[SPARK-30082][SQL] Do not replace Zeros when replacing NaNs

Do not cast `NaN` to an `Integer`, `Long`, `Short` or `Byte`. This is because casting `NaN` to those types results in a `0` which erroneously replaces `0`s while only `NaN`s should be replaced.

This Scala code snippet:
```
import scala.math;

println(Double.NaN.toLong)
```
returns `0` which is problematic as if you run the following Spark code, `0`s get replaced as well:
```
>>> df = spark.createDataFrame([(1.0, 0), (0.0, 3), (float('nan'), 0)], ("index", "value"))
>>> df.show()
+-----+-----+
|index|value|
+-----+-----+
|  1.0|    0|
|  0.0|    3|
|  NaN|    0|
+-----+-----+
>>> df.replace(float('nan'), 2).show()
+-----+-----+
|index|value|
+-----+-----+
|  1.0|    2|
|  0.0|    3|
|  2.0|    2|
+-----+-----+
```

Yes, after the PR, running the same above code snippet returns the correct expected results:
```
>>> df = spark.createDataFrame([(1.0, 0), (0.0, 3), (float('nan'), 0)], ("index", "value"))
>>> df.show()
+-----+-----+
|index|value|
+-----+-----+
|  1.0|    0|
|  0.0|    3|
|  NaN|    0|
+-----+-----+

>>> df.replace(float('nan'), 2).show()
+-----+-----+
|index|value|
+-----+-----+
|  1.0|    0|
|  0.0|    3|
|  2.0|    0|
+-----+-----+
```

Added unit tests to verify replacing `NaN` only affects columns of type `Float` and `Double`

Closes apache#26738 from johnhany97/SPARK-30082.

Lead-authored-by: John Ayad <[email protected]>
Co-authored-by: John Ayad <[email protected]>
Signed-off-by: Wenchen Fan <[email protected]>

Author: johnhany97

sql/core/src/main/scala/org/apache/spark/sql/DataFrameNaFunctions.scala

@@ -455,7 +455,7 @@ final class DataFrameNaFunctions private[sql](df: DataFrame) {
     val keyExpr = df.col(col.name).expr
     def buildExpr(v: Any) = Cast(Literal(v), keyExpr.dataType)
     val branches = replacementMap.flatMap { case (source, target) =>
-      Seq(buildExpr(source), buildExpr(target))
+      Seq(Literal(source), buildExpr(target))
     }.toSeq
     new Column(CaseKeyWhen(keyExpr, branches :+ keyExpr)).as(col.name)
   }

sql/core/src/test/scala/org/apache/spark/sql/DataFrameNaFunctionsSuite.scala

@@ -36,6 +36,14 @@ class DataFrameNaFunctionsSuite extends QueryTest with SharedSQLContext {
       ).toDF("name", "age", "height")
   }
 
+  def createNaNDF(): DataFrame = {
+    Seq[(java.lang.Integer, java.lang.Long, java.lang.Short,
+      java.lang.Byte, java.lang.Float, java.lang.Double)](
+      (1, 1L, 1.toShort, 1.toByte, 1.0f, 1.0),
+      (0, 0L, 0.toShort, 0.toByte, Float.NaN, Double.NaN)
+    ).toDF("int", "long", "short", "byte", "float", "double")
+  }
+
   test("drop") {
     val input = createDF()
     val rows = input.collect()
@@ -305,4 +313,40 @@ class DataFrameNaFunctionsSuite extends QueryTest with SharedSQLContext {
       )).na.drop("name" :: Nil).select("name"),
       Row("Alice") :: Row("David") :: Nil)
   }
+
+  test("replace nan with float") {
+    checkAnswer(
+      createNaNDF().na.replace("*", Map(
+        Float.NaN -> 10.0f
+      )),
+      Row(1, 1L, 1.toShort, 1.toByte, 1.0f, 1.0) ::
+      Row(0, 0L, 0.toShort, 0.toByte, 10.0f, 10.0) :: Nil)
+  }
+
+  test("replace nan with double") {
+    checkAnswer(
+      createNaNDF().na.replace("*", Map(
+        Double.NaN -> 10.0
+      )),
+      Row(1, 1L, 1.toShort, 1.toByte, 1.0f, 1.0) ::
+      Row(0, 0L, 0.toShort, 0.toByte, 10.0f, 10.0) :: Nil)
+  }
+
+  test("replace float with nan") {
+    checkAnswer(
+      createNaNDF().na.replace("*", Map(
+        1.0f -> Float.NaN
+      )),
+      Row(0, 0L, 0.toShort, 0.toByte, Float.NaN, Double.NaN) ::
+      Row(0, 0L, 0.toShort, 0.toByte, Float.NaN, Double.NaN) :: Nil)
+  }
+
+  test("replace double with nan") {
+    checkAnswer(
+      createNaNDF().na.replace("*", Map(
+        1.0 -> Double.NaN
+      )),
+      Row(0, 0L, 0.toShort, 0.toByte, Float.NaN, Double.NaN) ::
+      Row(0, 0L, 0.toShort, 0.toByte, Float.NaN, Double.NaN) :: Nil)
+  }
 }