试试这个-
我已将列名作为别名提供给 sum(column) ```
Dataset df = spark.range(2).withColumn("value", lit(2));
df.show(false);
df.printSchema();

    /**
     * +---+-----+
     * |id |value|
     * +---+-----+
     * |0  |2    |
     * |1  |2    |
     * +---+-----+
     *
     * root
     *  |-- id: long (nullable = false)
     *  |-- value: integer (nullable = false)
     */
    Map<String, String> map = new HashMap<>();
    for(String column:df.columns())
        map.put(column, "sum");

    List<Column> cols = map.entrySet().stream().map(c -> expr(String.format("%s(%s) as %s", c.getValue(), c.getKey(), c.getKey())))
            .collect(Collectors.toList());

    df.agg(cols.get(0), toScalaSeq(cols.subList(1, cols.size()))).show(false);
    /**
     * +---+-----+
     * |id |value|
     * +---+-----+
     * |1  |4    |
     * +---+-----+
     */

实用程序-

Buffer toScalaSeq(List list) {
return JavaConversions.asScalaBuffer(list);
}