When have you ever needed to know the exact complexity function for bubble sort (beyond the simple fact that it's in a class well above that of any algorithm you'd typically want to use) in order to solve a real, actual work-related problem? Right there, on the spot?
That's the point of the "hate" behind these questions.
Point one, it checks if you understand the most primitive sorting algorithm out there. Pretty low bar assessment of your general CS knowledge.
Point two, it checks whether the candidate has understanding of computational complexity.
Now you might argue that you don't need any of that in the day job, but that's on you. If the interviewer wants to check you have fairly basic minimum of understanding of very basic CS concepts, that's a suitable question to ask.
In theory there is a difference, in practice - none. People would at least know what that is (and hence have no problem gauging bubblesort complexity), or you'd have blank stare back.
Just wonder what would you suggest as an alternative, if you need to confirm basic algorithms proficiency and understanding of big-O? You inevitably arrive to an algorithm question, and bubblesort is as good as any.
Actually the blank stare means "What is this, sophomore year again? I can't believe anyone still cares about bubble sort."
Just wonder what would you suggest as an alternative?
Look at their GitHub/Mercurial account (which you've been ignoring all this time in your desperate search for something to nail them on, but if you would spend a second or two looking, you'd find is chuck full of algorithm stuff -- much of it way more intricate than bubble sort). And ask as many questions as you like based on some project you find there.
Or, pick a problem you're working on that's algorithm-related (but which you genuinely don't fully know how to solve). Use that as discussion material. What approach they'd suggest, given that the data are sparse / not evenly distributed, whatever.
You know, as if they were a peer. Not an interrogation subject.
The majority of my peers don't have a GitHub account, most of them quite competent programmers. Do you suggest to discriminate applicants based on their public activity? Do you suggest it's somehow more fair than a general question?
I like how you cast me into some mean soulless generalized interviewer and started bashing down your pain points, while I actually don't interview people. But a simple algorithm question is not out of line on a programming interview; it's not whiteboarding or take home assignments. It doesn't take much time to answer, it shows you have some very basic CS knowledge at least. I did not use bubblesort outside of CS101 class some 25 years ago and had to actually think about its runtime complexity, but it was not hard and it did not take long.
Again if you think basic CS knowledge has nothing to do with your job that's on you. Should add that I'm not really comfortable being grilled in any way on the interviews, it's not a very good dynamic. But I don't see any general, dignified way to filter out non-performers. Unfortunately just credentials are not enough in our trade.
If I have to do a very careful performance decision on something as well trodden as a sorting algorithm, I'll walk to the bookshelves, pick up The Art of Programming, Volume 3, and let Knuth handle it for me.
As a software engineer, there are things I do all the time: Reading unfamiliar code and third party documentation, do some debugging (often based on logs), and try to write code that is well factored and thoroughly tested. I want to know whether any prospective teammate is good at those things for their experience level.
Interestingly enough, the general algorithm questions are easier the newer you are, and the less you know about all the topics I mentioned, because more often than not those are learned on the jobs, while the basic algorithms are learned in college. I end up reading papers on algorithms at work, but they are not algorithms that I expect to be general knowledge, and that I'd expect people to read on the job when they are applicable: How many candidates are going to be able to explain hyper log log, Paxos or Lamport clocks on command? Is knowing all of those three by heart a sensible pass/no pass signal? I for one don't think so.
I do not think that these questions probe some corner cases. Rather these probe the general knowledge about the field.
Bubble sort question probes the understanding about the complexity theory, hopscotch hashing about concurrent algorithms and CAP theorem about problems with distributed databases.
Naturally when these questions are not open for discussion and instead only a short answer is expected then there is in my opinion something wrong with the interviewer or with the company.
You have to understand that the other side knowns in fact nothing about you and I find these questions to be quite fair to improve the situation.
Bubble sort question probes the understanding about the complexity theory,
It's just that, again, you're picking a very marginal example to do that with.
And more fundamentally -- simply asking someone "What's the complexity of $foo"? doesn't tell you anything about whether they "understand" complexity. It only tells you whether they've adequately memorized that particular cell on their crib sheet.
As they are thoroughly incentivized to do, thanks to people employing interview techniques like these.
I definitely agree with you that shallow quizzes are not very informative and are alone not a good methodology for job interviews.
Naturally this question in isolation is not very informative and it is hard to differentiate if the answer to it comes from a more general understanding or is just memorized. It has to be supported with other questions and wider discussion.
What I want to say is that in a larger context this question is still not unfounded and it is hard for me to consider this or any other listed questions as a signalling a dominant status.
If you do not know the complexity class of the bubble sort then you probably do not a have formal CS education. It also follows that it is possible that when you see a similar bad solution then you are not able to recognize it as such.
I think that these three questions showed are quite good and fair (for example I did not what is Hopscotch hashing, but it looks quite important after looking it up, so thanks for this).
Some of such questions might be possibly indifferent for the position you were applying, but seeing some hate behind these questions is in my opinion quite unjustified.
> If you do not know the complexity class of the bubble sort then you probably do not a have formal CS education. It also follows that it is possible that when you see a similar bad solution then you are not able to recognize it as such.
This is absolutely absurd. I have a CS degree from Berkeley (universally considered one of the best CS programs in the world), and I spent the last few years on an unusually CS-focused team within Google (specifically, artificial intelligence). Off the top of my head, I could easily imagine not remembering what exactly bubble sort is. What's the use of retaining the details of an algorithm that's literally held up as the "don't do this" solution to sorting? The most I can imagine being useful is knowing the theoretical best case average complexity of a sort. And my job is wayyy more CS-focused than the average engineering job at Google (let alone elsewhere). You heavily overestimate how useful rote memorization is to real-world productivity.
Whether a candidate _understands_ these concepts, on the other hand, can actually be illuminating. If you described bubble sort to someone and asked him to describe the complexity, then you're getting all the signals you're describing about CS understanding and education. (though there's controversy about the usefulness of these skills as well)
I've been on both sides of the interview process at Google over the last half decade (I no longer work there) and I can tell you that asking people to recite random facts by rote was not part of their interview guidelines for as long as I was there. This includes a couple dozen interviews I conducted and the candidate notes of the other interviewers.
I got hit with a barrage of random trivia during one of the phone screens I did last year.
Eventually got fed up with their process, told the interviewer what I thought of it and hung up.
(which, to point out at least one upside, is the only way I've ever discovered of getting a company I don't want to work for to stop bugging me with recruiter spam)
I agree with you. But I think that you actually prove that this is not that absurd question at all because you actually described exactly the content of this question and why it is asked about (at least in my opinion).
Naturally if there is no option for the discussion when somebody does not know the answer off the top of their head then this would be really absurd and pointless situation.
This question is not good for selecting out somebody for their specific knowledge, but it is in my option a good one to prune out candidates that manage to demonstrate their general ignorance about the field.
> Naturally if there is no option for the discussion when somebody does not know the answer off the top of their head then this would be really absurd and pointless situation.
Ah ok, I guess I misunderstood you. Though this "absurd and pointless situation" does in fact happen: There definitely is a class of interviewer who will expect off-the-top-of-your-head detailed knowledge (like being able to remember bubble sort's details from just its name).
If you do not know the complexity class of the bubble sort then you probably do not a have formal CS education.
No, it means you're smart and have learned to sift out unimportant matters of detail ("Was bubble sort, which I only ever had to think about for a week or two the first half of my sophomore algorithms class, and was only ever brought up as a negative example anyway, O(n^2) or O(n^3)?") in order to focus on much more interesting and useful stuff.
Pretty much spot on and that was my point. You can answer to this question, you know what is complexity class, you know that it is a negative example and it is rather bad.
See, I am not a recruiter or interviewer, but this question feels to me like an useful one to prune out ignorant fools who start to lament how unfair is to ask such questions instead of giving more or less correct answer.
Naturally if short direct answer is the only accepted answer and there is no way to discuss these things with the interviewer then I would also walk away.
I haven't received a formal education in CS, and I know about CAP and what it means.
Reading the Aphyr "Jepsen" series is what helped me learn it... when I was still in high school. And this was necessary because I was curious about which database I needed to use for a particular side project I was writing.
It's totally possible to just Google "bubble sort complexity" or "CAP" and learn about it.
Yes, you can today easily find quite a lot of information and learn about it, but this was not the situation the original poster lamented about.
The claim was that when faced with such question, one can just google it and read the answer. Problem solved. Because these are just some random facts like who is the president of France.
My point was that in my opinion these questions actually probe the general understanding about the CS field.
You act as though CS knowledge is some high priesthood of supreme mathematical reasoning insulated from the internet.
As a matter of practical use, most of the time details about algorithms and data structures, from implementation through complexity analysis, can easily be regarded as "mere facts." Those times when this doesn't apply are times when somebody applying formal academic level CS is surely required, but these are far less frequent than most programming tasks require--even at Google.
No, this was not my intention. Sure people do not know all the facts and all the implementation details and corner cases of some specific algorithm, but I would expect them to have some basic knowledge.
This is like knowing that lookup complexity of the hash table is usually O(1) (but could become O(n)) and one should use it instead of list when frequent lookups are necessary.
I am not an interviewer, but I think that the questions listed were actually quite good pruning questions and if you do not know them, well, there are the search engines and you can look them up for the next time instead of starting lamenting about the unfairness and showing the ignorance that is in in this case in my opinion demonstrating the serious negligence.
if someone is unaware of CAP, they're unlikely to have been working in an environment where it was relevant. if this same person starts working in an environment where it's relevant, i'd wager that it will be hard to remain unaware.
I have a CS degree. I learned about the CAP theorem, and I work with the nature of it daily. But my mind have ventured beyond the CAP theorem, it's implications are now mere facts of distributed systems, and the name "CAP theorem" had simply sliped out of my mind. If you want to quiz knowledge, rather than google-fu and name dropping, ask if you can make a distributed system with consistency and availability.
Instead of asking about Bubble sort, ask what is generally the complexity of a naive sorting algorithm and how good it can get for a general case.
I have investigated immutable hash tables but I do not know much about concurrent hash tables. This is my blind spot, so I do not know what the relevant question about the Hopscotch hashing would be.
Instead of asking what is CAP theorem, ask if you can make a distributed system with consistency and availability at the same time.
I think that this was the intention behind these questions after all and is the reason why I am trying to argue in support of them.
There's a middleground between those extremes. There's lots of knowledge that isn't relevant most of the time, but when it is relevant its worth its weight in gold. (And often these things aren't so obvious that people will google search them.)
When I'm trying to evaluate a candidate, I want people who proactively search out knowledge in that middle zone. People who understand complexity theory, and CPU caching behaviour, and the C10k/C10M problem just because they know they might need it some day. Because they want to work on problems where this stuff is relevant.
This sort of knowledge is the difference between hearing "our website is slow - I think its the database?" and "our website is slow - I took a look and our database server has run out of open file handles. But I'm still confused - that shouldn't explain the latency we're seeing."
The CAP theorem is a great example of this sort of middleground might-be-useful knowledge. Understanding how your database will behave in the face of network partitions is exactly the sort of thing that separates good engineers from great ones.
That's the point of the "hate" behind these questions.