Sort MemSQL partitions by ordinal in MemSQLRDD getPartitions()

Summary:
As of MemSQL 6, "SHOW PARTITIONS ON `${dbName}`" no longer guarantees that the result set is in sorted order by ordinal, so users were hitting error messages like `requirement failed: partitions(0).partition == 21, but it should equal 0 ` (https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/rdd/RDD.scala#L2 )

This diff causes MemSQLRDD getPartitions() to sort the partition array by index before returning it.

Test Plan: sbt test

Reviewers: carl, jdkaplan

Reviewed By: jdkaplan

Subscribers: engineering-list

JIRA Issues: AP-1693

Differential Revision: https://grizzly.memsql.com/D24036

CHANGELOG

@@ -1,4 +1,7 @@
-2017-8-31 Version 2.0.4
+2018-01-18 Version 2.0.5
+        * Explicitly sort MemSQLRDD partitions due to MemSQL 6.0 no longer returning partitions in sorted order by ordinal.
+
+2017-08-31 Version 2.0.4
         * Switch threads in LoadDataStrategy so that the parent thread reads from the RDD and the new thread writes
           to MemSQL so that Spark has access to the thread-local variables it expects
 

README.md

@@ -21,7 +21,7 @@ Installation
 Inside a project definition you can depend on the MemSQL Connector using sbt:
 
 ```
-libraryDependencies  += "com.memsql" %% "memsql-connector" % "2.0.4"
+libraryDependencies  += "com.memsql" %% "memsql-connector" % "2.0.5"
 ```
 
 or Maven:

build.sbt

@@ -10,7 +10,7 @@ lazy val testScalastyle = taskKey[Unit]("testScalastyle")
 
 lazy val commonSettings = Seq(
   organization := "com.memsql",
-  version := "2.0.4",
+  version := "2.0.5",
   scalaVersion := "2.11.8",
   assemblyScalastyle := org.scalastyle.sbt.ScalastylePlugin.scalastyle.in(Compile).toTask("").value,
   assembly <<= assembly dependsOn assemblyScalastyle,

src/main/scala/com/memsql/spark/connector/rdd/MemSQLRDD.scala

@@ -80,7 +80,7 @@ case class MemSQLRDD[T: ClassTag](@transient sc: SparkContext,
 
                       new MemSQLRDDPartition(ordinal, connInfo, Some(partitionQuery))
                     })
-                    .toArray
+                    .toArray.sortBy(_.index).asInstanceOf[Array[Partition]]
                 }
                 case None => {
                   getSinglePartition