|
| 1 | +package part5rddtransformations |
| 2 | + |
| 3 | +import generator.DataGenerator |
| 4 | +import org.apache.spark.sql.SparkSession |
| 5 | +import scala.collection.mutable |
| 6 | + |
| 7 | +object I2ITransformations { |
| 8 | + |
| 9 | + val spark = SparkSession.builder() |
| 10 | + .appName("I2I Transformations") |
| 11 | + .master("local[*]") |
| 12 | + .getOrCreate() |
| 13 | + |
| 14 | + val sc = spark.sparkContext |
| 15 | + |
| 16 | + /* |
| 17 | + Science project |
| 18 | + each metric has identifier, value |
| 19 | +
|
| 20 | + Return the smallest ("best") 10 metrics (identifiers + values) |
| 21 | + */ |
| 22 | + |
| 23 | + val LIMIT = 10 |
| 24 | + |
| 25 | + def readMetrics() = sc.textFile("src/main/resources/generated/metrics/metrics10m.txt") |
| 26 | + .map { line => |
| 27 | + val tokens = line.split(" ") |
| 28 | + val name = tokens(0) |
| 29 | + val value = tokens(1) |
| 30 | + |
| 31 | + (name, value.toDouble) |
| 32 | + } |
| 33 | + |
| 34 | + def printTopMetrics() = { |
| 35 | + val sortedMetrics = readMetrics().sortBy(_._2).take(LIMIT) |
| 36 | + sortedMetrics.foreach(println) |
| 37 | + } |
| 38 | + |
| 39 | + def printTopMetricsI2I() = { |
| 40 | + |
| 41 | + val iteratorToIteratorTransformation = (records: Iterator[(String, Double)]) => { |
| 42 | + /* |
| 43 | + i2i transformation |
| 44 | + - they are NARROW TRANSFORMATIONS |
| 45 | + - Spark will "selectively" spill data to disk when partitions are too big for memory |
| 46 | +
|
| 47 | + Warning: don't traverse more than once or convert to collections |
| 48 | + */ |
| 49 | + |
| 50 | + implicit val ordering: Ordering[(String, Double)] = Ordering.by[(String, Double), Double](_._2) |
| 51 | + val limitedCollection = new mutable.TreeSet[(String, Double)]() |
| 52 | + |
| 53 | + records.foreach { record => |
| 54 | + limitedCollection.add(record) |
| 55 | + if (limitedCollection.size > LIMIT) { |
| 56 | + limitedCollection.remove(limitedCollection.last) |
| 57 | + } |
| 58 | + } |
| 59 | + |
| 60 | + // I've traversed the iterator |
| 61 | + |
| 62 | + limitedCollection.toIterator |
| 63 | + } |
| 64 | + |
| 65 | + val topMetrics = readMetrics() |
| 66 | + .mapPartitions(iteratorToIteratorTransformation) |
| 67 | + .repartition(1) |
| 68 | + .mapPartitions(iteratorToIteratorTransformation) |
| 69 | + |
| 70 | + val result = topMetrics.take(LIMIT) |
| 71 | + result.foreach(println) |
| 72 | + } |
| 73 | + |
| 74 | + def main(args: Array[String]): Unit = { |
| 75 | + printTopMetrics() |
| 76 | + printTopMetricsI2I() |
| 77 | + Thread.sleep(1000000) |
| 78 | + } |
| 79 | +} |
0 commit comments