Validators

A validator is a program that reads a test's input and asserts that it satisfies the problem's constraints — that $N$ really is within the bounds the statement promises, that the graph really is connected, that the file has as many lines as its first number claims. It grades nothing and never sees a participant's output. Its job is to keep malformed tests out of the problem.

The Validator row is on the problem's Testing tab and shows the validator's runtime, or unset. You need permission to write problems.

The editor is split: the code on the left, the results of running it on the right.

Fig 1. A validation run, grouped by testset, with one coloured square per test.

Trying a validator before storing it

Run compiles the validator currently in the editor and checks every test in the problem with it — without saving it to the problem. This is the point of the split editor: you write a candidate validator, run it against the tests you already have, see which ones it rejects, and only then decide whether the validator or the test is wrong.

Save stores the validator. Reset removes it, after a confirmation.

Reading the results

Results arrive live while the run proceeds, grouped by testset. A collapsed group shows a strip of coloured squares, one per test; expanding it gives the per-test list.

The run as a whole moves through Pending, Provisioning, Compiling and Checking to Complete, or ends in Error, Failure, Skipped or Cancelled.

Each test gets one of three verdicts:

Broken is not a validation failure; it means there was nothing to validate. Look at the generator for that test.

For a rejected test, the row shows the first line of the validator's own output instead of the word Invalid, so a validator that explains itself gives you the reason without another click. Inspect opens a dialog with the validator's full output, its standard error merged in, and the input it was fed.

Starting a new validation supersedes any earlier one still running, which ends up Cancelled.

Validation and test status

Once a validator is stored, it takes part in the normal life of a test: generated data has to be generated and validated before a test reads Ready, and a test whose input the validator rejects shows Invalid with the reason on hover. See Test cases.

Secret validators

A validator marked secret cannot be displayed. The editor offers Override, which replaces the stored validator with a new one; the original is not recoverable through the console.