Composing Fexprs Preserves Short-circuit Evaluation.


; One of the well known advantages of special operators over
; functions is that special operators apply "short-cuirciting".
;
; For example, operator "and" that evaluates
;
;  (and (= 1 1) (= 2 2) (= 3 4) (= 5 (slow-function 6)))
;
; will never evaluate clause (= 5 (slow-function 6)). Instead,
; after (= 3 4) is recognized to be false, there is no theoretical
; possibility that whole expression evaluates to true, and
; nil is returned without evaluating other clauses.
;
; This is significant difference between operators and functions.
; If "and" was defined as function, all clauses would be evaluated
; first, and then passed to the operator "and."
;
; Only today I came to idea to test whether operator compose I
; defined few days ago, which can be applied on fexprs as well as
; on functions, preserves short-circuiting.
;
; I'll test it by defining two well known logical operators,
; nor and nand.

(set 'nor (compose not or))
(set 'nand (compose not and))

; Let us test whether 'nor' and 'nand' preserve short-circuit
; evaluation.

(nand (println "this should be done ")
      (println "and this should be done")
      (println (setf j nil))
      (println "but this shouldn't be done"))

; this should be done
; and this should be done
; nil
;

(println "---")

(nor (println (= 1 2))
     (println (= 1 3))
     (println "and something different")
     (println "this shouldn't be done"))
     
; ---
; nil
; nil
; and something different

; Everything works. Beautiful, isn't it?

(exit)



Bad weather in Croatia.

4 comments:

  1. Really bad weather :(

    >Beautiful
    compose() is a good enough approach to make code shorter, I like it :)

    ReplyDelete
  2. Actually, it is not that bad weather if one sits in the bar, drink tea and rum, and watch the world through windows. I have another similar function, it is "increase-order". On that way I can increase order of functions, so these can apply on other functions. For example, I can define +f

    (setf +f (increase-order +))

    and +f will add functions, not numbers.

    ((+f sin cos) x) ==> (+ (sin x) (cos x))

    It was one of my first posts, so maybe it doesn't work any more - Newlisp changed, but version in my library works.

    ReplyDelete
  3. Blogger Wolverrum said...

    As I understanding increasOrder properly it has type:
    increaseOrder :: (b->c) -> ((a->b) -> (a->c))

    At least above I was able to create using C#2 :)

    FuncX<FuncX<A, B>, FuncX<A, C>> Up< B, C>(FuncX<B, C> fn)
    {return delegate(FuncX<A, B> h) {
    return Compose(h, fn); }; }

    ReplyDelete