Spark-datasetops is a tiny library that aims to make Spark SQL Dataset more developer friendly by bringing back the operators we all love to use on key-value RDDs (the ones defined in PairRDDFunctions). I suppose these operators were nixed since they cannot easily be ported to Java and Python, but given how easy it is to bring them back it seems like a shame not to have them.
To use simply import RichPairDataset which is an implicit class:
import com.tresata.spark.datasetops.RichPairDataset
RichPairDataset adds methods to a any key-value Dataset (Dataset[(K, V)]). Some examples taken from unit tests:
scala> import spark.implicits._
scala> import com.tresata.spark.datasetops.RichPairDataset
scala> val x = Seq((1, 2), (1, 3), (2, 4)).toDS
scala> val y = Seq((1, "a"), (1, "b"), (2, "c")).toDS
scala> x.mapValues(_ + 1).show
+---+---+
| _1| _2|
+---+---+
| 1| 3|
| 1| 4|
| 2| 5|
+---+---+
scala> x.aggByKey(0, { (b: Int, a: Int) => b + a }, { (b1: Int, b2: Int) => b1 + b2 }, { (b: Int) => b }).show
+-----+-----------+
|value|anon$1(int)|
+-----+-----------+
| 2| 4|
| 1| 5|
+-----+-----------+
scala> x.countByKey.show
+-----+--------+
|value|count(1)|
+-----+--------+
| 2| 1|
| 1| 2|
+-----+--------+
scala> x.joinOnKey(y).show
+---+-----+
| _1| _2|
+---+-----+
| 1|[2,b]|
| 1|[2,a]|
| 1|[3,b]|
| 1|[3,a]|
| 2|[4,c]|
+---+-----+
Enjoy! Team @ Tresata
Starting with release 1.0.0 this library compiles against Spark 3. Because of this Spark 2 and Scala 2.11 are no longer supported. We are still compiling with Java 8.