Seasoned Schemer's get-first, get-next, and wa

(define get-first
  (lambda (l)
    (call-with-current-continuation
      (lambda (here)
        (set! leave here)
        (waddle l)
        (leave (quote ()))))))

(define get-first
  (lambda (l)
    (call-with-current-continuation
      (lambda (here)
        (set! leave here)
        (leave (waddle l))))))

For anybody not familiar with the book "The Seasoned Schemer", get-first, get-next, and waddle (last two not defined here) are procedures to apparently model coroutines to iterate through a tree passed to waddle that yields leaves only. Just prior to waddle's yield on its second-to-last re-entry, it sets the re-entry point to where it will only ever return the pure value '() i.e. instead of yielding '() , the actual value of waddle is '() , as if it were a pure function all along.

With this in mind, we can see what get-first sets up... When waddle returns "for real", it will be inside the call/cc in get-first and then (leave (quote ())) is the value of get-first (and, in turn, this leave is intended to return to get-next on the last iteration, therefore it is get-next that does the "actual" return of '()).

So why is the second version not equivalent, where waddle's value of '() would be the argument to leave?

The confusion is because "leave" is not the function I want it to be, but the function it evaluates to when it's evaluated, which appears to be left-to-right and thus before "waddle". That means it evaluates to what it was just set to, in the statement prior.

Moral: beware when using functions that are subject to redefining WITHIN the call to the function! If this was on a right-to-left interpreter, waddle would be evaluated before the symbol leave was looked up as the function that "leaves" to wherever, during which time it would be set to a DIFFERENT function.