結果 : dataframe remove duplicates keep first