[SPARK-19436][SQL] Add missing tests for approxQuantile

## What changes were proposed in this pull request?
1, check the behavior with illegal `quantiles` and `relativeError`
2, add tests for `relativeError` > 1
3, update tests for `null` data
4, update some docs for javadoc8

## How was this patch tested?
local test in spark-shell

Author: Zheng RuiFeng <ruifengz@foxmail.com>
Author: Ruifeng Zheng <ruifengz@foxmail.com>

Closes #16776 from zhengruifeng/fix_approxQuantile.

1 files changed, 68 insertions, 9 deletions

diff --git a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
index f52b18e27b..d0910e618a 100644
--- a/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
+++ b/sql/core/src/test/scala/org/apache/spark/sql/DataFrameStatSuite.scala
@@ -25,7 +25,7 @@ import org.apache.spark.internal.Logging
 import org.apache.spark.sql.execution.stat.StatFunctions
 import org.apache.spark.sql.functions.col
 import org.apache.spark.sql.test.SharedSQLContext
-import org.apache.spark.sql.types.DoubleType
+import org.apache.spark.sql.types.{DoubleType, StructField, StructType}
 
 class DataFrameStatSuite extends QueryTest with SharedSQLContext {
   import testImplicits._
@@ -159,16 +159,75 @@ class DataFrameStatSuite extends QueryTest with SharedSQLContext {
       assert(math.abs(md1 - 2 * q1 * n) < error_double)
       assert(math.abs(md2 - 2 * q2 * n) < error_double)
     }
-    // test approxQuantile on NaN values
-    val dfNaN = Seq(Double.NaN, 1.0, Double.NaN, Double.NaN).toDF("input")
-    val resNaN = dfNaN.stat.approxQuantile("input", Array(q1, q2), epsilons.head)
+
+    // quantile should be in the range [0.0, 1.0]
+    val e = intercept[IllegalArgumentException] {
+      df.stat.approxQuantile(Array("singles", "doubles"), Array(q1, q2, -0.1), epsilons.head)
+    }
+    assert(e.getMessage.contains("quantile should be in the range [0.0, 1.0]"))
+
+    // relativeError should be non-negative
+    val e2 = intercept[IllegalArgumentException] {
+      df.stat.approxQuantile(Array("singles", "doubles"), Array(q1, q2), -1.0)
+    }
+    assert(e2.getMessage.contains("Relative Error must be non-negative"))
+
+    // return null if the dataset is empty
+    val res1 = df.selectExpr("*").limit(0)
+      .stat.approxQuantile("singles", Array(q1, q2), epsilons.head)
+    assert(res1 === null)
+
+    val res2 = df.selectExpr("*").limit(0)
+      .stat.approxQuantile(Array("singles", "doubles"), Array(q1, q2), epsilons.head)
+    assert(res2 === null)
+  }
+
+  test("approximate quantile 2: test relativeError greater than 1 return the same result as 1") {
+    val n = 1000
+    val df = Seq.tabulate(n)(i => (i, 2.0 * i)).toDF("singles", "doubles")
+
+    val q1 = 0.5
+    val q2 = 0.8
+    val epsilons = List(2.0, 5.0, 100.0)
+
+    val Array(single1_1) = df.stat.approxQuantile("singles", Array(q1), 1.0)
+    val Array(s1_1, s2_1) = df.stat.approxQuantile("singles", Array(q1, q2), 1.0)
+    val Array(Array(ms1_1, ms2_1), Array(md1_1, md2_1)) =
+      df.stat.approxQuantile(Array("singles", "doubles"), Array(q1, q2), 1.0)
+
+    for (epsilon <- epsilons) {
+      val Array(single1) = df.stat.approxQuantile("singles", Array(q1), epsilon)
+      val Array(s1, s2) = df.stat.approxQuantile("singles", Array(q1, q2), epsilon)
+      val Array(Array(ms1, ms2), Array(md1, md2)) =
+        df.stat.approxQuantile(Array("singles", "doubles"), Array(q1, q2), epsilon)
+      assert(single1_1 === single1)
+      assert(s1_1 === s1)
+      assert(s2_1 === s2)
+      assert(ms1_1 === ms1)
+      assert(ms2_1 === ms2)
+      assert(md1_1 === md1)
+      assert(md2_1 === md2)
+    }
+  }
+
+  test("approximate quantile 3: test on NaN and null values") {
+    val q1 = 0.5
+    val q2 = 0.8
+    val epsilon = 0.1
+    val rows = spark.sparkContext.parallelize(Seq(Row(Double.NaN, 1.0), Row(1.0, 1.0),
+      Row(-1.0, Double.NaN), Row(Double.NaN, Double.NaN), Row(null, null), Row(null, 1.0),
+      Row(-1.0, null), Row(Double.NaN, null)))
+    val schema = StructType(Seq(StructField("input1", DoubleType, nullable = true),
+      StructField("input2", DoubleType, nullable = true)))
+    val dfNaN = spark.createDataFrame(rows, schema)
+    val resNaN = dfNaN.stat.approxQuantile("input1", Array(q1, q2), epsilon)
     assert(resNaN.count(_.isNaN) === 0)
-    // test approxQuantile on multi-column NaN values
-    val dfNaN2 = Seq((Double.NaN, 1.0), (1.0, 1.0), (-1.0, Double.NaN), (Double.NaN, Double.NaN))
-      .toDF("input1", "input2")
-    val resNaN2 = dfNaN2.stat.approxQuantile(Array("input1", "input2"),
-      Array(q1, q2), epsilons.head)
+    assert(resNaN.count(_ == null) === 0)
+
+    val resNaN2 = dfNaN.stat.approxQuantile(Array("input1", "input2"),
+      Array(q1, q2), epsilon)
     assert(resNaN2.flatten.count(_.isNaN) === 0)
+    assert(resNaN2.flatten.count(_ == null) === 0)
   }
 
   test("crosstab") {