Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Browse files

Add blog post on Code Jam qualification round.

  • Loading branch information...
commit 599e0bf5d193f51e1b24df7c7c4e348747ac1546 1 parent a4627b8
@hermannloose authored
Showing with 93 additions and 0 deletions.
  1. +93 −0 _posts/2012-04-15-codejam-qualification.md
View
93 _posts/2012-04-15-codejam-qualification.md
@@ -0,0 +1,93 @@
+---
+layout: post
+title: Code Jam 2012 – Qualification Round
+---
+
+I qualified for Google Code Jam 2012 yesterday.
+
+I began in the afternoon, well into the round, because I didn't want to stay up
+late—the start of qualification was 12am in Germany. Having solved only one
+practice problem before on Friday I didn't exactly feel "limbered up" at that
+point, but I had some basic Ruby stdin/stdout workflows memorized again.
+
+Upon reading the problems again, I just realized how ridiculously simple
+["Speaking in
+Tongues"](http://code.google.com/codejam/contest/1460488/dashboard#s=p0)
+must've really been (which is also what I was told on Google+ today) but back
+then, I panicked—even without the requirement to do particularly well—and went
+straight to the other problem descriptions.
+
+I settled on ["Recycled
+Numbers"](http://code.google.com/codejam/contest/1460488/dashboard#s=p2) as the
+first one to tackle. Brain partly locking up (I guess) it took me a short while
+to figure out actually how to traverse the possibilities to check for, which
+then turned into a straightforward implementation that was done rather swiftly.
+
+The authors' statement on them being "sure about the output to Case \#4" had me
+hunting the last bug—my output being 288, one larger—for quite some time.
+I realized that it had to be a duplicate pair of numbers somewhere, but at the
+time it didn't really occur to me how this should happen, so I took the
+*not-so-clever* approach—as in, I wouldn't want to ressort to that in an
+interview with the stakes being high—of manually going through the outputs of
+case \#4, until I stumbled upon 1212 generating both 2121 (in the form of
+212|1) and, well, 2121 (in the form of 2|121). Which made a lot of sense. Silly
+me.
+
+I fixed that rather bluntly by keeping track of the pairs generated, with the
+overhead of object creation probably contributing to less than amazing running
+time of the algorithm, but I wanted to be done with it and alas: first small
+input went through correctly in about two or three seconds.
+
+Satisfaction I then promptly ruined with the large input. I fired it up, sat
+back, figured it'd be running a little longer than three seconds—I only tried
+one test case with the bounds being 1 and 2000000 before that, to get a sense
+of the dimensions and whether my approach would be feasible at all, with using
+Ruby etc.—and got the silly idea of `tail ‑f`'ing the output file to see
+progress … of which there wasn't any. *At all.*
+
+Now, I'm not too sure about what *should* have happened there—straightforward
+suspicion being some buffering somewhere, as my test right now shows output
+after 50 test cases but then hangs, even after the program has finished—but
+after panicking and aborting the run two or three times (with maybe three
+minutes left in the end) I realized that output was in fact being written. Just
+not, you know, picked up by `tail`. So I started it up one last time aaaaand
+THERE GOES THE DEADLINE.
+
+*Really* disappointing.
+
+Wall time now says 2 minutes 49 seconds, with 2 minutes 46 seconds spent in
+userland crunching numbers. So it would've worked out, if only. Oh well.
+
+The last problem—enough points for the small input to qualify, which is where
+I stopped (after getting the large input correctly as well), because I wasn't
+feeling too frisky and wanted to do something else with the rest of the day—was
+["Dancing With The
+Googlers"](http://code.google.com/codejam/contest/1460488/dashboard#s=p1),
+a nice, kind of real world one.
+
+Start with the total scores sorted, divide each by three, decide whether they
+can reach the minimum best score in any way and decrement the allowed
+"surprising cases" whenever one occurs. Easy enough.
+
+Except I made a stupid mistake. Plus another. And then some.
+
+So it took me *seven* (!) incorrect small inputs until I finally got it right.
+Lots of "fixing that small error over there" and not properly thinking the
+whole thing through again. I'm not proud of that. Such performance in an
+interview would likely have resulted in an epic fail. But then again, my mind
+wasn't too fresh yesterday.
+
+### Conclusion ###
+
+I don't think I will stand any chance against the masses of lightning-fast
+coders with their razor-sharp minds from all over the world (though especially
+China and Russia seem to rule) to advance in one of the next rounds. I panic
+way too easily when faced with a clock almost audibly ticking, something I've
+just barely managed to keep in check in the phone screens leading up to my
+internship at Google in London last year. (I'm still a bit puzzled I got in
+back then and get to go back for another one this year, but they seemed to like
+my work.)
+
+But if anything, it will be a very good way of warming up to that pressure and
+dealing with it, so I guess I'll just keep at it.
+
Please sign in to comment.
Something went wrong with that request. Please try again.