Three Interview Questions
May 13, 2014
The first question (insert new node 40 in sorted linked list, given node 50) is a trick question. No, it is not possible to insert the node. You need a pointer to node 30 to perform the insertion, but there is no way to get a pointer to node 30 given only a pointer to node 50. This is a simple question to ensure the candidate understands linked lists.
The second question (subtract a linked list end to end) is tedious, but not hard. We use the tortoise-and-hare algorithm to find the middle:
(define (subtract-first-half xs)
(let loop ((front xs) (back (reverse xs)) (hare xs) (out (list)))
(if (or (null? hare) (null? (cdr hare)))
(append (reverse out) front)
(loop (cdr front) (cdr back) (cddr front)
(cons (- (car front) (car back)) out)))))
> (subtract-first-half '(5 4 3 2 1))
(4 2 0 2 1)
> (subtract-first-half '(4 3 2 1))
(3 1 2 1)
Another good question; the tortoise-and-hare algorithm should be familiar to all competent programmers.
The third question (number of trailing zeros in a factorial) is easy if you know the math and impossible if you don’t. We write numbers base 10, so there are as many trailing zeros as there are factors of 10 in the factorial of n, and since 10 = 2 * 5 and there are more factors of 2 than of 5 in any factorial, the number of trailing zeros will be the number of factors of 5. To find that, for each power of 5 up n add the number of times it divides n. For instance, if n = 487, there are ⌊487/5⌋ = 97 factors of 5, ⌊487/25⌋ = 19 factors of 52, and ⌊487/125⌋ = 3 factors of 53.
(define (trailing-zeros n)
(let loop ((k 5) (z 0))
(if (< n k) z
(loop (* k 5) (+ z (quotient n k))))))
> (trailing-zeros 487)
119
I don’t like this interview question. If you know the math, or if you have studied a book of programming interview questions, it’s easy; you can write the program as fast as you can type. If you don’t know the math, you will flail about looking for a solution, getting more and more nervous as time passes. In either case, the interviewer will learn nothing about your programming skills.
You can run the program at http://programmingpraxis.codepad.org/G17lhMD3.
Programming Praxis 
The first can be a double trick question… it depends on if the nodes are modifiable. If so, you change the value in 50 to 40, and insert a new 40 node.
I mean, a new 50 node.
Ceci n’est pas une Python.
def insert(links, elem): '''Insert in ascending order.''' return ( (elem, ()) if links == () else (elem, links) if elem < links[0] else (links[0], insert(links[1], elem)) ) print('Test result 1:', insert((10, (30, (50, (70, ())))), 40))Traditional Lisp exercises. Much fun.
The above and the below produce the expected results in this form:
Test result 1: (10, (30, (40, (50, (70, ())))))
Test result 2: (4, (2, (0, (2, (1, ())))))
def reverse(links, aux = ()): return ( aux if links == () else reverse(links[1], (links[0], aux)) ) def tail(links, meter): '''Return the last half-meter of links.''' return ( links if meter == () or meter[1] == () else tail(links[1], meter[1][1]) ) def sub(longer, shorter): return ( longer if shorter == () else (longer[0] - shorter[0], sub(longer[1], shorter[1])) ) def subrevtail(links): return sub(links, reverse(tail(links, links))) print('Test result 2:', subrevtail((5, (4, (3, (2, (1, ())))))))Oops. Yes, I did miss an initial condition in the first question.
;; 1) Consider a sorted singly-linked list having the following ;; nodes: 10 -> 30 -> 50 -> 70 -> NULL. You are given a pointer to ;; node 50 and a new node having the value 40. Can you insert node ;; 40 correctly in the list maintaining the ascending order? (defun insert-at-node (node data) (setf (cdr node) (cons (car node) (cdr node)) (car node) data) node) (let ((list (list 10 30 50 70))) (insert-at-node (nthcdr 2 list) 40) list) --> (10 30 40 50 70)This is correct, as long as the list is is originally in the ascending
order, and the node given to insert-at-node is effectively the node
containing the smallest item greater than the new data.
;; 2) Given a linked list 5 -> 4 -> 3 -> 2 -> 1 produce a linked ;; list 4 -> 2 -> 0 -> 2 -> 1 by subtracting the last node of the ;; list from the first, the next-to-last node from the second, and ;; so on, stopping at the midpoint of the list. (defun palinsub (list) (loop :for slow = list :then (cdr slow) :for fast = list :then (cddr fast) :while (cdr fast) :finally (return (append (mapcar (function -) list (reverse slow)) (if fast (cdr slow) slow))))) (assert (equalp (palinsub '(5 4 3 2 1)) '(4 2 0 2 1))) (assert (equalp (palinsub '(5 4 3 2)) '(3 1 3 2))) (assert (equalp (palinsub '(3)) '(0))) (assert (equalp (palinsub '()) '()));; 3) Write a program to output the number of consecutive trailing ;; zeros in the factorial of a number. For example, if the number is ;; 5, then 5! = 120, and there is one trailing zero. (defun count-factors-in-n (n f) (if (zerop n) 0 (loop :with quo :with rem :do (multiple-value-setq (quo rem) (truncate n f)) (setq n quo) :while (zerop rem) :sum 1))) (defun count-factors-in-factorial-n (n f) (loop :for i :from 1 :to n :sum (count-factors-in-n i f))) (progn (assert (= (count-factors-in-factorial-n 4 5) 0)) (assert (= (count-factors-in-factorial-n 5 5) 1)) (assert (= (count-factors-in-factorial-n 6 5) 1)) (assert (= (count-factors-in-factorial-n 24 5) 4)) (assert (= (count-factors-in-factorial-n 25 5) 6)) (assert (= (count-factors-in-factorial-n 27 5) 6)))For any n>=2, in the decomposition in prime factors of n!, the
exponent of 2 is greater than the exponent of 5,
since (log n 2)/(log n 5) = (log 5 2) ~= 2.321928 > 1
Since only the multiples of an exponent of 10 will add a 0 to the
factorial, and there are more multiples of 2 than multiples of 5, we
can just sum the exponents of 5 factor by factor.
For the last problem, we can just repeatedly divide by 5 and add:
fcount 0 k = k fcount n k = fcount n' (k+n') where n' = n `div` 5 main = print (fcount 487 0) >> print (fcount 100 0) >> print (fcount 1000 0) >> print (fcount 10000 0) >> print (fcount 100000 0) 119 24 249 2499 24999And for the second problem, a little Haskell function:
@informatimago –
Your factorial algorithm was interesting – building on top of it- if we just divide n by 5 – that seems to work although I did not test for very large numbers…
factorial(4) – will have 0 zeros (4/5)
factorial(10) – will have 2 zeros (10/5)
factorial(24) – will have 4 zeros (24/5)
factorial(25) – will have 5 zeros (25/5) and so on…
For number 3, counting factors in each iteration of a factorial is asymptotically less efficient than counting representation of progressively larger powers of the factor in n. Implemented in Python the faster solution looks as follows:
def trailing_zeroes(n): zs = 0 while n >= 5: n //= 5 zs += n return zsEven at small n, the difference in growth is significant:
(n was randomly chosen across range [1,10240], 1000 samples for each n were taken and averaged.)
n: Faster (s) Slower (s)
538: 3.74e-07 8.91e-05
1418: 4.32e-07 2.34e-04
3915: 4.85e-07 6.43e-04
4071: 4.86e-07 6.72e-04
6333: 4.85e-07 1.04e-03
6364: 4.85e-07 1.05e-03
8262: 5.03e-07 1.36e-03
8970: 5.00e-07 1.48e-03
9401: 4.99e-07 1.55e-03
9497: 5.01e-07 1.57e-03
After n=10240, the slower algorithm becomes unwieldy, while the faster method still runs in under 4e-04 seconds with n above 480 digits.
Vivek is correct. For the second problem, all we have to do is divide by 5. The reasoning is that the prime factors of 10 are 2 and 5. For n!, every 5th value of n will give us a factor of 5. Every second value of n will give us a factor of 2, so we’ll have plenty of factors of 2 (i.e. in the prime factorization of n!, we’ll always have at least as many factors of 2 as we do factors of 5). So every 5th value of n!, we’ll add another 0 to the end of the number.