A function f() returns multiple values, which are to be fed into a second function g(). I cannot change f(), but I can change g() and all calls to it. I’m wondering if it’s possible to say which variant performs best?

• Function header with one table argument

  function g(arg)
  --code
  end
  

  and function call with table constructor

  g({f()})
  

• Function header with variable number of arguments

  function g(...)
  --code
  end
  

  and regular function call

  g(f())
  

• Function header with multiple variables

  function g(x, y, z)
  --code
  end
  

  and function call with unpack

  g(unpack(f()))
  

• A completely different variant?