1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
|
package spark
import org.scalatest.FunSuite
import org.scalatest.BeforeAndAfter
import org.scalatest.matchers.ShouldMatchers
import SparkContext._
class SortingSuite extends FunSuite with LocalSparkContext with ShouldMatchers with Logging {
test("sortByKey") {
sc = new SparkContext("local", "test")
val pairs = sc.parallelize(Array((1, 0), (2, 0), (0, 0), (3, 0)), 2)
assert(pairs.sortByKey().collect() === Array((0,0), (1,0), (2,0), (3,0)))
}
test("large array") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 2)
val sorted = pairs.sortByKey()
assert(sorted.splits.size === 2)
assert(sorted.collect() === pairArr.sortBy(_._1))
}
test("large array with one split") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 2)
val sorted = pairs.sortByKey(true, 1)
assert(sorted.splits.size === 1)
assert(sorted.collect() === pairArr.sortBy(_._1))
}
test("large array with many splits") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 2)
val sorted = pairs.sortByKey(true, 20)
assert(sorted.splits.size === 20)
assert(sorted.collect() === pairArr.sortBy(_._1))
}
test("sort descending") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 2)
assert(pairs.sortByKey(false).collect() === pairArr.sortWith((x, y) => x._1 > y._1))
}
test("sort descending with one split") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 1)
assert(pairs.sortByKey(false, 1).collect() === pairArr.sortWith((x, y) => x._1 > y._1))
}
test("sort descending with many splits") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(1000) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 2)
assert(pairs.sortByKey(false, 20).collect() === pairArr.sortWith((x, y) => x._1 > y._1))
}
test("more partitions than elements") {
sc = new SparkContext("local", "test")
val rand = new scala.util.Random()
val pairArr = Array.fill(10) { (rand.nextInt(), rand.nextInt()) }
val pairs = sc.parallelize(pairArr, 30)
assert(pairs.sortByKey().collect() === pairArr.sortBy(_._1))
}
test("empty RDD") {
sc = new SparkContext("local", "test")
val pairArr = new Array[(Int, Int)](0)
val pairs = sc.parallelize(pairArr, 2)
assert(pairs.sortByKey().collect() === pairArr.sortBy(_._1))
}
test("partition balancing") {
sc = new SparkContext("local", "test")
val pairArr = (1 to 1000).map(x => (x, x)).toArray
val sorted = sc.parallelize(pairArr, 4).sortByKey()
assert(sorted.collect() === pairArr.sortBy(_._1))
val partitions = sorted.collectPartitions()
logInfo("Partition lengths: " + partitions.map(_.length).mkString(", "))
partitions(0).length should be > 180
partitions(1).length should be > 180
partitions(2).length should be > 180
partitions(3).length should be > 180
partitions(0).last should be < partitions(1).head
partitions(1).last should be < partitions(2).head
partitions(2).last should be < partitions(3).head
}
test("partition balancing for descending sort") {
sc = new SparkContext("local", "test")
val pairArr = (1 to 1000).map(x => (x, x)).toArray
val sorted = sc.parallelize(pairArr, 4).sortByKey(false)
assert(sorted.collect() === pairArr.sortBy(_._1).reverse)
val partitions = sorted.collectPartitions()
logInfo("partition lengths: " + partitions.map(_.length).mkString(", "))
partitions(0).length should be > 180
partitions(1).length should be > 180
partitions(2).length should be > 180
partitions(3).length should be > 180
partitions(0).last should be > partitions(1).head
partitions(1).last should be > partitions(2).head
partitions(2).last should be > partitions(3).head
}
}
|