[TAJO-976] HashPartitioner doesn't make desired number of partitions infrequently.

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Assignee: Hyoungjun Kim
• Reporter: Hyoungjun Kim
• Created: 24/Jul/14 08:42
• Updated: 29/Jul/14 16:08
• Resolved: 29/Jul/14 04:27

Description

If the number desired partition is a multiples of 31, HashPartitioner doesn't make desired number of partitions. I tested with the following code.

Random rand = new Random();
String[][] data = new String[1000][];

for (int i = 0; i < 1000; i++) {
  data[i] = new String[]{ String.valueOf(rand.nextInt(1000)), String.valueOf(rand.nextInt(1000)), String.valueOf(rand.nextInt(1000))};
}

int[] testNumPartitions = new int[]{31, 62, 124, 32, 63, 125};
for (int index = 0; index <  testNumPartitions.length; index++) {
  Set<Integer> ids = new TreeSet<Integer>();
  for (int i = 0; i < data.length; i++) {
    VTuple vtuple = new VTuple(new Datum[]{new TextDatum(data[i][0]), new TextDatum(data[i][1]), new TextDatum(data[i][2])});

    int hashValue = vtuple.hashCode();
    int partitionId = (hashValue & Integer.MAX_VALUE) % testNumPartitions[index];
    ids.add(partitionId);
  }

  System.out.println("Expected partition num: " + testNumPartitions[index] + ", real partition num:" + ids.size());
}

Expected partition num: 31, real partition num:10
Expected partition num: 62, real partition num:20
Expected partition num: 124, real partition num:40
Expected partition num: 32, real partition num:32
Expected partition num: 63, real partition num:63
Expected partition num: 125, real partition num:125