As we all know, in yesterday’s Starters there was a problem POSSPEW, and in that problem many of the O(t\cdot n^2) got passed without TLE, and it is obvious that with a constraint of N \leq 10^5, k \leq 10^9 and t \leq 10^4, it was not intended to pass.
So, to avoid such weak test cases, CodeChef can introduce “hacking solutions” feature like one Codeforces has. That feature proves to be really helpful in these circumstances.
Should CodeChef introduce hacking feature?
Yes
No
0voters
Any additional suggestions/feedback is much appreciated!
I think the @admin should stop everything else and do the plag checks first otherwise it won’t survive long. Already , codechef has a very very bad reputation in the community.
This problem is not an isolated incident and on codechef I have witnessed this multiple times. But I think codechef purposely tries to create such type of “weak” tests where the naive solution passes to reduce the load on server. I don’t think they intend for the naive solution to pass but it’s probably a side effect in trying to not create tests that increase the load on servers. I say this because it’s comparatively rare to see tests being “weak” in the sense that they miss some corner case, at least from my experience. So it’s not the case that the tests for the problems aren’t prepared thoroughly. This is just me speculating though, I could be wrong. So introducing hacks is still a problem because it will bring higher load on server with it. In case my speculation is wrong and this is simply a case of unintentional weak tests then hacks is definitely a good solution.
This isn’t true. The time limits are set keeping the server load in mind.
The weak test cases are just an unintended effect of having few test files.
CodeChef problemsetting platform is static, in that solution codes, validators, generators are run client side and only the final test cases are uploaded to the judge. Unless the platform is changed radically (making it similar to polygon), the feature for hacks isn’t possible.
Say I’m preparing a problem. I prepare the generators, validators and solution code. Then, on my computer, I run the generators to create test input files, validators to check if the test cases match the constraints (as in the problem) and finally the solution code to generate the output files.
At the end, only the generated input and output files are uploaded to CodeChef.
Yeah, so if someone says tests are weak for a problem, then we should agree that the problem is not with the platform, but with the generator program and the person who wrote it/approved it.
I mean it is always possible to generate strong tests using good generator programs, so there would not be any need for hacking submissions.
And I agree, more number of test files help ensure stronger tests. For example, Hackerrank allows around 50 test files. This will ensure every corner case is included in the test.
Hacks feature is prompting users to write obfuscated code making it hard for new users to read code and understand solutions while learning.
Some YouTube channels are uploading solutions before the end of contest. Making it easy for others who hasn’t solved the question to submit accepted answers. Which is very unethical.