cpp_re2c_lexer.hpp

Boost provides free peer-reviewed portable C++ source libraries. We emphasize libraries that work

/*=============================================================================
    Boost.Wave: A Standard compliant C++ preprocessor library

    Re2C based C++ lexer
    
    http://www.boost.org/

    Copyright (c) 2001-2008 Hartmut Kaiser. Distributed under the Boost
    Software License, Version 1.0. (See accompanying file
    LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
=============================================================================*/

#if !defined(CPP_RE2C_LEXER_HPP_B81A2629_D5B1_4944_A97D_60254182B9A8_INCLUDED)
#define CPP_RE2C_LEXER_HPP_B81A2629_D5B1_4944_A97D_60254182B9A8_INCLUDED

#include <string>
#include <cstdio>
#include <cstdarg>
#if defined(BOOST_SPIRIT_DEBUG)
#include <iostream>
#endif // defined(BOOST_SPIRIT_DEBUG)

#include <boost/concept_check.hpp>
#include <boost/assert.hpp>
#include <boost/spirit/include/classic_core.hpp>

#include <boost/wave/wave_config.hpp>
#include <boost/wave/language_support.hpp>
#include <boost/wave/token_ids.hpp>
#include <boost/wave/util/file_position.hpp>
#include <boost/wave/cpplexer/validate_universal_char.hpp>
#include <boost/wave/cpplexer/cpplexer_exceptions.hpp>
#include <boost/wave/cpplexer/token_cache.hpp>
#include <boost/wave/cpplexer/convert_trigraphs.hpp>

#include <boost/wave/cpplexer/cpp_lex_interface.hpp>
#include <boost/wave/cpplexer/re2clex/scanner.hpp>
#include <boost/wave/cpplexer/re2clex/cpp_re.hpp>
#if BOOST_WAVE_SUPPORT_PRAGMA_ONCE != 0
#include <boost/wave/cpplexer/detect_include_guards.hpp>
#endif

#include <boost/wave/cpplexer/cpp_lex_interface_generator.hpp>

// this must occur after all of the includes and before any code appears
#ifdef BOOST_HAS_ABI_HEADERS
#include BOOST_ABI_PREFIX
#endif

///////////////////////////////////////////////////////////////////////////////
namespace boost {
namespace wave {
namespace cpplexer {
namespace re2clex {

///////////////////////////////////////////////////////////////////////////////
// 
//  encapsulation of the re2c based cpp lexer
//
///////////////////////////////////////////////////////////////////////////////

template <typename IteratorT, typename PositionT = boost::wave::util::file_position_type>
class lexer 
{
public:
    typedef lex_token<PositionT>              token_type;
    typedef typename token_type::string_type  string_type;
    
    lexer(IteratorT const &first, IteratorT const &last, 
        PositionT const &pos, boost::wave::language_support language_);
    ~lexer();

    lex_token<PositionT>& get(lex_token<PositionT>&);
    void set_position(PositionT const &pos)
    {
        // set position has to change the file name and line number only
        filename = pos.get_file();
        scanner.line = pos.get_line();
//        scanner.column = scanner.curr_column = pos.get_column();
        scanner.file_name = filename.c_str();
    }
#if BOOST_WAVE_SUPPORT_PRAGMA_ONCE != 0
    bool has_include_guards(std::string& guard_name) const 
    { 
        return guards.detected(guard_name); 
    }
#endif
    
// error reporting from the re2c generated lexer
    static int report_error(Scanner const* s, int code, char const *, ...);

private:
    static char const *tok_names[];
    
    Scanner scanner;
    string_type filename;
    string_type value;
    bool at_eof;
    boost::wave::language_support language;
#if BOOST_WAVE_SUPPORT_PRAGMA_ONCE != 0
    include_guards<token_type> guards;
#endif
        
    static token_cache<string_type> const cache;
};

///////////////////////////////////////////////////////////////////////////////
// initialize cpp lexer 
template <typename IteratorT, typename PositionT>
inline
lexer<IteratorT, PositionT>::lexer(IteratorT const &first, 
        IteratorT const &last, PositionT const &pos, 
        boost::wave::language_support language_) 
:   filename(pos.get_file()), at_eof(false), language(language_)
{
    using namespace std;        // some systems have memset in std
    memset(&scanner, '\0', sizeof(Scanner));
    scanner.eol_offsets = aq_create();
    if (first != last) {
        scanner.first = scanner.act = (uchar *)&(*first);
        scanner.last = scanner.first + std::distance(first, last);  
    }
    scanner.line = pos.get_line();
    scanner.column = scanner.curr_column = pos.get_column();
    scanner.error_proc = report_error;
    scanner.file_name = filename.c_str();
    
#if BOOST_WAVE_SUPPORT_MS_EXTENSIONS != 0
    scanner.enable_ms_extensions = true;
#else
    scanner.enable_ms_extensions = false;
#endif

#if BOOST_WAVE_SUPPORT_VARIADICS_PLACEMARKERS != 0
    scanner.act_in_c99_mode = boost::wave::need_c99(language_);
#endif

#if BOOST_WAVE_SUPPORT_IMPORT_KEYWORD != 0
    scanner.enable_import_keyword = !boost::wave::need_c99(language_);
#else
    scanner.enable_import_keyword = false;
#endif

    scanner.detect_pp_numbers = boost::wave::need_prefer_pp_numbers(language_);
    scanner.single_line_only = boost::wave::need_single_line(language_);
}

template <typename IteratorT, typename PositionT>
inline
lexer<IteratorT, PositionT>::~lexer() 
{
    using namespace std;        // some systems have free in std
    aq_terminate(scanner.eol_offsets);
    free(scanner.bot);
}

///////////////////////////////////////////////////////////////////////////////
//  get the next token from the input stream
template <typename IteratorT, typename PositionT>
inline lex_token<PositionT>&
lexer<IteratorT, PositionT>::get(lex_token<PositionT>& result)
{
    if (at_eof) 
        return result = lex_token<PositionT>();  // return T_EOI

    unsigned int actline = scanner.line;
    token_id id = token_id(scan(&scanner));
    
    switch (static_cast<unsigned int>(id)) {
    case T_IDENTIFIER:
    // test identifier characters for validity (throws if invalid chars found)
        value = string_type((char const *)scanner.tok, 
            scanner.cur-scanner.tok);
        if (!boost::wave::need_no_character_validation(language))
            impl::validate_identifier_name(value, actline, scanner.column, filename); 
        break;
 
    case T_STRINGLIT:
    case T_CHARLIT:
    // test literal characters for validity (throws if invalid chars found)
        value = string_type((char const *)scanner.tok, 
            scanner.cur-scanner.tok);
        if (boost::wave::need_convert_trigraphs(language))
            value = impl::convert_trigraphs(value); 
        if (!boost::wave::need_no_character_validation(language))
            impl::validate_literal(value, actline, scanner.column, filename); 
        break;

#if BOOST_WAVE_SUPPORT_INCLUDE_NEXT != 0
    case T_PP_HHEADER:
    case T_PP_QHEADER:
    case T_PP_INCLUDE:
    // convert to the corresponding ..._next token, if appropriate
      {
          value = string_type((char const *)scanner.tok, 
              scanner.cur-scanner.tok);

      // Skip '#' and whitespace and see whether we find an 'include_next' here.
          typename string_type::size_type start = value.find("include");
          if (value.compare(start, 12, "include_next", 12) == 0)
              id = token_id(id | AltTokenType);
          break;
      }
#endif

    case T_LONGINTLIT:  // supported in C99 and long_long mode
        value = string_type((char const *)scanner.tok, 
            scanner.cur-scanner.tok);
        if (!boost::wave::need_long_long(language)) {
        // syntax error: not allowed in C++ mode
            BOOST_WAVE_LEXER_THROW(lexing_exception, invalid_long_long_literal, 
                value.c_str(), actline, scanner.column, filename.c_str());
        }
        break;

    case T_OCTALINT:
    case T_DECIMALINT:
    case T_HEXAINT:
    case T_INTLIT:
    case T_FLOATLIT:
    case T_FIXEDPOINTLIT:
    case T_CCOMMENT:
    case T_CPPCOMMENT:
    case T_SPACE:
    case T_SPACE2:
    case T_ANY:
    case T_PP_NUMBER:
        value = string_type((char const *)scanner.tok, 
            scanner.cur-scanner.tok);
        break;
        
    case T_EOF:
    // T_EOF is returned as a valid token, the next call will return T_EOI,
    // i.e. the actual end of input
        at_eof = true;
        value.clear();
        break;
        
    case T_OR_TRIGRAPH:
    case T_XOR_TRIGRAPH:
    case T_LEFTBRACE_TRIGRAPH:
    case T_RIGHTBRACE_TRIGRAPH:
    case T_LEFTBRACKET_TRIGRAPH:
    case T_RIGHTBRACKET_TRIGRAPH:
    case T_COMPL_TRIGRAPH:
    case T_POUND_TRIGRAPH:
        if (boost::wave::need_convert_trigraphs(language)) {
            value = cache.get_token_value(BASEID_FROM_TOKEN(id));
        }
        else {
            value = string_type((char const *)scanner.tok, 
                scanner.cur-scanner.tok);
        }
        break;
        
    case T_ANY_TRIGRAPH:
        if (boost::wave::need_convert_trigraphs(language)) {
            value = impl::convert_trigraph(
                string_type((char const *)scanner.tok)); 
        }
        else {
            value = string_type((char const *)scanner.tok, 
                scanner.cur-scanner.tok);
        }
        break;
        
    default:
        if (CATEGORY_FROM_TOKEN(id) != EXTCATEGORY_FROM_TOKEN(id) ||
            IS_CATEGORY(id, UnknownTokenType))
        {
            value = string_type((char const *)scanner.tok, 
                scanner.cur-scanner.tok);
        }
        else {
            value = cache.get_token_value(id);
        }
        break;
    }
    
//     std::cerr << boost::wave::get_token_name(id) << ": " << value << std::endl;

    // the re2c lexer reports the new line number for newline tokens
    result = token_type(id, value, PositionT(filename, actline, scanner.column));

#if BOOST_WAVE_SUPPORT_PRAGMA_ONCE != 0
    return guards.detect_guard(result);
#else
    return result;
#endif
}

template <typename IteratorT, typename PositionT>
inline int 
lexer<IteratorT, PositionT>::report_error(Scanner const *s, int errcode, 
    char const *msg, ...)
{
    BOOST_ASSERT(0 != s);
    BOOST_ASSERT(0 != msg);

    using namespace std;    // some system have vsprintf in namespace std
    
    char buffer[200];           // should be large enough
    va_list params;
    va_start(params, msg);
    vsprintf(buffer, msg, params);
    va_end(params);
    
    BOOST_WAVE_LEXER_THROW_VAR(lexing_exception, errcode, buffer, s->line, 
        s->column, s->file_name);
//    BOOST_UNREACHABLE_RETURN(0);
    return 0;
}

///////////////////////////////////////////////////////////////////////////////
//   
//  lex_functor
//   
///////////////////////////////////////////////////////////////////////////////
     
template <typename IteratorT, typename PositionT = boost::wave::util::file_position_type>
class lex_functor 
:   public lex_input_interface_generator<
        typename lexer<IteratorT, PositionT>::token_type
    >
{    
public:
    typedef typename lexer<IteratorT, PositionT>::token_type token_type;
    
    lex_functor(IteratorT const &first, IteratorT const &last, 
            PositionT const &pos, boost::wave::language_support language)
    :   re2c_lexer(first, last, pos, language)
    {}
    virtual ~lex_functor() {}
    
// get the next token from the input stream
    token_type& get(token_type& result) { return re2c_lexer.get(result); }
    void set_position(PositionT const &pos) { re2c_lexer.set_position(pos); }
#if BOOST_WAVE_SUPPORT_PRAGMA_ONCE != 0
    bool has_include_guards(std::string& guard_name) const 
        { return re2c_lexer.has_include_guards(guard_name); }
#endif    

private:
    lexer<IteratorT, PositionT> re2c_lexer;
};

///////////////////////////////////////////////////////////////////////////////
template <typename IteratorT, typename PositionT>
token_cache<typename lexer<IteratorT, PositionT>::string_type> const
    lexer<IteratorT, PositionT>::cache = 
        token_cache<typename lexer<IteratorT, PositionT>::string_type>();
    
}   // namespace re2clex

///////////////////////////////////////////////////////////////////////////////
//  
//  The new_lexer_gen<>::new_lexer function (declared in cpp_lex_interface.hpp)
//  should be defined inline, if the lex_functor shouldn't be instantiated 
//  separately from the lex_iterator.
//
//  Separate (explicit) instantiation helps to reduce compilation time.
//
///////////////////////////////////////////////////////////////////////////////

#if BOOST_WAVE_SEPARATE_LEXER_INSTANTIATION != 0
#define BOOST_WAVE_RE2C_NEW_LEXER_INLINE
#else
#define BOOST_WAVE_RE2C_NEW_LEXER_INLINE inline
#endif 

///////////////////////////////////////////////////////////////////////////////
//
//  The 'new_lexer' function allows the opaque generation of a new lexer object.
//  It is coupled to the iterator type to allow to decouple the lexer/iterator 
//  configurations at compile time.
//
//  This function is declared inside the cpp_lex_token.hpp file, which is 
//  referenced by the source file calling the lexer and the source file, which
//  instantiates the lex_functor. But is is defined here, so it will be 
//  instantiated only while compiling the source file, which instantiates the 
//  lex_functor. While the cpp_re2c_token.hpp file may be included everywhere,
//  this file (cpp_re2c_lexer.hpp) should be included only once. This allows
//  to decouple the lexer interface from the lexer implementation and reduces 
//  compilation time.
//
///////////////////////////////////////////////////////////////////////////////

template <typename IteratorT, typename PositionT>
BOOST_WAVE_RE2C_NEW_LEXER_INLINE
lex_input_interface<lex_token<PositionT> > *
new_lexer_gen<IteratorT, PositionT>::new_lexer(IteratorT const &first,
    IteratorT const &last, PositionT const &pos, 
    boost::wave::language_support language)
{
    using re2clex::lex_functor;
    return new lex_functor<IteratorT, PositionT>(first, last, pos, language);
}

#undef BOOST_WAVE_RE2C_NEW_LEXER_INLINE

///////////////////////////////////////////////////////////////////////////////
}   // namespace cpplexer
}   // namespace wave
}   // namespace boost 
     
// the suffix header occurs after all of the code
#ifdef BOOST_HAS_ABI_HEADERS
#include BOOST_ABI_SUFFIX
#endif

#endif // !defined(CPP_RE2C_LEXER_HPP_B81A2629_D5B1_4944_A97D_60254182B9A8_INCLUDED)