5.1 Output control

When Felix emits primitive functions, by default it will have to put parentheses around every subexpression, since it does not actually parse the C code. This would make the output look like Scheme rather than C++!

5.1.1 Specifying precedence

You can specify the precedence to avoid gratuitous parentheses like this:

  fun add: int * int -> int = "$1:add + $2:mult" is add;

The precedences are named, from tightest binding to loosest binding:

  atom
  primary
  postfix
  unary
  cast
  pm
  mult
  add
  shift
  rel
  eq
  band
  bxor
  bor
  and
  xor
  or
  cond
  assign
  comma
  expr

5.1.2 Special cases

Felix does parse some special cases (exactly as written no spaces allowed):

  $1++
  $1--

  ~$1
  +$1
  -$1
  !$1
  &$1
  *$1
  ++$1
  --$1
  $1+$2
  $1-$2
  $1*$2
  $1/$2
  $1%$2

  $1<<$2
  $1>>$2
  $1&$2
  $1|$2
  $1^$2
  $1&&$2
  $1||$2

  $1+=$2
  $1-=$2
  $1*=$2
  $1/=$2
  $1%=$2
  $1<<=$2
  $1>>=$2
  $1&=$2
  $1|=$2
  $1^=$2

  $1<$2
  $1>$2
  $1>=$2
  $1<=$2
  $1==$2
  $1!=$2

  $1
  $1[$2]
  $1->$2
  $1.*$2
  $1->*$2
  $1,$2

  &::std::cout
  &::std::cerr
  $1.size
  $1.data[$2]

  ::flx::rtl::strutil::str<int>
  ::flx::rtl::strutil::str<#1>
  static_cast<#0>
  reinterpret<?1>

and the form

  name(expr)

which is assumed to be a function application. Additional forms may be recognised in the future. If you don't mis-specify precedences, Felix should always generated correct code (it just might have more parentheses than it needs).