refs rather than scalars in ast

$VAR1 = [
          'S',
          'pair',
          [
            'S',
            'bracketed',
            '<',
            [
              ${\$VAR1->[0]},
              ${\$VAR1->[1]},
              [
                ${\$VAR1->[0]},
                ${\$VAR1->[1]},
                [
                  'S',
                  'non-bracketed',
                  [
                    'non_parens',
                    'non_parens',
                    ' words in '
                  ]
                ],
                [
                  ${\$VAR1->[2][0]},
                  ${\$VAR1->[2][1]},
                  '<',
                  [
                    ${\$VAR1->[2][3][2][2][0]},
                    ${\$VAR1->[2][3][2][2][1]},
                    [
                      ${\$VAR1->[2][3][2][2][2][0]},
                      ${\$VAR1->[2][3][2][2][2][1]},
                      'nested angle'
                    ]
                  ],
                  '>'
                ]
              ],
              [
                ${\$VAR1->[2][3][2][2][0]},
                ${\$VAR1->[2][3][2][2][1]},
                [
                  ${\$VAR1->[2][3][2][2][2][0]},
                  ${\$VAR1->[2][3][2][2][2][1]},
                  ' brackets '
                ]
              ]
            ],
            '>'
          ],
          [
            ${\$VAR1->[2][3][2][2][0]},
            ${\$VAR1->[2][3][2][2][1]},
            [
              ${\$VAR1->[2][3][2][2][2][0]},
              ${\$VAR1->[2][3][2][2][2][1]},
              '
'
            ]
          ]
        ];

refs-in-ast.t

use 5.010;
use strict;
use warnings;

use Data::Dumper;

use Marpa::R2;

my $g = Marpa::R2::Scanless::G->new( { 
    bless_package => 'main',
    source => \(<<'END_OF_SOURCE'),
:default ::= action => [symbol, name, values]
lexeme default = action => [symbol, name, values] latm => 1

        S ::= S S       name => 'pair' |
              '<' S '>' name => 'bracketed' |
              '<' '>'   name => 'empty bracketed' |
              non_parens name => 'non-bracketed'

        non_parens ~ non_paren*
        non_paren ~ [^<>]

END_OF_SOURCE
} );

my $input = <<EOI;
< words in <nested angle> brackets >
EOI

my $r = Marpa::R2::Scanless::R->new( { grammar => $g } );
$r->read(\$input);
say Dumper ${ $r->value() };