If you have written some programming problems, and have prepared test cases, you will probably experience the terrible feeling that some test cases may be invalid (meaning it does not agree with the constraints in problem statement): upper bound can be violated, your graph not satisfied connectivity requirements or is not at tree... It is reasonable to feel that way. Even experienced problem setters make mistakes sometimes (for example, in the prestigious ACM ICPC World final 2007).

It is strictly recommended to write a special program (called validator) to formally check each test to satisfy all requirements from problem statements. Validators are strictly required for problems on Codeforces. Polygon has built-in support of validators.

It is really easy to write a validator using testlib.h.

Example

Following is a validator that could be used for problem 100541A - Stock Market:

#include "testlib.h"

int main(int argc, char* argv[]) {
    registerValidation(argc, argv);
    int testCount = inf.readInt(1, 10, "testCount");
    inf.readEoln();
    
    for (int i = 0; i < testCount; i++) {
        setTestCase(i + 1);
        int n = inf.readInt(1, 100, "n");
        inf.readSpace();
        inf.readInt(1, 1'000'000, "w");
        inf.readEoln();

        inf.readInts(n, 1, 1000, "p");
        inf.readEoln();
    }

    inf.readEof();
}

#include "testlib.h"

int main(int argc, char* argv[]) {
    registerValidation(argc, argv);
    int testCount = inf.readInt(1, 10, "testCount");
    inf.readEoln();
    
    for (int i = 0; i < testCount; i++) {
        int n = inf.readInt(1, 100, "n");
        inf.readSpace();
        inf.readInt(1, 1000000, "w");
        inf.readEoln();

        for(int i = 0; i < n; ++i) {
            inf.readInt(1, 1000, "p_i");
            if (i < n-1)
                inf.readSpace();
        }
        inf.readEoln();
    }

    inf.readEof();
}

The wonderful thing about this validator is that it is very simple and it is very difficult to write something incorrect.

More examples can be found at the Github repo

Available methods

The first line of your code should be registerValidation(argc, argv) which does some magic in the background, so that you can use the necessary methods. Most methods for validators are methods of the input stream inf, they start with prefix read and do the same thing: moves input stream pointer to next suitable place after reading something. It also detect violations (input does not match what you are trying to read), and then throw error. Usually the method's name is inf.readT[s], where T is one of Int, Long, Double, Token, Line, Space, Eoln, Eof, etc, depending on the data you want to read, while the s suffix is needed when you want to read an array of similar data. The complete list of such functions is given at the end of this post. Apart from these functions, the following are useful:

variableName argument

It is recommended to additionally pass a string argument called variableName to read* functions to make the error message human-readable. So it is preferred to use inf.readInt(1, 100, "n") instead of inf.readInt(1, 100). The first statement will fail with human-readable message like FAIL Integer parameter [name=n] equals to 0, violates the range [1, 100].

Using ensure/ensuref

To check a requirement (like a graph doesn't contain loops, i. e. x_i ≠ y_i) use ensuref(x_i != y_i, "Graph can't contain loops"). It is allowed to use C-language format specifiers like ensuref(s.length() % 2 == 0, "String 's' should have even length, but s.length()=%d", int(s.length())). Also you can use simple form like ensure(x > y), it will print failed condition if it doesn't hold in form FAIL Condition failed: "x > y".

Notes

• Validator is strict. It cares about correct placing of spaces. For example, when you're trying to read an integer and the next character is a space (and then an integer), the validator will throw error.
• Some methods have "regex" feature. It is not a full-featured regex as you may have used in many programming languages. It is a very simple version, which supports the following:
  • Set of character, e.g: [a-z] is a lowercase latin letter, [^a-z] matches anything but a lowercase latin letter.
  • Range, e.g. [a-z]{1,5} is a string of length 1 to 5 consists of only lowercase latin letter.
  • Or operator, e.g. mike|john is either mike or john.
  • Optional character, e.g. -?[1-9][0-9]{0,3} will match non-zero integers from -9999 to 9999 (note the optional minus sign).
  • Repetition, e.g. [0-9]* will match sequences (empty or non-empty) of digits, and [0-9]+ will match non-empty sequences of digits.
• Also regarding regex, very simple greedy algorithm is used. For example, pattern [0-9]?1 will not match 1, because of greedy nature of matching.

Following is full list of methods of the input stream (inf.<method>):

Reference: Github page of testlib.h