# See: Tieins.eyp
# See: http://www.gnu.org/software/bison/manual/html_mono/bison.html#Lexical-Tie_002dins
# for a more detailed explanation.

%strict

%token ID INT INTEGER 

%syntactic token HEX

%{
my %st;
%}

%lexer  {
    my $hexflag = $self->{HEXFLAG};

    m{\G\s*(\#.*)?}gc;

    m{\G(HEX\b|INT\b)}igc and return (uc($1), $1);

    m{(\G\d+)}gc and return ('INTEGER', $hexflag? hex($1) : $1);


    m{\G([a-zA-Z_]\w*)}gc and do {
        my $match = $1;
        $hexflag and !exists($st{$match}) and $match =~ m{^([A-F0-9]+$)}gc and return ('INTEGER', hex($match)); 
        return ('ID', $1);
    };

    m{\G(.)}gc and return ($1, $1);

    return('',undef);
  }

%right '='
%left '+'

%tree bypass alias

%%
stmt:
    decl <* ';'> expr <%name EXPS + ';'>
      {
        # make the symbol table an attribute 
        # of the root node
        $_[2]->{st} = { %st };
        $_[2];
      }
;

decl:
    INT ID <+ ','>
      {
        # insert identifiers in the symbol table
        $st{$_->{attr}} = 1 for $_[2]->children();
      }
;

expr:
    %name ID
    ID
  | %name NUM
    INTEGER
  | %name HEX
    HEX '(' { $_[0]->{HEXFLAG} = 1; } $expr ')'
      { 
        $_[0]->{HEXFLAG} = 0; 
        $expr;
      }
  | %name ASSIGN
    id '=' expr 
  | %name PLUS
    expr '+' expr
;

id : ID
;

%%

# Context-dependant lexer

=head1 SYNOPSIS

Compile it with:

                   $ eyapp -C SemanticInfoInTokens.eyp

Run it with:

                  $ ./SemanticInfoInTokens.pm -t -i -f inputforsemanticinfo.txt

try also:

            ./SemanticInfoInTokens.pm -t -i -f inputforsemanticinfo2.txt

=head1 THE TYPENAME-IDENTIFIER PROBLEM WHEN PARSING THE C<C> LANGUAGE

The C language has a context dependency: the way an identifier is used depends
on what its current meaning is. For example, consider this:

  T(x);

This looks like a function call statement, but if C<T> is a typedef name, then
this is actually a declaration of C<x>. How can a parser for C decide how to
parse this input?

Here is another example:

  {
    T * x;
    ...
  }

What is this, a declaration of C<x> as a pointer to C<T>, 
or a void multiplication of the variables C<T> and C<x>?

The usual method to solve this problem is to have two different token types, C<ID> and C<TYPENAME>.
When the lexer finds an identifier, it looks up in the symbol table 
the current declaration of the identifier in order to 
decide which token type to return: C<TYPENAME> if the
identifier is declared as a typedef, C<ID> otherwise.

=head1 THIS EXAMPLE

One way to handle context-dependency is the lexical tie-in: a flag which is set
by the semantic actions, whose purpose is to alter the way tokens are parsed.

In this "Calc"-like example we have a language with a special construct C<hex
(hex-expr)>. After the keyword C<hex> comes an C<expression> in parentheses in
which all integers are hexadecimal. In particular, strings in C</[A-F0-9]+/>
like C<A1B> must be treated as an hex integer unless they were previously
declared.

Here the lexer looks at the value of the hexflag attribute; when it is nonzero,
all integers are parsed in hexadecimal, and tokens starting with letters are
parsed as integers if possible.

=head1 SEE ALSO

=over 2

=item *  File: Tieins.eyp

=item * L<http://www.gnu.org/software/bison/manual/html_mono/bison.html#Lexical-Tie_002dins>

=item * L<http://en.wikipedia.org/wiki/The_lexer_hack>

=item * L<http://eli.thegreenplace.net/2007/11/24/the-context-sensitivity-of-cs-grammar/>

=back