使用值使用 groupbykey 后对 rdd 进行排序

Question

I have JavaPairRDD as

JavaPairRDD<String, Iterable<Row>> rdd = mydataset.orderBy("orderfield1", "orderfield2").javaRDD().mapToPair(row -> new Tuple2<>(row.getAs("id").toString(), row)).groupByKey()

As groupbykey() doesn't maintain order orderby doesn't work here. I want to order the Iterable<Row> using some of the fields from dataset.

Answer 1

You could transform the Iterable into a List and then sort that list like below. I assume that your sorting field is called x and that it is of type String but you can obviously adapt that to your specific case.

String sortingField = "x"
JavaPairRDD<String, List<Row>> rdd = mydataset
    .javaRDD()
    .mapToPair(row -> new Tuple2<>(row.getAs("id").toString(), row))
    .groupByKey()
    .mapValues(it -> {
        List<Row> rows = new ArrayList<>();
        it.forEach(rows::add);
        rows.sort(
            (Row a, Row b) -> a.<String>getAs(sortingField).compareTo(b.<String>getAs(sortingField))
        );
        return rows;
    });

Note that this is much simpler to write in scala:

val rdd = mydataset
    .rdd
    .map(row => (row.getAs("id").toString, row))
    .groupByKey
    .mapValues( _.toSeq.sortBy(_.getAs[String]("x")))