Approximating pi/4 using Wallis Product - SICP exercise 1.31

Question

From SICP

The sum procedure is only the simplest of a vast number of similar abstractions that can be captured as higher-order procedures. Write an analagous procedure called product that returns the product of the values of a function at points over a given range.

...

"Also use product to compute approximations to pi using the formula
pi/4 = 2*4*4*6*6*8*.../3*3*5*5*7*7*...

Here is a procedure called product:

(define (product term a next b)
  (if (> a b)
      1
      (* (term a) (product term (next a) next b))))

I have made two solutions:

Here is the first solution. This one navigates through the series by a period of 2 (E.G. if n is 1, it will evaluate 2*4/3*3). I did this because I wanted to find a pattern from the series and what I got was 2*4 and 4*6 and 6*8, and the denominators 3*3 and 5*5 and 7*7.

(define (pi-product2 n)
    (define (inc x) (+ x 1))
    (define (numer-term x)
      (* (* 2.0 x)
         (+ (* 2.0 x)
            2.0)))
    (define (denom-term x)
      (square (+
               (* 2.0 x)
               1.0)))
    (define (term x)
      (/ (numer-term x) (denom-term x)))
    (product term 1.0 inc n))

Here is my second solution. I just recently thought of how to evaluate them by each term (E.G. if n is 1, it will evaluate 2/3).

  (define (pi-product n)
    (define (inc x) (+ x 1))
    (define (term x)
      (if (even? x)
          (/ (+ x 2.0) (+ x 1.0))
          (/ (+ x 1.0) (+ x 2.0))))
    (product term 1.0 inc n))

• Is my first solution still a correct representation of the formula even if it evaluates two terms on each approximation? Or is it the first solution with the correct representation?
• I would think my first solution, though very ugly, is at least as twice as fast as the second solution. Is that right? Is the performance difference huge?

Whitch of my two solution is better? How can I improve them?

Answer 1

Is my first solution still a correct representation of the formula even if it evaluates two terms on each approximation? Or is it the first solution with the correct representation?

Although both solutions may be mathematically equivalent in theory, they are vastly different in practice. The first uses floating-point constants, which leads to inexact math. The second one uses exact math.

How can I improve them?

In general, you should avoid using recursion directly when there are higher-level abstractions that fit. For example, here are some ways you could restructure your code around streams:

• Generate one stream for terms in the numerator and one stream for terms in the denominator. Then use stream-map to divide successive terms. Take the first n terms of the result, and reduce with +.
• Produce a stream of successive approximations of \$π\$. There are several ways to do this. For example, you could use cons-stream, but it would also be possible (if you define a few auxiliary functions) to tweak the strategy outlined in the preceding point. Then take the nth item of this stream.