Throughout chapter 2 of SICP and particularly section 2.2. and its corresponding solutions on the Scheme wiki, I am regularly seeing the same pattern when dealing with tree structures:
(cond ((null? tree) nil)
((not (pair? tree)) (foo tree))
(else (cons (bar (car tree))
(bar (cdr tree)))))
The second line of these structures are a frequent source of confusion for me. In any other language, I'd expect to see the final two branches swapped as follows:
(cond ((null? tree) nil)
((pair? tree) (cons (bar (car tree))
(bar (cdr tree))))
(else (foo tree)))
So why am I not seeing that in Scheme? Is there some benefit to preferring (not? (pair? x) [...]) over swapping the branches and removing the not?.
Cases usually start with the smallest or most special case and and end with the largest or most general.
With a tree, that would be