[MAPREDUCE-5569] FloatSplitter is not generating correct splits

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.0-beta, 1.3.0, 0.23.9
• Fix Version/s: 1.3.0, 0.23.10, 2.3.0
• Hadoop Flags: Reviewed
• Assignee: Nathan Roberts
• Reporter: Nathan Roberts
• Created: 07/Oct/13 16:34
• Updated: 10/Mar/15 04:30
• Resolved: 09/Oct/13 16:55

Description

The closing split is not calculated correctly:

     // Catch any overage and create the closed interval for the last split.
     if (curLower <= maxVal || splits.size() == 1) {
       splits.add(new DataDrivenDBInputFormat.DataDrivenDBInputSplit(
-          lowClausePrefix + Double.toString(curUpper),
+          lowClausePrefix + Double.toString(curLower),
           colName + " <= " + Double.toString(maxVal)));
     }

For the case of min=5.0, max=7.0, 2 splits, the current code returns splits of (column1 >=5.0, column1 <6.0), (column1 >=7.0, column1 <=7.0). The second split is obviously not correct.

Issue Links

• relates to: MAPREDUCE-5102 fix coverage org.apache.hadoop.mapreduce.lib.db and org.apache.hadoop.mapred.lib.db