Slashdot Mirror
Programmers Are Confessing Their Coding Sins To Protest a Broken Job Interview Process (theoutline.com)
A number of programmers have taken it to Twitter to bring it to everyone's, but particularly recruiter's, attention about the grueling interview process in their field that relies heavily on technical questions. David Heinemeier Hansson, a well-known programmer and the creator of the popular Ruby on Rails coding framework, started it when he tweeted, "Hello, my name is David. I would fail to write bubble sort on a whiteboard. I look code up on the internet all the time. I don't do riddles." Another coder added, "Hello, my name is Tim. I'm a lead at Google with over 30 years coding experience and I need to look up how to get length of a python string." Another coder chimed in, "Hello my name is Mike, I'm a GDE and lead at NY Times, I don't know what np complete means. Should I?" A feature story on The Outline adds: This interview style, widely used by major tech companies including Google and Amazon, typically pits candidates against a whiteboard without access to reference material -- a scenario working programmers say is demoralizing and an unrealistic test of actual ability. People spend weeks preparing for this process, afraid that the interviewer will quiz them on the one obscure algorithm they haven't studied. "A cottage industry has emerged that reminds us uncomfortably of SAT prep," Karla Monterroso, VP of programs for Code2040, an organization for black and Latino techies, wrote in a critique of the whiteboard interview. [...] This means companies tend to favor recent computer science grads from top-tier schools who have had time to cram; in other words, it doesn't help diversify the field with women, older people, and people of color.
Hi, my name is Vince. I interviewed for Amazon, specifically for their PHP API for AWS development team. Despite an entire background of 10+ years of developing front-facing PHP APIs for other businesses, plus having a major part of my code available for public review on GitHub, I failed their interview process because they wanted me to write a specific type of searching and sorting algorithm, by hand, on white-board. This type of code would never have been used on the job, ever. Yet this is what they interview on. The job was to build a PHP API so PHP developers can call basic PHP functions, and the library would translate them over to HHTPS calls to AWS. All of the complex computing/searching/sorting is handled by the existing AWS services.
Honestly, I see all of the complaints above as whines by people who don't understand what a whiteboard interview is.
* If I ask you to write a sorting algorithm on a white board, I am not asking you to by rote copy out a bubble sort - in fact, if you do it by rote, I'm likely to go "oh, well that was uninformative, he didn't solve any problems, he just copied something out" (and that's why I'm unlikely to ask you to sort things, but instead, something more obscure)
* If you "don't do riddles" then I actively don't want to hire you - the entire purpose of a software engineer (i.e. not a flunky programmer) is to do riddles, all day, every day. If you don't want to do that, you don't want to do the job I'm interviewing you for.
* If you need to look up how to find the length of a string in python, I don't give any shits. I don't care if you write x.length(), x.size(), x.count, length(x), I'm asking for you to solve a problem, not get the code in the exact shape it'll compile in.
* If you don't know what NP complete means, I don't care. Of course I'm going to probe a bit into whether you can analyse the performance of your code - that's absolutely part of the job, and something I need to know you can do. But not knowing what one term means is not going to not get you hired.
Long story short - don't assume that everything in a whiteboard interview should be taken literally. I don't want to find out if you can write exact algorithm x perfectly so that it will compile. I want to know if you can solve a problem, and can talk rationally about your solution, its trade offs, its performance, etc.
Bubble sort is also good for almost sorted datasets (pretty much n in this case). It's used for very fast broad phase collision detection where overlaps are detected during swaps. Since the sort happens every timestep, the endpoints stay pretty much sorted and the broad phase collection detection runs in near n time.