an array of token groups is calculated successfully and it's probably fast enough.
made row/column positions make sense