Re-writing church numerals function - scheme

Question

In SICP it defines the church numerals for positive numbers as follows:
(define zero (lambda (f) (lambda (x) x)))
(define (add-1 n)
(lambda (f) (lambda (x) (f (n f) x))))
The following is my 'best attempt' to rewrite this for my own understanding, here passing explicit arguments to one function:
(define (church f x n)
(cond
((= n 0) x) ; zero case: return x
(else (f (church f x (- n 1)))))) ; otherwise f(f(f...(x))) n times
(church square 3 2)
81
And then redefining zero I would have:
(define (zero2 f)
(lambda (x) (church f x 0)))
And add-one as:
(define (add-1 n f)
(lambda (x) (church f x (+ n 1))))
Or, if we have to defer the f argument then adding a wrapper-lambda:
(define (add-1 n)
(lambda (f) (lambda (x) (church f x (+ n 1)))))
Do I have a correct understanding of this? if so, why the oh-so-complicated-syntax at the top for the add-1 or zero procedures (note: I'm guessing it's not that complicated and I'm just not fully understanding what it's doing). Any help would be greatly appreciated!

Answer 1

lambda calculus is a sub set of Scheme that does not allow more than one argument and lambda. With combinations of lambdas you can make any construct:
(define false (lambda (true) (lambda (false) false)))
(define true (lambda (true) (lambda (false) true)))
(define if (lambda (pred) (lambda (consequence) (lambda (alternative) ((pred consequence) alternative)))))
You might be wondering why I allow define since it isn't lambda. Well you don;t need it. It is just for convenience since with it you can try it out:
(((if true)
'result-true)
'result-false)
; ==> result-true
Instead of using the totally equal version:
((lambda (pred)
(lambda (consequence)
(lambda (alternative)
((pred consequence) alternative))))
(lambda (true) (lambda (false) true))
'result-true
'result-false)
Your function church is not lambda calculus since it does not return a church number and it takes more than one argument which is a violation. I have seen scheme functions to produce chuck numbers but any chuck number you should be able to do this to get the integer value:
((church-number add1) 0)
eg. zero:
(((lambda (f) (lambda (x) x)) add1) 0) ; ==> 0

Answer 2

Your version presupposes the existence of primitives like cond, 0, 1, =, and -. The point of all this is to show that you can implement such primitives starting from nothing but lambda.

Answer 3

SICP defines the Church numerals for positive numbers as follows:
(define zero (lambda (f) (lambda (x) x)))
(define (add-1 n)
(lambda (f) (lambda (x) (f (n f) x))))
No, it doesn't. The correct definitions are
(define zero (lambda (f) (lambda (x) x)))
(define (add-1 n)
(lambda (f) (lambda (x) (f ((n f) x)))))
f is a "successor step", and x is "zero value".
(f ((n f) x)) means, do with f and x whatever n would be doing with f and x, and then do f one more time to the result.
In other words, transform the "zero value" with the "successor step" function one more times than n would be transforming it.
Now,
> ((zero add1) 0)
0
> (((add-1 zero) add1) 0)
1
> (((add-1 (add-1 zero)) add1) 0)
2
etc. Or,
> (define plus1 (lambda (x) (cons '() x)))
> ((zero plus1) '(NIL))
'(NIL)
> (((add-1 zero) plus1) '(NIL))
'(() NIL)
> (((add-1 (add-1 zero)) plus1) '(NIL))
'(() () NIL)
Hopefully you can see how the Church numbers could be defined as binary functions as well:
(define zero (lambda (f x) x))
(define (add-1 n)
(lambda (f x) (f (n f x))))
(define plus1 (lambda (x) (cons '() x)))
(zero add1 0) ;=> 0
((add-1 zero) add1 0) ;=> 1
((add-1 (add-1 zero)) add1 0) ;=> 2
(zero plus1 '(NIL)) ;=> '(NIL)
((add-1 zero) plus1 '(NIL)) ;=> '(() NIL)
((add-1 (add-1 zero)) plus1 '(NIL)) ;=> '(() () NIL)
producing the same results as before.