3 ms·
but programmer has to know what ds/algorithm to choose, of course you can look up complexities and stuff but it doesn't help in cases where performance tuning o
by dominotw 10y ago
but programmer has to know what ds/algorithm to choose, of course you can look up complexities and stuff but it doesn't help in cases where performance tuning of caches ect is necessary. you have to understand what that particular algorithm/ds is actually doing to tune it properly.
- cookiecaper 10y agoIt's much more important to get a smart and adaptable programmer than one who memorized his compsci textbooks. Yes, an understanding of data structures and algorithms can be helpful, but unless messing around in that space is really going to comprise a large percentage of on-the-job duties (and as the grandparent indicated, it's not for 99% of us who can just use the canned implementations from the stdlib), it's stupid to make it a large percentage of success in the interview process. I'm glad to hear that Amazon kept the textbook-style questions light -- it's not what you usually hear about happening in interviews with big tech companies.
- vonmoltke 10y ago> I'm glad to hear that Amazon kept the textbook-style questions light -- it's not what you usually hear about happening in interviews with big tech companies. Yeah, this was a pleasant surprise about the process for me.
- johnwheeler 10y agoYour example of performance tuning caches is a perfect. It's enough to generally know about caches. Databases have query caches. Memcache and Redis are memory caches. Retrieval is O(1) because they're based on hashes. You can influence object expiration and turn on LRU. I can't implement a cache with LRU by the seat of my pants though, but I'm experienced enough to know you shouldn't do that anyways. And, I know enough to read the docs and ask intelligent questions. None of that changes the fact that when my cache blows up in production, I'm going to be in the exact same position as my all-algo-knowing colleague -- scared, grasping at straws, and applying grit to fix the problem.
- plandis 10y agoI've always considered an LRU cache question to be pretty decent. The general idea is fairly obvious (linked list + hash table) and straightforward to get something working and is a good indicator on how someone can synthesize data structures to fit the needs of the problem. Plus there is a lot of open endedness that makes it easy to drill down into someone's thought process.
- greenleafjacob 10y agoWow, I could not agree more. What you are calling grit is massively under-measured in the current bullshit whiteboarding interviews for software developers today.
- johnwheeler 10y agoThanks. Another name for it is StackOverflow :-P
- collyw 10y agoI learned that sort of stuff 15 years ago. By now it is forgotten, but I am perfectly capable of reading up on stuff and relearning it. I learn new stuff all the time in this line of work.
- dominotw 10y ago> but I am perfectly capable of reading up on stuff and relearning it How do you prove this to a potential employer though? Also, it should be possible to read up on algorithms for the interview.correct?.
- johnwheeler 10y agoYes, but in that case you're reading up on algorithms for the sake of the interview and not the job. There, it becomes a filter or rite of passage. There are better filters, and rites of passage (in employment at least) are complete bullshit. My favorite filter is to just ask the candidate to talk about technology, open ended, like they would to a friend. You get a real sense of what they care about, and you give them a chance to lead the discussion. Do they ask questions back to you (I hope so), or do they monopolize the discussion. Are they even prepared to have an open ended discussion? You'd be surprised how many 'software engineering' candidates can't do that.