modify/replace in LISP - replace

I have an implementation of a function called modify list shown below but it only works for top level lists.
(defun modify-list (old new a-list)
(cond
((null a-list) nil)
((eql (car a-list) old) (cons new (modify-list old new (cdr a-list))))
(T (cons (car a-list)(modify-list old new (cdr a-list))))))
CL-USER 16 : 6 > (modify-list 'a 'x '(p a d g c a))
(P X D G C X) <-- GOOD!
CL-USER 17 : 6 > (modify-list 'a 'x '(p a d (g a) c a))
(P X D (G A) C X) <----NOT GOOD!
can anyone help me make this function work on nested lists?

Why not working at an higher level? It would make the code simpler...
(defun modify (old new x)
(cond
((eq x old) new)
((listp x)
(mapcar (lambda (y) (modify old new y)) x))
(t x)))
Basically instead of assuming x must be a list (actually a tree) you just return new if x is old, recursively map if it's a list or otherwise return x unchanged...
With this approach also (modify 'a 'x 'a) --> X (and that IMO seems right).

Here's an idea:
(defun modify-list (old new a-list)
(cond ((null a-list) nil)
((not (listp (car a-list)))
(if (eql (car a-list) old)
(cons new (modify-list old new (cdr a-list)))
(cons (car a-list) (modify-list old new (cdr a-list)))))
(T (cons (modify-list old new (car a-list))
(modify-list old new (cdr a-list))))))
I don't have access to a LISP interpreter (anyone can verify the above procedure, please?), so you'll have to test it first!

Related

Why does this arithmetic code return lists?

I'm currently working my way through exercise 3.17 of SICP. I know that I'm messing it up and I intend to fix that later. I will not link the exercise, as it is not relevant. This was one of my attempts:
#lang sicp
(define (count-pairs x)
(define encountered '())
(define counter 0)
(define (loop x)
(set! counter (+
(cond
((null? x) 0)
((not (pair? x)) 0)
((null? encountered) (set! encountered (list (car x))) 1)
((eq? (car x) (car encountered)) 0)
(else 1))
counter))
(if (not (pair? x)) counter (begin (loop (car x))
loop (cdr x))))
(loop x))
(count-pairs (list 'a 'b 'c))
(define second (cons 'a 'b))
(define third (cons 'a 'b))
(define first (cons second third))
(set-car! third second)
(count-pairs first)
(define 3rd (cons 'a 'b))
(define 2nd (cons 3rd 3rd))
(define 1st (cons 2nd 2nd))
(count-pairs 1st)
To my shock, this returned:
(b c)
((a . b) . b)
((a . b) a . b)
How is this possible? I know that this code isn't even close to doing as intended, but as far as I can see it should only do arithmetic and therefore return numbers. How is it possible for this code to return list structures?
Get a new IDE. Stack Overflow's syntax highlighting makes this a dead giveaway.
(begin (loop (car x))
loop (cdr x))))

Return a list without the last element

I've just started to learn Racket.
I have this code:
#lang racket
(define l1 '(1 2 3 4))
(car l1)
(cdr l1)
(car l1) returns 1.
(cdr l1) returns '(2 3 4)
Is there a function that returns '(1 2 3)?
I've tried this:
#lang racket
(define l1 '(1 2 3 4))
(map
(lambda (l i)
(if (not (= i (sub1 (length l1)))) l '()))
l1 (range 0 (length l1)))
But, it returns: '(1 2 3 ())
And I have also tried:
#lang racket
(define l1 '(1 2 3 4))
(map
(lambda (l i)
(cond ((not (= i (sub1 (length l1)))) l )))
l1 (range 0 (length l1)))
But, it returns: '(1 2 3 #<void>)
The map function always returns a list the same length as its input. You want an output list that is shorter than its input. The function you are looking for is traditionally called but-last:
(define (but-last xs) (reverse (cdr (reverse xs))))
What about something like this ?
(define (myCdr l)
(if (not (pair? (cdr l)))
'()
(cons (car l) (myCdr (cdr l)))
)
)
length is generally an anti-pattern in Scheme because the entire list needs to be read in order to get the result. W. Ness remarks that map does not alter the structure of the list, and the behavior of filter is based on the list's values, neither of which suit your needs.
Instead of making potentially expensive computations first or awkwardly applying the library functions, you can compute the init of a list using direct recursion -
(define (init l)
(cond ((null? l)
(error 'init "cannot get init of empty list"))
((null? (cdr l))
null)
(else
(cons (car l)
(init (cdr l))))))
(init '(a b c d e)) ;; '(a b c d)
(init '(a)) ;; '(a)
(init '()) ;; init: cannot get init of empty list
Or a tail-recursive form that only uses one reverse -
(define (init l)
(let loop ((acc null)
(l l))
(cond ((null? l)
(error 'init "cannot get init of empty list"))
((null? (cdr l))
(reverse acc))
(else
(loop (cons (car l) acc)
(cdr l))))))
(init '(a b c d e)) ;; '(a b c d)
(init '(a)) ;; '(a)
(init '()) ;; init: cannot get init of empty list
And lastly a tail-recursive form that does not use length or reverse. For more intuition on how this works, see "How do collector functions work in Scheme?" -
(define (init l (return identity))
(cond ((null? l)
(error 'init "cannot get init of empty list"))
((null? (cdr l))
(return null))
(else
(init (cdr l)
(lambda (r)
(return (cons (car l) r)))))))
(init '(a b c d e)) ;; '(a b c d)
(init '(a)) ;; '(a)
(init '()) ;; init: cannot get init of empty list
Here's one more, via zipping:
#lang racket
(require srfi/1)
(define (but-last-zip xs)
(if (null xs)
xs ; or error, you choose
(map (lambda (x y) x)
xs
(cdr xs))))
Here's another, emulating filtering via lists with appending, where empty lists disappear by themselves:
(define (but-last-app xs)
(if (null? xs)
xs
(let ((n (length xs)))
(apply append ; the magic
(map (lambda (x i)
(if (= i (- n 1)) '() (list x)))
xs
(range n))))))
Or we could use the decorate--filter--undecorate directly, it's even more code!
(define (but-last-fil xs)
(if (null? xs)
xs
(let ((n (length xs)))
(map car
(filter (lambda (x) (not (null? x)))
(map (lambda (x i)
(if (= i (- n 1)) '() (list x)))
xs
(range n)))))))
Here's yet another alternative, assuming that the list is non-empty. It's efficient (it performs a single pass over the list), and it doesn't get any simpler than this!
(define (delete-last lst)
(drop-right lst 1))
(delete-last '(1 2 3 4))
=> '(1 2 3)
Here is an equivalent of Will Ness's beautiful but-last-zip which does not rely on srfi/1 in Racket: without srfi/1 Racket's map insists that all its arguments are the same length (as does the R5RS version in fact) but it is common in other Lisps to have the function terminate at the end of the shortest list.
This function uses Racket's for/list and also wires in the assumption that the result for the empty list is the empty list.
#lang racket
(define (but-last-zip xs)
(for/list ([x xs] [y (if (null? xs) xs (rest xs))])
x))
I think Will's version is purer: mapping functions over things is a very Lisp thing to do I think, while for/list feels less Lispy to me. This version's only advantage is that it does not require a module.
My own solution using recursion:
#lang racket
(define but-last
(lambda (l)
(cond ((null? (cdr l)) '())
(else (cons (car l) (but-last (cdr l)))))))
And another solution using filter-not and map:
#lang racket
(define l1 '(1 2 3 4))
(filter-not empty? (map
(lambda (l i)
(if (not (= i (sub1 (length l1)))) l empty))
l1 (range 0 (length l1))))

Merging jumping pairs

How do I recursively merge jumping pairs of elements of a list of lists? I need to have
'((a b c) (e d f) (g h i))
from
'((a b) c (e d) f (g h) i)
My attempt
(define (f lst)
(if (or (null? lst)
(null? (cdr lst)))
'()
(cons (append (car lst) (list (cadr lst)))
(list (append (caddr lst) (cdddr lst))))))
works if I define
(define listi '((a b) c (d e) f))
from which I obtain
((a b c) (d e f))
by doing simply
(f listi)
but it does not work for longer lists. I know I need recursion but I don't know where to insert f again in the last sentence of my code.
A simpler case that your algorithm fails: (f '((1 2) 3)) should result in '((1 2 3)), but yours results in an error.
We will define some terms first:
An "element" is a regular element, like 1 or 'a.
A "plain list" is simply a list of "element"s with no nested list.
E.g., '(1 2 3) is a plain list. '((1 2) 3) is not a plain list.
A "plain list" is either:
an empty list
a cons of an "element" and the next "plain list"
A "list of jumping pairs" is a list of even length where the odd index has a "plain list", and the even index has an element. E.g., '((1) 2 (a) 4) is a "list of jumping pairs". A "list of jumping pairs" is either:
an empty list
a cons of
a "plain list"
a cons of an "element" and the next "list of jumping pairs"
We are done with terminology. Before writing the function, let's start with some examples:
(f '()) equivalent to (f empty)
should output '()
equivalent to empty
(f '((1 2) 3)) equivalent to (f (cons (cons 1 (cons 2 empty))
(cons 3
empty)))
should output '((1 2 3))
equivalent to (cons (cons 1 (cons 2 (cons 3 empty)))
empty)
(f '((1 2) 3 (4) a)) equivalent to (f (cons (cons 1 (cons 2 empty))
(cons 3
(cons (cons 4 empty)
(cons 'a
empty)))))
should output '((1 2 3) (4 a))
equivalent to (cons (cons 1 (cons 2 (cons 3 empty)))
(cons (cons 4 (cons 'a empty))
empty))
So, f is a function that consumes a "list of jumping pairs" and returns a list of "plain list".
Now we will write the function f:
(define (f lst)
???)
Note that the type of lst is a "list of jumping pairs", so we will perform a case analysis on it straightforwardly:
(define (f lst)
(cond
[(empty? lst) ???] ;; the empty list case
[else ??? ;; the cons case has
(first lst) ;; the "plain list",
(first (rest lst)) ;; the "element", and
(rest (rest lst)) ;; the next "list of jumping pairs"
???])) ;; that are available for us to use
From the example:
(f '()) equivalent to (f empty)
should output '()
equivalent to empty
we know that the empty case should return an empty list, so let's fill in the hole accordingly:
(define (f lst)
(cond
[(empty? lst) empty] ;; the empty list case
[else ??? ;; the cons case has
(first lst) ;; the "plain list",
(first (rest lst)) ;; the "element", and
(rest (rest lst)) ;; the next "list of jumping pairs"
???])) ;; that are available for us to use
From the example:
(f '((1 2) 3)) equivalent to (f (cons (cons 1 (cons 2 empty))
(cons 3
empty)))
should output '((1 2 3))
equivalent to (cons (cons 1 (cons 2 (cons 3 empty)))
empty)
we know that we definitely want to put the "element" into the back of the "plain list" to obtain the resulting "plain list" that we want:
(define (f lst)
(cond
[(empty? lst) empty] ;; the empty list case
[else ;; the cons case has:
???
;; the resulting "plain list" that we want
(append (first lst) (cons (first (rest lst)) empty))
;; the next "list of jumping pairs"
(rest (rest lst))
;; that are available for us to use
???]))
There's still the next "list of jumping pairs" left that we need to deal with, but we have a way to deal with it already: f!
(define (f lst)
(cond
[(empty? lst) empty] ;; the empty list case
[else ;; the cons case has:
???
;; the resulting "plain list" that we want
(append (first lst) (cons (first (rest lst)) empty))
;; the list of "plain list"
(f (rest (rest lst)))
;; that are available for us to use
???]))
And then we can return the answer:
(define (f lst)
(cond
[(empty? lst) empty] ;; the empty list case
[else ;; the cons case returns
;; the resulting list of "plain list" that we want
(cons (append (first lst) (cons (first (rest lst)) empty))
(f (rest (rest lst))))]))
Pattern matching (using match below) is insanely useful for this kind of problem -
(define (f xs)
(match xs
;; '((a b) c . rest)
[(list (list a b) c rest ...)
(cons (list a b c)
(f rest))]
;; otherwise
[_
empty]))
define/match offers some syntax sugar for this common procedure style making things even nicer -
(define/match (f xs)
[((list (list a b) c rest ...))
(cons (list a b c)
(f rest))]
[(_)
empty])
And a tail-recursive revision -
(define (f xs)
(define/match (loop acc xs)
[(acc (list (list a b) c rest ...))
(loop (cons (list a b c) acc)
rest)]
[(acc _)
acc])
(reverse (loop empty xs)))
Output for each program is the same -
(f '((a b) c (e d) f (g h) i))
;; '((a b c) (e d f) (g h i))
(f '((a b) c))
;; '((a b c))
(f '((a b) c x y z))
;; '((a b c))
(f '(x y z))
;; '()
(f '())
;; '()
As an added bonus, this answer does not use the costly append operation
There is no recursive case in your code so it will just work statically for a 4 element list. You need to support the following:
(f '()) ; ==> ()
(f '((a b c) d (e f g) h)) ; ==> (cons (append '(a b c) (list 'd)) (f '((e f g) h)))
Now this requires exactly even number of elements and that every odd element is a proper list. There is nothing wrong with that, but onw might want to ensure this by type checking or by adding code for what should happen when it isn't.

Appending reversed list in Scheme

I am learning Scheme and wanted to write a recursive program that reverses a given list.
In one test case however, I noticed that a (b c) e -> e (b c) a.
What I'm trying to get is a (b c) e -> e (c b) a.
This is what I have:
(define (deep-reverse lst)
(if (null? lst)
'()
(begin
(display (car lst))
(display "\n")
(if (null? (cdr lst))
'()
(append (deep-reverse (cdr lst)) (list (reverse (car lst))))
) //End of inner if
))) //End of begin, outer if, define
When I attempt to run the code with
(deep-reverse '(1 (b c) (a b)))
I get:
1
(b c)
(a b)
mcdr: contract violation
expected: mpair?
given: 1
The issue is with (list (reverse (car lst))), although in an isolated test case it works fine. Which leads me to believe that the issue may have to do with append.
Thank you in advance.
Edit: Going from (list (reverse (car lst))) to (reverse (list(car lst))) makes the code run without an error but doesn't reverse (a b) to (b a).
As the error message explains, your problem is that you are trying to reverse a number. Firstly, let's remove some of the unnecessary conditions and debugging stuff in your program, arriving at this simpler program. Let's step through this program to see what's going on:
(define (deep-reverse lst)
(if (null? lst)
'()
(append (deep-reverse (cdr lst)) (list (reverse (car lst))))))
We start with
(deep-reverse '(1 (b c) (a b)))
Substituting the argument we get
(if (null? '(1 (b c) (a b)))
'()
(append (deep-reverse (cdr '(1 (b c) (a b))))
(list (reverse (car '(1 (b c) (a b)))))))
Because the condition is #f, this simplifies to
(append (deep-reverse (cdr '(1 (b c) (a b))))
(list (reverse (car '(1 (b c) (a b))))))
To evaluate the first argument, first find the cdr, and call deep-reverse on that. I will skip the steps here but you should easily be able to test that it works correctly.
(append '((b a) (c b)) (list (reverse (car '(1 (b c) (a b))))))
Next we evaluate the car:
(append '((b a) (c b)) (list (reverse 1)))
And here we see what the problem is: we can't reverse a single number!
The issue is that your deep-reverse should have two distinct behaviours recursively:
on a number, or symbol, or other non-list entity, don't do anything, because it does not make sense to reverse a number
on a list, deep reverse it
There are two reasons why your current program does not do this properly:
it only does a shallow reverse on the elements of the list; that is, it won't deep reverse '(((a b) (c d)) ((e f) (g h))) correctly
it fails if it ever encounters a number or other non-list, like a symbol
The easy fix is to add a condition to check if it's a pair? first before attempting to reverse it. If it's not pair?, then lst must either be nil (which we may leave as-is) or a non-list object (which we may also leave as-is)
(define (deep-reverse lst)
(if (pair? lst)
(append (deep-reverse (cdr lst)) (list (deep-reverse (car lst))))
lst))
Finally, I should note that the pattern we are using here is really a foldr pattern. We can abstract away this pattern with foldr:
(define (deep-reverse xs)
(cond ((pair? xs)
(foldr (lambda (x y) (append y (list (deep-reverse x)))) '() xs))
(else xs)))
But we note also that this is inefficient, because append is an expensive operation. Modifying the algorithm to a tail recursive one makes it clear that this is actually a foldl:
(define (deep-reverse xs)
(cond ((pair? xs)
(foldl (lambda (x y) (cons (deep-reverse x) y)) '() xs))
(else xs)))
which is how such a function might be written in typical idiomatic Scheme, or as pointed out by Will Ness,
(define (deep-reverse xs)
(cond ((pair? xs) (reverse (map deep-reverse xs)))
(else xs)))

Scheme extract unique atoms from list

I'm trying to write a scheme function that will return the unique atoms found in the input list such that.
> (unique-atoms '(a (b) b ((c)) (a (b))))
(a c b)
> (unique-atoms '(a . a))
(a)
> (unique-atoms '())
()
I was thinking something like this as a start
(define (unique-atoms l)
(if (null? l)
'()
(eq? (car (l) unique-atoms(cdr (l))))))
but I don't know how to collect the atoms that are unique, and create a new list while checking everything recursively.
The following walks list, term by term. If the next value is a list itself, then a recursive call is made with (append next rest) - that is, as list is walked we are flattening sublists at the same time.
We use a (tail) recursive function, looking, to walk the list and to accumulate the rslt. We add to the result when next is not alreay in rslt.
(define (uniquely list)
(let looking ((rslt '()) (list list))
(if (null? list)
rslt
(let ((next (car list))
(rest (cdr list)))
(if (list? next)
(looking rslt (append next rest))
(looking (if (memq next rslt)
rslt
(cons next rslt))
rest))))))
> (uniquely '(a b (a b) ((((a))))))
(b a)
If you really want the code to work for 'improper lists' like '(a . a) then the predicates null? and list? probably need to change.
This problem has two parts:
You need to find a way to visit each element of the given form, recursing into sublists.
You need a way to collect the unique elements being visited.
Here's a solution to the first part:
(define (recursive-fold visitor initial x)
(let recur ((value initial)
(x x))
(cond ((null? x) value)
((pair? x) (recur (recur value (car x)) (cdr x)))
(else (visitor x value)))))
I leave it for you to implement the second part.
I found a half solution where the non unique items are removed, although this wont work for an atom b and a list with b such as '(b (b))
(define (uniqueAtoms l)
(cond ((null? l)
'())
((member (car l) (cdr l))
(uniqueAtoms (cdr l)))
(else
(cons (car l) (uniqueAtoms (cdr l))))))
The easiest way to solve this problem with all kinds of list structures is to divide it into two parts
1) flatten then list - this results in a proper list with no sublists
; if you use Racket, you can use the build-in flatten procedure
; otherwise this one should do
(define (flatten expr)
(let loop ((expr expr) (res '()))
(cond
((empty? expr) res)
((pair? expr) (append (flatten (car expr)) (flatten (cdr expr))))
(else (cons expr res)))))
2) find all unique members of this proper list
(define (unique-atoms lst)
(let loop ((lst (flatten lst)) (res '()))
(if (empty? lst)
(reverse res)
(let ((c (car lst)))
(loop (cdr lst) (if (member c res) res (cons c res)))))))
Tests:
; unit test - Racket specific
(module+ test
(require rackunit)
(check-equal? (unique-atoms '(a (b) b ((c)) (a (b)))) '(a b c))
(check-equal? (unique-atoms '(a (b) b ((c . q)) (a (b . d)))) '(a b c q d))
(check-equal? (unique-atoms '(a . a)) '(a))
(check-equal? (unique-atoms '(a b (a b) ((((a)))))) '(a b))
(check-equal? (unique-atoms '()) '()))