Implementing last-non-zero without continuations - recursion

last-non-zero takes a list of numbers and return the last cdr whose car is 0.
So, I can implement it using continuations, but how do I do this with natural recursion.
(define last-non-zero
(lambda (ls)
(let/cc return
(letrec
((lnz
(lambda (ls)
(cond
((null? ls) '())
((zero? (car ls)) ;; jump out when we get to last 0.
(return (lnz (cdr ls))))
(else
(cons (car ls) (lnz (cdr ls))))))))
(lnz ls)))))

Here's an obvious version which is not tail-recursive:
(define (last-non-zero l)
;; Return the last cdr of l which does not contain zero
;; or #f if there is none
(cond
((null? l)
#f)
((zero? (car l))
(let ((lnzc (last-non-zero (cdr l))))
;; This is (or lnzc (cdr l)) but that makes me feel bad
(if lnzc
lnzc
(cdr l))))
(else
(last-non-zero (cdr l)))))
Here is that version turned into a tail-recursive equivalent with also the zero test moved around a bit.
(define (last-non-zero l)
(let lnzl ([lt l]
[r #f])
(if (null? lt)
r
(lnzl (cdr lt) (if (zero? (car lt)) (cdr lt) r)))))
It's much clearer in this last version that the list is traversed exactly once.

Please indicate if I have correctly understood the problem:
#lang scheme
; returns cdr after last zero in lst
(define (last-non-zero lst)
; a helper function with 'saved' holding progress
(define (lnz-iter lst saved)
(if (null? lst)
saved
(if (zero? (car lst))
(lnz-iter (cdr lst) (cdr lst))
(lnz-iter (cdr lst) saved))))
(lnz-iter lst '()))
(last-non-zero '(1 2 3 0 7 9)) ; result (7 9)

Racket's takef-right can do it:
> (takef-right '(1 2 0 3 4 0 5 6 7) (lambda (n) (not (zero? n))))
'(5 6 7)
But assuming you have an assignment where you're supposed to write the logic yourself instead of just using a built in function, one easy if not very efficient approach is to reverse the list, build a new list out of everything up to the first zero, and return that. Something like:
(define (last-non-zero ls)
(let loop ([res '()]
[ls (reverse ls)])
(if (or (null? ls) (zero? (car ls)))
res
(loop (cons (car ls) res) (cdr ls)))))

Using your implementation where you return the argument in the event there are no zero you can just have a variable to keep the value you think has no zero values until you hit it and then update both:
(define (last-non-zero lst)
(let loop ((lst lst) (result lst))
(cond ((null? lst) result)
((zero? (car lst)) (loop (cdr lst) (cdr lst)))
(else (loop (cdr lst) result)))))
(last-non-zero '()) ; ==> ()
(last-non-zero '(2 3)) ; ==> (2 3)
(last-non-zero '(2 3 0)) ; ==> ()
(last-non-zero '(2 3 0 1 2)) ; ==> (1 2)

(define last-non-zero
(lambda (l)
((lambda (s) (s s l (lambda (x) x)))
(lambda (s l* ret)
(if (null? l*)
(ret '())
(let ((a (car l*))
(r (cdr l*)))
(if (zero? a)
(s s r (lambda (x) x))
(s s r
(lambda (r)
(ret (cons a r)))))))))))

Also possible, to use foldr:
(define (last-non-zero l)
(reverse (foldl (lambda (e res) (if (zero? e) '() (cons e res))) 0 l)))
Or use recursion:
(define (last-non-zero l (res '()))
(cond ((empty? l) res)
((zero? (car l)) (last-non-zero (cdr l) (cdr l)))
(else (last-non-zero (cdr l) res))))

Related

How to remove a list of length 1 from a nested list in lisp?

I have a nested list (1 (4 (5) 3) 9 10) and I want to delete the lists of length 1 so the result would be (1 (4 3) 9 10).
This is what I have tried so far, which does not remove (5) and returns the original list.
(defun remove (l)
(cond
((null l) nil)
((and (listp (car l)) (= (length l) 1)) (remove (cdr l)))
((atom (car l)) (cons (car l) (remove (cdr l))))
(T (cons (remove (car l)) (remove (cdr l))))
))
Two things: first, remove is a predefined function in package CL, so I strongly advice to use a different name, let's say my-remove.
Second, you are testing the length of l instead of the sublist (car l), which is what you want to eliminate.
The correct form would be:
(defun my-remove (l)
(cond
((null l) nil)
((and (listp (car l)) (= (length (car l)) 1)) (my-remove (cdr l)))
((atom (car l)) (cons (car l) (my-remove (cdr l))))
(T (cons (my-remove (car l)) (my-remove (cdr l))))
))
Tail call recursive version. Plus: Without the test (atom (car l)) to be permissive for non-list and non-atom components in the list. (e.g. vectors or other objects as element of the list - they are treated like atoms.
(defun my-remove (l &optional (acc '()))
(cond ((null l) (nreverse acc))
((listp (car l)) (if (= 1 (length (car l))) ;; list objects
(my-remove (cdr l) acc) ;; - of length 1
(my-remove (cdr l) (cons (my-remove (car l)) acc)))) ;; - longer
(t (my-remove (cdr l) (cons (car l) acc))))) ;; non-list objects

Option type encoding / robustness in Lisp

(define (nth n lst)
(if (= n 1)
(car lst)
(nth (- n 1)
(cdr lst) )))
is an unsafe partial function, n may go out of range. An error can be helpful,
(define (nth n lst)
(if (null? lst)
(error "`nth` out of range")
(if (= n 1)
(car lst)
(nth (- n 1)
(cdr lst) ))))
But what would a robust Scheme analogue to Haskell's Maybe data type look like?
data Maybe a = Nothing | Just a
nth :: Int -> [a] -> Maybe a
nth _ [] = Nothing
nth 1 (x : _) = Just x
nth n (_ : xs) = nth (n - 1) xs
Is just returning '() adequate?
(define (nth n lst)
(if (null? lst) '()
(if (= n 1)
(car lst)
(nth (- n 1)
(cdr lst) ))))
It's easy to break your attempt. Just create a list that contains an empty list:
(define lst '((1 2) () (3 4)))
(nth 2 lst)
-> ()
(nth 100 lst)
-> ()
The key point that you're missing is that Haskell's Maybe doesn't simply return a bare value when it exists, it wraps that value. As you said, Haskell defines Maybe like so:
data Maybe a = Nothing | Just a
NOT like this:
data Maybe a = Nothing | a
The latter is the equivalent of what you're doing.
To get most of the way to a proper Maybe, you can return an empty list if the element does not exist, as you were, but also wrap the return value in another list if the element does exist:
(define (nth n lst)
(if (null? lst) '()
(if (= n 1)
(list (car lst)) ; This is the element, wrap it before returning.
(nth (- n 1)
(cdr lst) ))))
This way, your result will be either an empty list, meaning the element did not exist, or a list with only one element: the element you asked for. Reusing that same list from above, we can distinguish between the empty list and a non-existant element:
(define lst '((1 2) () (3 4)))
(nth 2 lst)
-> (())
(nth 100 lst)
-> ()
Another way to signal, that no matching element was found, would be to use multiple return values:
(define (nth n ls)
(cond
((null? ls) (values #f #f))
((= n 1) (values (car ls) #t))
(else (nth (- n 1) ls))))
This comes at the expense of being a little bit cumbersome for the users of this function, since they now have to do a
(call-with-values (lambda () (nth some-n some-list))
(lambda (element found?)
... whatever ...))
but that can be alleviated by using some careful macrology. R7RS specifies the let-values syntax.
(let-values (((element found?) (nth some-n some-list)))
... whatever ...)
There are several ways to do this.
The direct equivalent would be to mimic the Miranda version:
#!r6rs
(library (sylwester maybe)
(export maybe nothing maybe? nothing?)
(import (rnrs base))
;; private tag
(define tag-maybe (list 'maybe))
;; exported tag and features
(define nothing (list 'nothing))
(define (maybe? v)
(and (pair? v)
(eq? tag-maybe (car v))))
(define (nothing? v)
(and (maybe? v)
(eq? nothing (cdr v))))
(define (maybe v)
(cons tag-maybe v)))
How to use it:
#!r6rs
(import (rnrs) (sylwester maybe))
(define (nth n lst)
(cond ((null? lst) (maybe nothing))
((zero? n) (maybe (car lst)))
(else (nth (- n 1) (cdr lst)))))
(nothing? (nth 2 '()))
; ==> #t
Exceptions
(define (nth n lst)
(cond ((null? lst) (raise 'nth-nothing))
((zero? n) (car lst))
(else (nth (- n 1) (cdr lst)))))
(guard (ex
((eq? ex 'nth-nothing)
"nothing-value"))
(nth 1 '())) ; ==> "nothing-value"
Default value:
(define (nth n lst nothing)
(cond ((null? lst) nothing)
((zero? n) (car lst))
(else (nth (- n 1) (cdr lst)))))
(nth 1 '() '())
; ==> '()
Deault value derived from procedure
(define (nth index lst pnothing)
(cond ((null? lst) (pnothing))
((zero? n) (car lst))
(else (nth (- n 1) (cdr lst)))))
(nth 1 '() (lambda _ "list too short"))
; ==> "list too short"
Combination of exception and default procedure
Racket, a Scheme decent, often has a default value option that defaults to an exception or a procedure thunk. It's possible to mimic that behavior:
(define (handle signal rest)
(if (and (not (null? rest))
(procedure? (car rest)))
((car rest))
(raise signal)))
(define (nth n lst . nothing)
(cond ((null? lst) (handle 'nth-nothing nothing))
((zero? n) (car lst))
(else (nth (- n 1) (cdr lst)))))
(nth 1 '() (lambda () 5)) ; ==> 5
(nth 1 '()) ; exception signalled
As a non-lisper I really can't say how idiomatic this is, but you could return the Church encoding of an option type:
(define (nth n ls)
(cond
((null? ls) (lambda (default f) default))
((= n 1) (lambda (default f) (f (car ls))))
(else (nth (- n 1) ls))))
But that's about as complicated to use as #Dirk's proposal. I'd personally prefer to just add a default argument to nth itself.

Leave only the elements that are not repeated in a given list IN SCHEME [duplicate]

I'm trying to teach myself functional language thinking and have written a procedure that takes a list and returns a list with duplicates filtered out. This works, but the output list is sorted in the order in which the last instance of each duplicate item is found in the input list.
(define (inlist L n)
(cond
((null? L) #f)
((= (car L) n) #t)
(else (inlist (cdr L) n))
))
(define (uniquelist L)
(cond
((null? L) '())
((= 1 (length L)) L)
((inlist (cdr L) (car L)) (uniquelist (cdr L)))
(else (cons (car L) (uniquelist (cdr L))))
))
So..
(uniquelist '(1 1 2 3)) => (1 2 3)
...but...
(uniquelist '(1 2 3 1)) => (2 3 1)
Is there a simple alternative that maintains the order of the first instance of each duplicate?
The best way to solve this problem would be to use Racket's built-in remove-duplicates procedure. But of course, you want to implement the solution from scratch. Here's a way using idiomatic Racket, and notice that we can use member (another built-in function) in place of inlist:
(define (uniquelist L)
(let loop ([lst (reverse L)] [acc empty])
(cond [(empty? lst)
acc]
[(member (first lst) (rest lst))
(loop (rest lst) acc)]
[else
(loop (rest lst) (cons (first lst) acc))])))
Or we can write the same procedure using standard Scheme, as shown in SICP:
(define (uniquelist L)
(let loop ((lst (reverse L)) (acc '()))
(cond ((null? lst)
acc)
((member (car lst) (cdr lst))
(loop (cdr lst) acc))
(else
(loop (cdr lst) (cons (car lst) acc))))))
The above makes use of a named let for iteration, and shows how to write a tail-recursive implementation. It works as expected:
(uniquelist '(1 1 2 3))
=> '(1 2 3)
(uniquelist '(1 2 3 1))
=> '(1 2 3)

Vector multiply function in scheme, manipulating initial recursive parameter

I've been working on a vector multiply function in scheme and have found myself in rut. I dont want to use any looping and I dont want to use any scheme built in functions other than the ones I've already included. I've created a helper function called rotate and dotproduct. I can get the correct values if I do this in racket (vectormult '(1 2 -1) (rotate '((0 2 3) (1 2 0) (1 0 3)))). How can I rotate the initial parameter without re-rotating after every recursive call? NOTE: I dont want to introduce additional paramaters. If my logic/approach to this is all wrong please help me get on the right track.
Code
(define dotproduct
(lambda (l1 l2)
(if (or (null? l1) (null? l2))
0
(+ (* (car l1) (car l2)) (dotproduct (cdr l1) (cdr l2))))))
(define getFirsts
(lambda (l)
(cond
((null? l) `())
(else (cons (first* l) (getFirsts (cdr l)))))))
(define removeFirsts
(lambda (l)
(cond
((null? l) `())
((null? (car l)) `())
(else (cons (cdr (car l)) (removeFirsts (cdr l)))))))
(define rotate
(lambda (l)
(cond
((null? l) `())
((null? (first* l)) `())
(else (cons (getFirsts l) (rotate (removeFirsts l)))))))
(define vectormult
(lambda (l1 l2)
(cond
((null? l2) `())
(else (cons (dotproduct l1 (car l2)) (vectormult l1 (cdr l2)))))))
If I understand the question, you could rename your current vectormult to, say, rotatedvectormult (and change its recursive call accordingly), and then have vectormult just rotate the parameter before calling rotatedvectormult. This way, rotatedvectormult would know the parameter was already rotated, but vectormult could still take an unrotated vector.
I ended up ditching the rotate function in favor of adding 2 functions getFirsts and removeFirsts.
Code
(define getFirsts
(lambda (l)
(cond
((null? l) `())
(else (cons (first* l) (getFirsts (cdr l)))))))
(define removeFirsts
(lambda (l)
(cond
((null? l) `())
((null? (car l)) `())
(else (cons (cdr (car l)) (removeFirsts (cdr l)))))))
;(define rotate
; (lambda (l)
; (cond
; ((null? l) `())
; ((null? (first* l)) `())
; (else (cons (getFirsts l) (rotate (removeFirsts l)))))))
(define vectormult
(lambda (l1 l2)
(cond
((null? (first* l2)) `())
(else (cons (dotproduct l1 (getFirsts l2)) (vectormult l1 (removeFirsts l2)))))))

Scheme reverse a list

I am trying to reverse a list in Scheme using DrRacket.
Code:
(define rev
(lambda(l)
(if (null? l)
'()
(append (rev (cdr l)) (list (car l))))))
If I input (rev '(a((b)(c d)(((e)))))), the output is (((b) (c d) (((e)))) a).
I want it to be (((((e)))(d c)(b))a). I looked here: How to Reverse a List? but I get an even worse output. What am I doing wrong? Any help would be appreciated!
This is trickier than it looks, you're trying to do a "deep reverse" on a list of lists, not only the elements are reversed, but also the structure … here, try this:
(define (rev l)
(let loop ((lst l)
(acc '()))
(cond ((null? lst) acc)
((not (pair? lst)) lst)
(else (loop (cdr lst)
(cons (rev (car lst))
acc))))))
It works as expected:
(rev '(a ((b) (c d) (((e))))))
=> '(((((e))) (d c) (b)) a)
This code will do it:
(define (rev-list lst)
(if (null? lst)
null
(if (list? lst)
(append (rev-list (cdr lst)
(list (rev-list (car lst))))
lst)))
And the result is:
>>> (display (rev-list '((1 7) 5 (2 4 (5 9))) ))
(((9 5) 4 2) 5 (7 1))
The idea is simple: Return the arg if it's not a list, return rev-list(arg) otherwise.

Resources