Efficient learning of One-Variable Patttern Languages From Positive Data