expr.cxx

ecos实时嵌入式操作系统

        }
        if (!isdigit(current_char)) {
            throw CdlParseException("Invalid floating point constant, expected a digit for the exponent.\n" +
                                    get_error_location());
        }
        is_float = true;
        do {
            tmp += (char) current_char;
            next_char();
        } while(isdigit(current_char));
    }

    if (is_float) {
        if (!Cdl::string_to_double(tmp, current_double)) {
            throw CdlParseException("Invalid floating point constant `" + tmp + "'.\n" + get_error_location());
        } else {
            current_token = T_Double;
        }
    } else {
        if (!Cdl::string_to_integer(tmp, current_int)) {
            throw CdlParseException("Invalid integer constant `" + tmp + "'.\n" + get_error_location());
        } else {
            current_token = T_Integer;
        }
    }
    
    check_number_termination();
    CYG_REPORT_RETURN();
}

//}}}
//{{{  process_alphanumeric()           

// The start of an alphanumeric sequence has been detected. This may
// be a reference, a function call, or an operator like eq or to. All
// such sequences must be a valid C preprocessor name, so the only
// characters allowed are underscore, upper and lower case characters,
// and digits. The first character cannot be a digit, but that has
// been checked already.
//
// Some care has to be taken with locale's, the C library may decide
// that a character is a letter even though the same character is not
// valid as far as the preprocessor is concerned.
static void
process_alphanumeric()
{
    CYG_REPORT_FUNCNAME("process_alphanumeric");

    do {
       current_reference += (char) current_char;
       next_char();
    } while (('_' == current_char) || isdigit(current_char) ||
             (('a' <= current_char) && (current_char <= 'z')) ||
             (('A' <= current_char) && (current_char <= 'Z')));

    CYG_REPORT_RETURN();
}

//}}}
//{{{  process_special()                

// Usually an alphanumeric sequence of characters is a reference, e.g.
// CYGPKG_KERNEL. However there are only so many special characters
// available so some operators are implemented as a sequence, e.g. 
// "to". CDL also supports functions like is_substr().
//
// The data will have been collected into the current_reference string
// by a call to process_alphanumeric().

static bool
process_special()
{
    CYG_REPORT_FUNCNAMETYPE("process_special", "special %d");
    bool result = false;
    
    if ("to" == current_reference) {
        current_token  = T_Range;
        result = true;
    } else if ("implies" == current_reference) {
        current_token  = T_Implies;
        result = true;
    } else if ("xor" == current_reference) {
        current_token  = T_Xor;
        result = true;
    } else if ("eqv" == current_reference) {
        current_token  = T_Eqv;
        result = true;
    } else if (CdlFunction::is_function(current_reference.c_str(), current_function_id)) {
        current_token  = T_Function;
        result = true;
    }

    if (result) {
        current_special     = current_reference;
        current_reference   = "";
    }
    CYG_REPORT_RETVAL(result);
    return result;
}

//}}}

//}}}
//{{{  next_token()                     

// ----------------------------------------------------------------------------
// Work out what the next token is. This includes the handling of
// strings, integers, doubles, and references.
static void
next_token()
{
    CYG_REPORT_FUNCNAMETYPE("next_token", "token %d");

    // Make sure there is no dross left lying around from the previous call.
    current_token       = T_Invalid;
    current_string      = "";
    current_reference   = "";
    current_special     = "";
    current_int         = 0;
    current_double      = 0.0;
    current_format      = CdlValueFormat_Default;
    current_function_id = 0;

    // Skip leading white space. This includes newlines, tabs, etc,
    // consider the case of:
    //    ...
    //    legal_values {
    //        1
    //        2
    //        4
    //        ..
    //    }
    //    ...
    // which is perfectly legitimate. White space inside strings
    // is handled by the string literal code, and does not get filtered
    // out here.
    //
    // Exactly which characters are white-space is implementation-defined,
    // so a special check for EOF is in order.
    while ((EOF != current_char) && isspace(current_char)) {
        next_char();
    }

    // Remember the token starting point. next_char() has actually moved
    // the index on by one.
    token_start = current_index - 1;

    // The simple cases can be handled inline, the more complicated cases
    // involve other functions
    switch(current_char) {

      case EOF:
          current_token = T_EOD;
          break;

      case '"':
          process_string();
          break;

      case '(':
          current_token = T_OpenBracket;
          next_char();
          break;

      case ')':
          current_token = T_CloseBracket;
          next_char();
          break;

          // At this level it is not possible to distinguish between
          // unary and binary operators, so no attempt is made to
          // turn - and + into part of a number.
      case '-':
          current_token = T_Minus;
          next_char();
          break;

      case '+':
          current_token = T_Plus;
          next_char();
          break;

      case '*':
          current_token = T_Times;
          next_char();
          break;

      case '/':
          current_token = T_Divide;
          next_char();
          break;

      case '!':
          next_char();
          if ('=' == current_char) {
              current_token = T_NotEqual;
              next_char();
          } else {
              current_token = T_Exclamation;
          }
          break;

      case '~':
          current_token = T_Tilde;
          next_char();
          break;

      case '?':
          current_token = T_Questionmark;
          next_char();
          break;

      case '%':
          current_token = T_Remainder;
          next_char();
          break;

      case '<':
          next_char();
          if ('<' == current_char) {
              current_token = T_LeftShift;
              next_char();
          } else if ('=' == current_char) {
              current_token = T_LessEqual;
              next_char();
          } else {
              current_token = T_LessThan;
          }
          break;

      case '>':
          next_char();
          if ('>' == current_char) {
              current_token = T_RightShift;
              next_char();
          } else if ('=' == current_char) {
              current_token = T_GreaterEqual;
              next_char();
          } else {
              current_token = T_GreaterThan;
          }
          break;

      case '=':
          next_char();
          if ('=' != current_char) {
              throw CdlParseException(std::string("Incomplete == operator in expression.\n") + get_error_location());
          } else {
              current_token = T_Equal;
              next_char();
          }
          break;

      case '&':
          next_char();
          if ('&' == current_char) {
              current_token = T_And;
              next_char();
          } else {
              current_token = T_BitAnd;
          }
          break;

      case '^':
          current_token = T_BitXor;
          next_char();
          break;

      case '|':
          next_char();
          if ('|' == current_char) {
              current_token = T_Or;
              next_char();
          } else {
              current_token = T_BitOr;
          }
          break;

      case ':':
          current_token = T_Colon;
          next_char();
          break;

      case '.':
          current_token = T_StringConcat;
          next_char();
          break;

      case ',':
          current_token = T_Comma;
          next_char();
          break;
        
      default:
          // String constants have been handled already. The only
          // valid tokens that are left are numbers, references,
          // "specials" such as the range and string equality
          // operators, and functions.
          //
          // Numbers should begin with a digit (plus and minus are
          // tokenized separately).
          //
          // References must be valid C preprocessor symbols, i.e.
          // they must begin with either a letter or an underscore.
          // The range operator is handled most conveniently as
          // a special case of a reference.
          if (isdigit(current_char)) {
              process_number();
          } else if (('_' == current_char) ||
                     (('a' <= current_char) && (current_char <= 'z')) ||
                     (('A' <= current_char) && (current_char <= 'Z'))) {
              process_alphanumeric();
              if (!process_special()) {
                  current_token = T_Reference;
              }
          } else {
              std::string msg = "Unexpected character '";
              msg += (char) current_char;
              msg += "' in expression.\n";
              msg += get_error_location();
              throw CdlParseException(msg);
          }
          break;
    }

    CYG_REPORT_RETVAL(current_token);
}

//}}}
//{{{  initialise_tokenisation()        

// ----------------------------------------------------------------------------
// This is called at the start of expression parsing. It
// sets up the appropriate statics, and provides initial
// values for current_char and current_token.
static void
initialise_tokenisation(std::string data, int index)
{
    CYG_REPORT_FUNCNAME("initialise_tokenization");

    current_data        = data;
    current_index       = static_cast<unsigned int>(index);
    token_start         = current_index;
    next_char();
    next_token();

    CYG_REPORT_RETURN();
}

//}}}

//}}}
//{{{  Syntactic analysis               

// ----------------------------------------------------------------------------
// Syntactic analysis.
//
// The BNF of CDL expressions is something like this:
//
//   <expression>   ::= <conditional>
//   <conditional>  ::= <implies> ? <conditional> : <conditional> | <implies>
//   <implies>      ::= <eqv>    [<implies op>  <implies>]      implies
//   <eqv>          ::= <or>     [<eqv op>      <eqv>]          xor, eqv        
//   <or>           ::= <and>    [<or op>       <or>]           ||
//   <and>          ::= <bitor>  [<and op>      <and>]          &&
//   <bitor>        ::= <bitxor> [<bitor op>    <bitor>]        |
//   <bitxor>       ::= <bitand> [<bitxor op>   <bitxor>]       ^
//   <bitand>       ::= <eq>     [<bitand op>   <and>]          &
//   <eq>           ::= <comp>   [<eq op>       <eq>]           == !=
//   <comp>         ::= <shift>  [<comp op>     <comp>]         < <= > >=
//   <shift>        ::= <add>    [<shift op>    <shift>]        << >>
//   <add>          ::= <mult>   [<add op>      <add>]          + - .
//   <mult>         ::= <unary>  [<mult op>     <mult>]         * / %
//   <unary>        ::= -<unary> | +<unary> | !<unary> | *<unary> | ?<unary> |
//                      ~<unary> |
//                      <string constant> | <integer constant> |
//                      <double constant> | <reference> |
//                      ( <expression> ) | <function>
//
// There are separate functions for each of these terms.

// A forward declaration, needed for bracketed subexpressions.
static void parse_expression(CdlExpression);

// A utility to add a reference to the current expression, returning
// the index.
static int
push_reference(CdlExpression expr, const std::string& reference)
{
    CYG_REPORT_FUNCNAMETYPE("push_reference", "new index %d");