fix UT

clockfly · clockfly · commit 3dda1ae90cea · 2016-08-30T17:10:34.000+08:00
diff --git a/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/ApproximatePercentile.scala b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/ApproximatePercentile.scala
@@ -91,8 +91,10 @@ case class ApproximatePercentile(
       // Rule ImplicitTypeCasts can cast other numeric types to double
       case (_, num: Double) => (false, Array(num))
       case (ArrayType(baseType: NumericType, _), arrayData: ArrayData) =>
-         val numericArray = arrayData.toArray(baseType)(baseType.classTag)
-        (true, numericArray.map(baseType.numeric.toDouble))
+         val numericArray = arrayData.toObjectArray(baseType)
+        (true, numericArray.map {x =>
+          baseType.numeric.toDouble(x.asInstanceOf[baseType.InternalType])
+        })
       case other =>
         throw new AnalysisException(s"Invalid data type ${other._1} for parameter percentage")
     }
@@ -183,8 +185,8 @@ object ApproximatePercentile {
    *                   underlying quantileSummaries is compressed.
    */
   class PercentileDigest(
-    private var summaries: QuantileSummaries,
-    private var isCompressed: Boolean) {
+      private var summaries: QuantileSummaries,
+      private var isCompressed: Boolean) {
 
     // Trigger compression if the QuantileSummaries's buffer length exceeds
     // compressThresHoldBufferLength. The buffer length can be get by
@@ -287,10 +289,10 @@ object ApproximatePercentile {
   // An case class to wrap fields of QuantileSummaries, so that we can use the expression encoder
   // to serialize it.
   case class QuantileSummariesData(
-      val compressThreshold: Int,
-      val relativeError: Double,
-      val sampled: Array[Stats] = Array.empty,
-      val count: Long = 0L) {
+      compressThreshold: Int,
+      relativeError: Double,
+      sampled: Array[Stats],
+      count: Long) {
     def this(summary: QuantileSummaries) = {
       this(summary.compressThreshold, summary.relativeError, summary.sampled, summary.count)
     }
diff --git a/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/ExpressionToSQLSuite.scala b/sql/hive/src/test/scala/org/apache/spark/sql/catalyst/ExpressionToSQLSuite.scala
@@ -155,7 +155,11 @@ class ExpressionToSQLSuite extends SQLBuilderTest with SQLTestUtils {
 
   test("aggregate functions") {
     checkSqlGeneration("SELECT approx_count_distinct(value) FROM t1 GROUP BY key")
-    checkSqlGeneration("SELECT percentile_approx(value) FROM t1 GROUP BY key")
+    checkSqlGeneration("SELECT percentile_approx(value, 0.25) FROM t1 GROUP BY key")
+    checkSqlGeneration("SELECT percentile_approx(value, array(0.25, 0.75)) FROM t1 GROUP BY key")
+    checkSqlGeneration("SELECT percentile_approx(value, 0.25, 100) FROM t1 GROUP BY key")
+    checkSqlGeneration(
+      "SELECT percentile_approx(value, array(0.25, 0.75), 100) FROM t1 GROUP BY key")
     checkSqlGeneration("SELECT avg(value) FROM t1 GROUP BY key")
     checkSqlGeneration("SELECT corr(value, key) FROM t1 GROUP BY key")
     checkSqlGeneration("SELECT count(value) FROM t1 GROUP BY key")
