Re^2: Order of Precedence in Parse::RecDescent grammar

[suaveant]

This seems to work... I am guessing it is basically the shortcut way to do what blokhead proposed... I was able to get it to return a little less depth in the tree, though it isn't quite perfect.

Here is my grammar, for posterity's sake.

my $grammar = <<'GRAMMAR'

  {
      use strict;
      use warnings;

      sub treeify {
         my $t = shift;
#     my($l,$r) = (shift,shift);
     while(@_) {
         my($l,$r) = (shift,shift);
         
         $t = [ 'op',$l, (ref $t eq 'ARRAY' && !$#$t ? @$t : $t), (ref
+ $r eq 'ARRAY' && !$#$r ? @$r : $r) ]
     }
         return $t;
      }
   }

  startrule: bin_op

  bin_op: comp_op 
 # Lowest precendance.
  comp_op: <leftop: add_op COMP add_op>
              { treeify(@{$item[1]}); }

   add_op: <leftop: prod_op SUM prod_op >
              { treeify(@{$item[1]}); }

   prod_op : <leftop: mod_op PROD mod_op >
              { treeify(@{$item[1]}); }
   # Highest precendance.

   mod_op : <leftop: term MOD term >
              { treeify(@{$item[1]}); }

   SUM      : '+' | '-'
   PROD     : '*' | '/'
   MOD      : '%'
   COMP     : />=?|<=?|!=|==?|le|ge|eq|ne|lt|gt/


  term: function 
            | '(' bin_op ')' { $item[2] }
            | number
            | string

  if: /if/i '(' list_bin_op list_bin_op bin_op ')' { ['func',@item[1,3
+..5]] }
  concat: /concat/i '(' list_bin_op(s) bin_op ')' { ['func',$item[1],@
+{$item[3]},$item[4]] }
  left: /left/i '(' list_bin_op bin_op ')' { ['func',@item[1,3,4]] }
  right: /right/i '(' list_bin_op bin_op ')' { ['func',@item[1,3,4]] }
  ifnull: /ifnull/i '(' list_bin_op bin_op ')' { ['func',@item[1,3,4]]
+ }

  function: if | concat | left | right | ifnull

  number: /[+-]?\d+/ { $item[1] }
  string: { $_ = extract_quotelike($text); chop; substr($_,0,1,''); $_
+; } { [@item] }

#$thisparser->startrule($item[2],1,$type)

  list_bin_op: bin_op ',' { $item[1] }

GRAMMAR
;

This is actually parsing MySQL-like statements, so there is a bit in there for some function definitions, seems to work, though. Thanks all!

[ikegami]

You're giving % higher precedence than *. Are you sure that's correct? According to your grammar, 4*2%3 is equivalent to 4*(2%3) (8) rather than (4*2)%3 (0).

The words op and func are probably redundant. Do you really need to know that something is a op rather than specifically an addition, substratction, multiplication, etc.?

Does extract_quotelike remove the quotes? If so, you have a problem if someone create a string called op or func. That's why I had term in the tree. This is the only level you removed, and it hurts you.

Update: Oops, I'm mistaken on that last one. If it's a ref, it's an op or func. It's it's not, it's a number or string. It's not very orthogonal to flattern terms, but it works for this grammar.

[suaveant]

Oh.. my bad... apparently they are equal in precedence... oops. Oh well... that wasn't the only problem, so it was still useful :)