The first records of the input dataset will be used