[plt-scheme] critique of some code

Todd, your solution looks OK to me, and you have good tests for some 
edge cases.

FWIW, here's how I would have done it, if I were constrained to using 
"substring" (which will do lots of allocations), because I love using 
named-"let" as much as possible:

(define (string-index-of str sub)
  (let* ((str-len   (string-length str))
         (sub-len   (string-length sub))
         (max-start (- str-len sub-len)))
    (let loop ((start 0))
      (cond ((> start max-start) #f)
            ((string=? (substring str start (+ start sub-len))
                       sub)
             start)
            (else (loop (+ 1 start)))))))

For efficiency, but constrained to using only R5RS Scheme primitives, 
I'd try two more approaches: (1) bursting both strings into lists of 
characters and comparing them as lists somehow; (2) using "string-ref" 
on each character, rather than "substring", to reduce allocations.

Small nit: I'd return #f rather than -1.  Java uses the cryptic -1 value 
because its crude type system encourages it to, but you don't have that 
problem.  Using #f also conveniently means that you can use the result 
of the procedure as a Boolean value, including with the sometimes very 
convenient "cond" "=>" syntax.

-- 
http://www.neilvandyke.org/