Advent of Code 2020 • Hacker News Board

Re: Advent of Code 2020

#81

Post by igouy » Wed, Dec 02, 2020, 6:04 PM UTC

tasubotadas wrote: post #79

AoC is great. But I also think that it's a missed opportunity to measure productivity and runtime performance of programming languages. In "Benchmarks Game" you get extremely optimized programs which is not usually achievable by an average developer. On AoC you would get a wide distribution of submissions and you could get the right estimate of how performant are the programs submitted by PHP, Python, or Rust develop…

> In "Benchmarks Game" you get…

You get what people choose to contribute — if you want to see Python or PHP programs that don't use GMP then please contribute those programs.

The benchmarks game shows up-to 10 programs for each language implementation: some will be all about performance, some will be about the language, some will be about program size.

https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

and

https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

Sort by gz and notice the Chapel programs seem to be written to show program size rather than performance:

https://benchmarksgame-team.pages.debian.net/benchmarksgame/...

Re: Advent of Code 2020

#82

Post by andy_ppp » Fri, Dec 04, 2020, 12:07 PM UTC

xyzzy_plugh wrote: post #2

I've always found AoC to be added stress during a time of the year I'm not really running a stress-deficit. I find the competitive aspect a pretty big turn off. I usually get a few puzzles in and then burn out, because I can't keep up with the people who do it every day (I don't even use my computer every day). I wish the scoring didn't prioritize timeliness. This is really the sort of thing I'd enjoy cranking out on…

I hadn't even realised it was a competition - I'm not really interested in that... I just want to improve my coding exercise ability should I ever interview again and I'm looking forward to it getting absurdly hard and 100% CHEATING. But I will still learn something so I'm happy to be taking part this year!

Re: Advent of Code 2020

#83

Post by sathish316 » Fri, Dec 04, 2020, 4:28 PM UTC

Is Advent of code used mainly for learning new languages? I find it very different from Topcoder style questions. I tried 3 puzzles and part1 and part2 in each question makes this a good test of how to design code in order to make it extensible for future changes. Can someone who has done Advent of code in the past confirm if learning new languages and making your code extensible are the primary objective of AoC questions?

Re: Advent of Code 2020

#84

Post by will_pseudonym » Sat, Dec 05, 2020, 1:29 AM UTC

sammorrowdrums wrote: post #76

Jtsummers wrote: post #71

Earlier quoted context omitted.

I suspect that thy learned these things, but it didn't stick. From memory of 2018/2019 things that the younger colleagues didn't know (or didn't know well enough to get without prompting): - When to use CFG versus regex for parsing. Many inputs could be parsed with regexes if you made certain assumptions and got lucky. But CFGs were much easier for some. - Shunting-yard algorithm. This came up, I think, in 2018. - Us…

Excellent comment, I did both parts last year up to day 17 and just had to stop, but learning about things like A* were really what made it worthwhile. I have no formal CS training, but did a lot of online courses and so I use events like this to help plug gaps and learn new languages, and it would have been great to have a mentor like yourself through some of these puzzles last year as it can be quite hard to solve…

I'm in the same boat as you having no formal college/university CS training. You've given me a good idea for a blog series. Writing posts that go through the solutions I came up with for each day of AoC if I can do so without benefit of knowledge of useful algorithms & data structure concepts, and then reading the best solutions I can find on reddit etc, and write a blog post on how it works, as well as what underlying concepts it utilizes, and how they work. If I do that for all days/years of AoC, I could post a summary of the most useful concepts that I learned, that were applied to the most problems of AoC.

Re: Advent of Code 2020

#85

Post by abhibansal530 » Sat, Dec 05, 2020, 4:18 PM UTC

gsinclair wrote: post #74

a_imho wrote: post #42

Why Advent of Code, how is it different to the million other coding challenges online? I never participated, only looked at some of the puzzles.

My experience of other coding challenges is: here's the problem, and you submit source code which is run against many test cases to confirm correctness. Advent Of Code only asks you to submit an answer, not source code.

https://projecteuler.net has the same concept, but it's more oriented towards math problems.

Re: Advent of Code 2020

#86

Post by thom » Sun, Dec 06, 2020, 1:53 PM UTC

ojnabieoot wrote: post #9

recursive wrote: post #5

Earlier quoted context omitted.

If you must look at the scores, you can now make your own private leaderboards. If you're the only member of the leaderboard, then you'll win every challenge.

This is dismissive and unnecessarily condescending. Is it possible to opt out of the public leaderboard entirely? Because if not - perhaps things have changed - the stressful part about AoC is having a publicly viewable permanent score based on arbitrary and unfair criteria. Simply “not looking at it” isn’t enough. I did AoC a few years ago and realized that I had my GitHub profile associated to “middle-of-the-road”…

I just find the idea that anybody would judge you on the scores absolutely crazy. Have you looked at people's code?! Do I want to hire people who can create unreadable one liners, then rewrite them into another unreadable one liner for the second part of each puzzle? Do I want to make it impossible to hire people with jobs, kids, or people in different timezones? Would you want to work somewhere that used this sort of thing to decide how to build teams?

Advent of Code for me has always been about creativity, learning new languages, some fun with co-workers. Occasionally after finding the simplest solution, I'll try and work out algorithms with better performance, or even read papers on the underlying fundamental computer science problem involved. But I try to write the code as I would in the day job, and so I always use the example inputs to write unit tests and I always try to refactor stuff to be readable. I'd much rather see code I could imagine working with in real life on someone's GitHub profile than something absurdly terse.

All that said, the puzzles this year have been pretty simple and do seem slightly more optimised to people on speedruns than some years previously, so it's not been that stimulating so far.