Code Golf at Google
blog.zmxv.comFun fact: there has been a problem with Googlers calculating hash functions that pass all the test cases for certain questions.
If you think about it the right way, that's what everyone is doing: calculating a hash from the problem statement into stream of symbols with certain constraints.
So the problem here is not calculating hash functions per se, but over-fitting the solution to those test cases.
i think the sad pattern I see in a lot of these code golf implementations is the solution score is based on Lines of Code, which we already agree is a POOR representation of either complexity or simplicity.
I think it would be far more interesting to measure number of instructions to complete the objective. Though I suspect the folks familiar with assembly would run circles, perhaps you could segment by solution language?
Given the entirely whimsical and (hopefully) intentionally rebellious nature of Code Golf, I don't think that worrying about the usefulness of the metric is really worth it.
You may be interested in http://www.zachtronics.com/tis-100/. It's essentially a code golf game with many different metrics to optimize including number of instructions.
wow, thanks! great line: "It’s the assembly language programming game you never asked for!"
The demoscene may be of interest if you're into optimising the binary for size and not the source:
http://www.hugi.scene.org/compo/
That one seems abandoned for a few years now but you can find plenty of tiny graphical effects here, executables are 512 bytes and below:
http://www.pouet.net/prodlist.php?type%5B%5D=32b&type%5B%5D=...
The reason lines of code is a poor way to score solutions is that you should be counting characters! :-)
My favorite Perl golf challenge is something I got from Boston Perl Mongers many years ago:
You have the numbers 1 to 9 (in order), and in between each number you can insert a plus, a multiply, or nothing. Your program should print all expressions that evaluate to 2002. For instance it should print the solution
(and one other)1*23+45*6*7+89I have a command-line solution in 80 characters, including the `perl ` at the beginning. It works on 5.10 and 5.16. I think there was a better solution back in the 5.8 days, but my notes are unclear.
Once you know the other solution you can do much better than 80 characters!
Lines of code is a poor representation, but non-whitespace characters is a somewhat better one. Perl golf and vim golf both use that.
You are not describing code golf, though. "A sad pattern in chess is checkmate; the goal should be to eliminate the other player entirely."
I agree what you describe would be fun, too, but it isn't code golf.
Hah, I remember spending hours on this during work hours when I was interning last year. I used to come up with very clever and short solutions only then to find out I wasn't even top 10 :(
Perl is a way more fun for golfing but it's missing from the supported languages
I had to do a code test for a job (sigh). But it actually was kind of fun. The website takes your code and checks correctness and efficiencies.
The company posts "challenges" every month. But they also have lessons which contain some interesting code challenges. The only problems is 1) they're timed 2) sometimes the descriptions are kinda mathy which may put people off.
Can a googler put the questions in public for others to see?
I'm trying to run the c++ version, but it won't compile because the loop variables in the range for loops do not have declarations.
There was a proposal[1][2] for C++17 to allow:
which would be equivalent to:for (e : c)
Gcc implements this when compiling with -std=c++1z. My understanding is that it was rejected by the committee.for (auto &&e : c)[1] http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n385...
[2] http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n399...
How does this work, some sort of TLD magic? (That shouldn't be possible, right?) Is it just routed within Google's internal network?
Search paths. The full address is go.corp.google.com (which is simply a URL shortener), IIRC; however, I think the resolvers are also configured to respond to a bare name in a lot of cases as an optimization. They talk a little bit about corp in their BeyondCorp paper[0], which is well worth a read, and I'm speaking to ancient memory so I might be wrong these days.
[0]: http://static.googleusercontent.com/media/research.google.co...
Thanks for the link! Figured it was a bit more complex than just a host file, being Google and all.
When you're on a corp device, your DNS search path includes google.com. It's not magic at all, it's over here: http://goto.google.com/
(You're going to see a login page that you can't authenticate to)
By default foo resolves to foo.mydomain.mytld (man resolv.conf, section domain), so as long as the machines are configured with a local domain, this will work.
It's probably resolved by the internal network DNS. You can do the same even with your home router.
I'd assume the same way localhost works on your computer.
It works the same as any other corporate network: either custom DNS resolution on the machine (i.e. /etc/hosts), DNS resolution in the internal network, or DNS resolution on the VPN. Or, more likely, a mix of the 3.
To get to go links, you've got to be connected to Google's network (either physically, or through VPN). It's essentially the same as accessing a computer on your LAN with its host name (in which case, the router can do the DNS lookup and resolution for the internal network).
Fascinating but totally incorrect analysis. All anyone needs to get to go/ is to put corp.google.com in their resolver search paths. Of course, such people will also need a google.com login to actually be redirected to the destination.
I watched the wolf of wall street recently. I see some of the solutions to these code golf exercises as similar to the bankers who beat their chests and celebrate their masculinity in selling penny stocks to gullible investors.These code golf exercises seem like the software developer equivalent: "look at how brilliant and superior I am that I can obfusicate code into something no-one understands in only 2 lines".
My concern here is that this Code Golf mentality infects the normal day-to-day coding at Google. Just because an implementation can spawn a few lines it doesn't make it a good implementation. What about clean code, maintainability, automated testing, self documenting code...
I pity the codebase I would have to maintain from one of these code golfers.
Nice how many off-handed insults you've managed to pack into such small space, based on no data at all except extrapolating from your own prejudices. It's almost like some kind of HN comment golf.
A playful programming competition is really not at all comparable to cheating innocent people out of their money. Like, seriously. People are generally able to behave appropriately in situations with different expectations. There's no reason to think that it's not equally applicable to using the appropriate programming style for the project. It's like you're looking at somebody walking on the street in jeans and a T-shirt, and complaining that they're not properly dressed for the opera. And no code gets committed without a code review at Google anyway.
Finally, code golf is rarely about obfuscation; obfuscation just for its own sake isn't compatible with minimizing code length. Once you e.g. know the basic Perl golf tricks, the code can be surprisingly readable since it really just contains the core of the algorithm.
Insult golf sounds kinda fun. Try to come up with the most insulting thing possible in the least characters.
On a more serious note, I agree with you. There's some alarmist reactions to things like this, but in reality I don't think it's an issue at all. Sure, there might be bragging and such, but that's the point of something like this, isn't it? It's a game, not a serious discussion on optimization. Short code is more "fun" than superoptimized to my mind.
My concern here is
If people write a haiku
They might never stop.