Zhichao Zhang created CARBONDATA-1625:
------------------------------------------
Summary: Introduce new datatype of varchar(size) to store column
length more than short limit.
Key: CARBONDATA-1625
URL: https://issues.apache.org/jira/browse/CARBONDATA-1625
Project: CarbonData
Issue Type: New Feature
Components: file-format
Reporter: Zhichao Zhang
Priority: Minor
I am using Spark 2.1 + CarbonData 1.2, and find that if
enable.unsafe.sort=true, the length of bytes of column exceed 32768, it will
load data unsuccessfully.
My test code:
val longStr = sb.toString() // the getBytes length of longStr exceeds
32768
println(longStr.length())
println(longStr.getBytes("UTF-8").length)
import spark.implicits._
val df1 = spark.sparkContext.parallelize(0 to 1000)
.map(x => ("a", x.toString(), longStr, x, x.toLong, x * 2))
.toDF("stringField1", "stringField2", "stringField3", "intField",
"longField", "int2Field")
val df2 = spark.sparkContext.parallelize(1001 to 2000)
.map(x => ("b", x.toString(), (x % 2).toString(), x, x.toLong, x * 2))
.toDF("stringField1", "stringField2", "stringField3", "intField",
"longField", "int2Field")
val df3 = df1.union(df2)
val tableName = "study_carbondata_test"
spark.sql(s"DROP TABLE IF EXISTS ${tableName} ").show()
val sortScope = "LOCAL_SORT" // LOCAL_SORT GLOBAL_SORT
spark.sql(s"""
| CREATE TABLE IF NOT EXISTS ${tableName} (
| stringField1 string,
| stringField2 string,
| stringField3 string,
| intField int,
| longField bigint,
| int2Field int
| )
| STORED BY 'carbondata'
| TBLPROPERTIES('DICTIONARY_INCLUDE'='stringField1, stringField2',
| 'SORT_COLUMNS'='stringField1, stringField2, intField,
longField',
| 'SORT_SCOPE'='${sortScope}',
| 'NO_INVERTED_INDEX'='stringField3, int2Field',
| 'TABLE_BLOCKSIZE'='64'
| )
""".stripMargin)
df3.write
.format("carbondata")
.option("tableName", "study_carbondata_test")
.option("compress", "true") // just valid when tempCSV is true
.option("tempCSV", "false")
.option("single_pass", "true")
.mode(SaveMode.Append)
.save()
The error message:
*java.lang.NegativeArraySizeException
at
org.apache.carbondata.processing.newflow.sort.unsafe.UnsafeCarbonRowPage.getRow(UnsafeCarbonRowPage.java:182)
at
org.apache.carbondata.processing.newflow.sort.unsafe.holder.UnsafeInmemoryHolder.readRow(UnsafeInmemoryHolder.java:63)
at
org.apache.carbondata.processing.newflow.sort.unsafe.merger.UnsafeSingleThreadFinalSortFilesMerger.startSorting(UnsafeSingleThreadFinalSortFilesMerger.java:114)
at
org.apache.carbondata.processing.newflow.sort.unsafe.merger.UnsafeSingleThreadFinalSortFilesMerger.startFinalMerge(UnsafeSingleThreadFinalSortFilesMerger.java:81)
at
org.apache.carbondata.processing.newflow.sort.impl.UnsafeParallelReadMergeSorterImpl.sort(UnsafeParallelReadMergeSorterImpl.java:105)
at
org.apache.carbondata.processing.newflow.steps.SortProcessorStepImpl.execute(SortProcessorStepImpl.java:62)
at
org.apache.carbondata.processing.newflow.steps.DataWriterProcessorStepImpl.execute(DataWriterProcessorStepImpl.java:87)
at
org.apache.carbondata.processing.newflow.DataLoadExecutor.execute(DataLoadExecutor.java:51)
at
org.apache.carbondata.spark.rdd.NewDataFrameLoaderRDD$$anon$2.<init>(NewCarbonDataLoadRDD.scala:442)
at
org.apache.carbondata.spark.rdd.NewDataFrameLoaderRDD.internalCompute(NewCarbonDataLoadRDD.scala:405)
at
org.apache.carbondata.spark.rdd.CarbonRDD.compute(CarbonRDD.scala:62)
at org.apache.spark.rdd.RDD.computeOrReadCheckpoint(RDD.scala:323)
at org.apache.spark.rdd.RDD.iterator(RDD.scala:287)*
Currently, the length of column was stored by short type.
Introduce new datatype of varchar(size) to store column length more than short
limit.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
