Hello, Codeforces!

If you're developing problems in Polygon, you might like this image to attract attention.

For the new members of the community, I would like to remind you that Polygon (https://polygon.codeforces.com/) is a system developed and maintained by Codeforces for preparing programming problems. It is there that authors and coordinators develop all the problems for the rounds. Moreover, I believe a significant (large?) portion of the problems for other competitions is also developed there: various stages of ICPC, national competitions of different levels, educational problems for various courses, etc. In 2023, more than 50000 problems were prepared in Polygon (only those for which a package was compiled are counted)!

We often implement such improvements in Polygon that make the process of preparing a problem more reliable. These often include various automatic warnings and tips for problem developers and/or coordinators.

Today I want to talk about two recent such innovations in Polygon. Both of these innovations exploit validators written in Testlib. They will work only if you write the validator correctly:

• Use variables exactly as they are named in the problem statement. For example, if $$$t$$$ is the number of test cases and $$$a$$$ is a given array, then int t = inf.readInt(1, 10'000, "t"); and vector<int> a = inf.readInts(n, -100, 100, "a"); are correct, but int t = inf.readInt(1, 10'000, "testCases"); and vector<int> a = inf.readInts(n, -100, 100, "ai"); are not.
• Use the built-in features of Testlib: for reading arrays, use readInts/readLongs etc. Whenever possible, check constraints through the arguments of read functions, not with custom code afterward.

In the Polygon/Codeforces ecosystem, a properly written validator not only ensures the correctness of the test but also helps to analyze and process the test. Here, read https://codeforces.com/blog/entry/105779 about how we highlight input data sets in the test when displayed in the problem statement.

Warnings about Incomplete Tests

When you build a package, you may see a warning like this from the system. Please do not ignore it!

We extract this information from the validator. Most variables from the test in the validator are read either as int t = inf.readInt(1, 10'000, "t"); (a single number) or as vector<int> a = inf.readInts(n, -100, 100, "a"); (an array of n numbers). The validator can report externally that on some of the tests a lower/upper limit of the variable's constraints was (or was not) reached. And if a limit was not reached in a test set, a warning will appear.

Sometimes (although quite rarely) the limit is indeed not reached or there is no reason to require its achievement. For example, in a problem with test cases, the case $$$t=1$$$ may not have special meaning (but often it does, since maximum tests are often possible only in this case). Then you can use a notation like int t = inf.readInt(1, 10'000, "~t");. The tilde to the left of the variable simply indicates that there should be no warning that the lower limit was not reached in the tests. Similarly, the tilde can be placed to the right (upper limit), or even on both sides.

Once again, I urge the use of methods to suppress warnings only in rare cases when they are indeed absolutely irrelevant. In any other case, just add a test of the required type, and this will make the problem better.

For developers of Codeforces round problems – it would be good to check that there are no such warnings for pretests. You can simply run a solution on a pretests testset and at the bottom of the invocation report, there will be warnings (if they are present).

Warnings about Mismatches between Constraints in Problem Statements and Validators

I thought about this feature for a long time but got down to implementation after a recent incident https://codeforces.com/blog/entry/124696?#comment-1107429 Fortunately, it affected only a very small number of participants. But such a mistake can significantly affect the course of the contest.

Now, if you set constants while checking limits in the validator (like here int t = inf.readInt(1, 10'000, "t");), these bound constants can be returned to the system by the validator. Besides, I wrote code that, in simple cases, extracts similar information from LaTeX markup. In case of a mismatch of limits, you will see a message like this.

Of course, this automation works only for simple constant constraints. However, they are most common. There will probably be rare cases of false positives. Please report them to me.