve all information known with regard to a macro definition. The parameter name specifies the name of the macro the information should be returned for. The parameters is_function_style and is_predefined return, whether the macro has been defined as a function style macro or as a predefined macro resp. The parameter pos will contain the position the macro was defined at. The parameter parameters will contain the names of the parameters the macro was defined with and the parameter definition will contain the token sequence for the definition (macro body).

The function returns true is the macro was defined and the requested information has been successfully retrieved, false otherwise.

remove_macro_definition

    template <typename String>
bool remove_macro_definition(
String const &name, bool even_predefined);

Removes the definition of the macro with the given name from the macro symbol table. This operation is equivalent to an #undef directive with this name executed from within the input stream. If the parameter even_predefined is true, then the macro is removed from the macro symbol table even, if it is defined as a predefined macro.

Note though, that the following macros are not undefinable in any case: __FILE__, __LINE__, __DATE__, __TIME__, __cplusplus, __STDC__. If the parameter even_predefined is not given, it defaults to false.

The function returns false, if the macro to undefine was not defined and returns true otherwise.

If the macro to remove may not be undefined (it is a predefined macro and the parameter even_predefined is set to false or it is one of the mentioned not undefinable macros above) the function throws a preprocess_exception.

reset_macro_definitions

    void reset_macro_definitions(); 

Resets the macro symbol table to it's initial state, i.e. undefines all user defined macros and inserts the internal predefined macros as described here.

macro_names_begin, macro_names_end

    name_iterator macro_names_begin();
name_iterator macro_names_end();
const_name_iterator macro_names_begin() const;
const_name_iterator macro_names_end() const;

The macro_names_begin and macro_names_end functions return iterators allowing to iterate on the names of all defined macros.

Get Version information

get_version

    static std::string get_version(); 

Returns a string containing the current Wave version formatted as 0xvvrsbbbb (this is a string representation of the equivalent hexadecimal number), where 'vv' is the version number, 'r' the release number, 's' the subrelease number and 'bbbb' the build number. A possible return value looks like 0x00910454. The returned value is the same as is inserted in the preprocessed token stream, when the predefined macro __WAVE_VERSION__ is expanded.

get_version_str

    static std::string get_version_str(); 

Returns a string containing the current Wave version formatted as "v.rr.ss.bbbb", where 'v' is the version number, 'rr' the release number, 'ss' the subrelease number and 'bbbb' the build number. A possible return value looks like "0.9.1.454". The returned value is the same as it will be inserted in the preprocessed token stream, when the predefined macro __WAVE_VERSION_STR__ is expanded.

Control extended options

set_language
get_language

    void set_language(language_support language, 
bool reset_macros = true);
language_support get_language() const;

This functions allows to specify the language mode, in which the Wave library should work. The possible language modes are defined by the enumerated type language_support:

    enum language_support {
// support flags for C++98
support_normal = 0x01,
support_cpp = support_normal,

// support flags for C99

support_option_long_long = 0x02,
support_option_variadics = 0x04,
support_c99 = support_option_variadics | support_option_long_long | 0x08,

// support flags for C++11

support_option_no_newline_at_end_of_file = 0x20,
support_cpp0x = support_option_variadics | support_option_long_long |
support_option_no_newline_at_end_of_file | 0x10,
support_cpp11 = support_cpp0x,

// support flags for C++17
support_option_has_include = 0x10000,
support_cpp1z = support_option_variadics | support_option_long_long |
support_option_no_newline_at_end_of_file | support_option_has_include | 0x20000,
support_cpp17 = support_cpp1z

// support flags for C++20
support_option_va_opt = 0x40000,
support_cpp2a = support_option_variadics | support_option_long_long |
support_option_no_newline_at_end_of_file | support_option_has_include |
support_option_va_opt | 0x80000,
support_cpp20 = support_cpp2a,

// the mask for the main language settings

support_option_mask = 0xFFC0,

// additional fine tuning of the general behavior
support_option_emit_contline = 0x0040,
support_option_insert_whitespace = 0x0080,
support_option_preserve_comments = 0x0100,
support_option_no_character_validation = 0x0200,
support_option_convert_trigraphs = 0x0400,
support_option_single_line = 0x0800,
support_option_prefer_pp_numbers = 0x1000,
support_option_emit_line_directives = 0x2000,
support_option_include_guard_detection = 0x4000,
support_option_emit_pragma_directives = 0x8000
};

When used with support_option_variadics the support for variadics, placemarkers and the operator _Pragma() is enabled in normal C++ mode. When used with the support_option_long_long the support for long long suffixes is enabled in C++ mode.

The support_c99 switches on the C99 language support, which enables variadics, placemarkers, the operator _Pragma and long long suffixes by default. Additionally it disables the C++ specific tokens such as '->*', '.*', '::' etc.

The support_option_... values are to be used in conjunction with the main language settings (support_normal and support_c99) .

During this process the library will recognize two forms of include guards:

 #ifndef INCLUDE_GUARD_MACRO
#define INCLUDE_GUARD_MACRO
...
#endif
or
#if !defined(INCLUDE_GUARD_MACRO)
#define INCLUDE_GUARD_MACRO
...
#endif

Note, that the parenthesis in the second form are optional (i.e. !defined INCLUDE_GUARD_MACRO will work as well). The code allows for any whitespace, newline and single '#' tokens before the #if/#ifndef and after the final #endif for the include guard to be recognized.

This flag is on by default, so if you do not want the library to automatically recognize include guards you will need to reset this explicitly.

If the parameter reset_macros is true the set_language function internally resets the list of defined macros, so please be careful not to call it after you have defined your own macros already.

set_max_include_nesting_depth
get_max_include_nesting_depth

    void set_max_include_nesting_depth(size_t new_depth); 
size_t get_max_include_nesting_depth() const;

This functions allow to set or to get the maximal possible include file nesting depth supported by the Wave library. The initial value for this is determined by the preprocessing constant WAVE_MAX_INCLUDE_LEVEL_DEPTH (see here).