Spark HiveContext -从外部分区配置单元表分隔符读取问题

Question

我有一个外部分区Hive表，其下划线文件行格式的分隔字段以'|‘结尾，通过Hive直接读取数据是可以的，但当使用Spark的Dataframe API时，分隔符'|’不被考虑。

创建外部分区表：

hive> create external table external_delimited_table(value1 string, value2 string)
partitioned by (year string, month string, day string) ROW FORMAT DELIMITED FIELDS TERMINATED BY '|'
location '/client/edb/poc_database/external_delimited_table';

创建只包含一行的数据文件，并将其放到外部分区表位置：

shell>echo "one|two" >> table_data.csv
shell>hadoop fs -mkdir -p /client/edb/poc_database/external_delimited_table/year=2016/month=08/day=20
shell>hadoop fs -copyFromLocal table_data.csv /client/edb/poc_database/external_delimited_table/year=2016/month=08/day=20

使分区处于活动状态：

hive> alter table external_delimited_table add partition (year='2016',month='08',day='20');

健全性检查：

hive> select * from external_delimited_table;
select * from external_delimited_table;
+----------------------------------+----------------------------------+--------------------------------+---------------------------------+-------------------------------+--+
| external_delimited_table.value1  | external_delimited_table.value2  | external_delimited_table.year  | external_delimited_table.month  | external_delimited_table.day  |
+----------------------------------+----------------------------------+--------------------------------+---------------------------------+-------------------------------+--+
| one                              | two                              | 2016                           | 08                              | 20 

Spark代码：

import org.apache.spark.sql.DataFrame
import org.apache.spark.sql.hive.HiveContext
import org.apache.spark.{SparkContext, SparkConf}
object TestHiveContext {

  def main(args: Array[String]): Unit = {

    val conf = new SparkConf().setAppName("Test Hive Context")

    val spark = new SparkContext(conf)
    val hiveContext  = new HiveContext(spark)

    val dataFrame: DataFrame = hiveContext.sql("SELECT * FROM external_delimited_table")
    dataFrame.show()

    spark.stop()
  }

dataFrame.show()输出：

+-------+------+----+-----+---+
| value1|value2|year|month|day|
+-------+------+----+-----+---+
|one|two|  null|2016|   08| 20|
+-------+------+----+-----+---+

Answer 1

这被证明是Spark版本1.5.0的一个问题。在版本1.6.0中不会发生问题：

scala> sqlContext.sql("select * from external_delimited_table")
res2: org.apache.spark.sql.DataFrame = [value1: string, value2: string, year: string, month: string, day: string]

scala> res2.show
+------+------+----+-----+---+
|value1|value2|year|month|day|
+------+------+----+-----+---+
|   one|   two|2016|   08| 20|
+------+------+----+-----+---+