manishnalla1994 commented on a change in pull request #3182: [CARBONDATA-3343]
Compaction for Range Sort
URL: https://github.com/apache/carbondata/pull/3182#discussion_r280737189
##########
File path:
integration/spark-common/src/main/scala/org/apache/carbondata/spark/rdd/CarbonMergerRDD.scala
##########
@@ -589,9 +601,91 @@ class CarbonMergerRDD[K, V](
// Change string type to support all types
val sampleRdd = scanRdd
.map(row => (row.get(0, sparkDataType), null))
+ val sortedRdd = sampleRdd.sortBy(_._1, true)(objectOrdering,
classTag[AnyRef])
val value = new DataSkewRangePartitioner(
- defaultParallelism, sampleRdd, true)(objectOrdering, classTag[Object])
- CarbonCompactionUtil.getRangesFromVals(value.rangeBounds, value.minMaxVals)
+ defaultParallelism, sortedRdd, true)(objectOrdering, classTag[Object])
+ if(minVal == null && maxVal == null) {
+ CarbonCompactionUtil
+ .getRangesFromVals(value.rangeBounds, value.rangeBounds(0),
value.rangeBounds(1))
+ } else {
+ CarbonCompactionUtil.getRangesFromVals(value.rangeBounds, minVal, maxVal)
+ }
+ }
+
+ def getOverallMinMax(carbonInputSplits: mutable.Seq[CarbonInputSplit],
Review comment:
It is not giving exact values for min/max in case of big data. For example,
in case for 1-10000, RangePartitioner was storing only a few intermediate
values in its ordered set in which the last value is not necessarily the max
value. So I had to take the overall min/max for the column.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
