Ask HN: What are other 'Fizzbuzz' type interview questions? | Hacker News Reader

Ask HN: What are other 'Fizzbuzz' type interview questions?

15 pointsjabs8322 comments

As a tangent from this article: http://www.gayle.com/blog/2015/5/31/the-problem-with-the-fizzbuzz-problem

What are some other similar type of interview questions that employers might ask to test an applicant's coding competency? Regardless of the oppinion of article's author I do like the 'fizzbuzz' problem as one that takes 'just enough coding knowledge/thinking'. Here's another example from the article: 'Given a list of people with their birth and end years (all between 1900 and 2000), find the year with the most number of people alive'

Please list other similar type questions. Thanks!

View on HN

There are lots of such questions:

1. Reverse a String

2. Find min/max in an array

3. Find/remove duplicates

4. IsPalindrome?

5. Fibonacci (iterative especially)

6. Primes (IsPrime, print first few etc)

7. Calculate Factorial

8. Find one missing number from 1 to N.

9. Basic Run length encoding (values and their counts)

etc.

It's trivial to come up with such questions, but it's not trivial to deliver them and judge a person on it. Before asking any of such questions, I'd urge you to exactly know what you want that person to do and have a sense of where they are coming from.

(I do this stuff for a living. About me: http://InterviewKickstart.com)

Plus some of these require background knowledge, for example some candidates won't know what a Fibonacci number is or how to calculate one (since they literally haven't covered it for twenty+ years).

So be careful what you're ACTUALLY testing.

The point of FizzBuzz is that you can give candidates a description of the problem and also give them the mod operator and a sample of its usage and some people would still fail.

It would be great if there were some actual research on this. (How many people who claim to be able to program cannot create fizzbuzz in an hour even if you give them a description of modulo)

I suspect that a lot of it comes from "programming sheep and goats", which was later withdrawn by its author.

You don't actually need the modulo operator though.

Modulo is effectively x - x * (x/y) with integers but you don't even need that.

You can do something like this.

#include <stdio.h> #include <stdbool.h>

int main(void) {

int five = 5; int three = 3; int count; bool numprint;

for (count = 1; count <= 100; count++){ numprint = true; if (count - three == 0) { three += 3; printf("Fizz"); numprint = false; } if (count - five == 0) { five += 5; printf("Buzz"); numprint = false; } if (numprint) { printf("%d", count); } printf("\n"); }

return 0; }

Yep, point is not to test if they know Fibonacci or prime numbers. If the candidate has forgotten, then you just tell them what those things are. And then expect them to answer it.

But here is the (sad?) reality: There will always be someone else applying for the same job, who knows what fibonacci numbers are. You have to be careful to give equal opportunity to both those kinds of people.

>8. Find one missing number from 1 to N.

That's a horrible interview question, because it's easy if you know the arithmetic/xor trick. If you haven't seen it before, it's harder than what you'd expect a candidate to get on an interview.

The arithmetic here is high-school arithmetic. Even if one doesn't know it, it is easy to derive/see it during the interview.

Not saying it's an excellent question. Just saying it's pretty much as quirky/inefficient as any other technical interview question.

Most interview questions can be summarized as:

1. Have you heard this question before?

2. What number am I thinking of?

You'd be surprised how often (2) comes up. The interviewer is looking for a specific answer, and any other answer is wrong.

You don't need to use xor, unless you are trying to be clever or have a number representation already in bits: n(n+1)/2 - sum(array) is your answer and is O(n);

xor is the solution that uses the least storage

I heard a variation of that question that began "using the least possible storage space...", with the goal being to find the xor solution.

The arithmetic solution also has problems when you overflow an integer, although that matters more if your CPU has 16 bit ints.

I'm probably going to get lynched for this. I don't have a CS background, but I don't see the difficulty in reversing a string?

>string[::-1]

Maybe it's because of my limited experience with (so far) Python, and not a lot of exposure to C/C++. What is the answer that interviewers are looking for here?

I think the interviewer would be looking for you to show how you might implement string[::-1] as an algorithm - rather than how to invoke the existing string reversal algorithm in cPython.

That's the problem with these questions, the expected answers are steeped in poor engineering practice. When a string needs reversing a library call is always the best first answer. Always. Always. Always.

Because the library call has already been debugged.

Hypothetical response:

"Of course, brudgers, and with your experience I'd hope that you would understand that we don't expect you to replace the $LANGUAGE standard library on this job. However, it's still important to us to know that you have some idea how to do that, even if only as a proxy for the sort of work we do expect you to do on this job. So, mind humoring us and taking a stab at it?"

1. When the absolutely best correct technical answer is unacceptable then how can the question measure technical competence accurately?

2. When the acceptable answers encourage suboptimal engineering practice what in an empirical and non-magical world is the question likely to select for?

3. When an organization is committed to an empirical approach, what is there besides:

  :does it pass the test?
  :does it need refactoring?

4. Once one notices how quickly the hypothetical response elided into measuring the willingness of the candidate to produce objectively worse code upon instruction, should you be surprised if the question selects for people?

The formal rejecting the correct and best answer is a tribal more steeped in "real men write their own string reversal algorithms" culture. For a junior engineer it sends the message that writing your own string reversal algorithm is a good thing. For a senior engineer, absent the sort of conversation we are having -- and isn't this the sort of conversation one would want to have with a senior engineer? -- the question selects for something other than sound engineering judgement.

It's not that the question is bad, per se. It's that when the best objectively correct answer is unacceptable, the belief that it objectively measures competence is a rationalized superstition because it mostly measures, "is this someone who I want to drink beer with?"

One I've had asked of me before is reversing a string, and then I was asked to re-implement it so that the string is reversed in place. I like this one over fizzbuzz because it doesn't require the candidate to know about the modulus operator. It lets you see how the candidate reasons around memory and edge cases, which come up far more frequently than modulus.

A problem with this is that it's asked so frequently people just memorize the answer.

This seems to imply that you can't learn about a candidate without giving them a problem them had not previously memorized. I have most of the solutions to common problems memorized in a few languages and can complete them on a board or a txt doc. Does that make me less of a good candidate? Or more?

I wouldn't say you can't learn about them but it makes it more difficult. Some people are really good at memorizing and when you ask them a common problem it's not useful -- they will just regurgitate the answer and reasoning without thinking about it. Some people are bad at memorizing and asking them the problem can be useful, since they have to work through it again, and fundamentally you want someone who can take a problem and work their way through it.

Really you should just create simple, unique, problems and ask those problems to candidates. It's not as easy as just asking the question you were asked, that you know the answer to, but it will force people to actually show if they're capable of solving problems and explaining their thought process.

>I have most of the solutions to common problems memorized in a few languages and can complete them on a board or a txt doc. Does that make me less of a good candidate? Or more?

I would say this probably makes you good at interviewing but has no impact on whether I think you would be a good candidate.

I don't know why people seem to think that you need to use a modulus operator to do fizzbuzz. There are several other ways to tell if one integer is divisible by another.