假设您有一个名为 numbers.txt 如下所示：

10
5
8
7
3
6
9
11
3
1

你可以这样实现你的目标：

int desiredSum = 15;
SparkSession spark = SparkSession
        .builder()
        .appName("My App")
        .master("local[*]")
        .getOrCreate();
Dataset<Row> rdd = spark
        .read()
        .text("numbers")
        .withColumnRenamed("value", "number")
        .withColumn("number", col("number").cast(DataTypes.LongType));
rdd.createOrReplaceTempView("myTable");
spark.sql("select first.number, second.number as number_2 from myTable  first inner join myTable second on first.number + second.number =" + desiredSum + " where first.number <= second.number").show();

+------+--------+
|number|number_2|
+------+--------+
|     5|      10|
|     7|       8|
|     6|       9|
+------+--------+

或者，如果数据很小，您可以使用spark中的笛卡尔积实现您的目标，如下所示：

int desiredSum = 15;
SparkSession spark = SparkSession
        .builder()
        .appName("My App")
        .master("local[*]")
        .getOrCreate();
Dataset<Row> rdd = spark
        .read()
        .text("numbers.txt")
        .withColumnRenamed("value", "number")
        .withColumn("number", col("number").cast(DataTypes.LongType));
Dataset<Row> joinedRdd = rdd.crossJoin(rdd.withColumnRenamed("number", "number_2")).filter("number <= number_2");
UserDefinedFunction mode = udf((UDF2<Long, Long, Object>) Long::sum, DataTypes.LongType);
joinedRdd = joinedRdd.withColumn("sum", mode.apply(col("number"), col( "number_2"))).filter("sum = " + desiredSum);
joinedRdd.show();

结果如下：

+------+--------+---+
|number|number_2|sum|
+------+--------+---+
|     5|      10| 15|
|     7|       8| 15|
|     6|       9| 15|
+------+--------+---+

**take into account the Order of time and space complexity when you use Cross join**