Github user bersprockets commented on a diff in the pull request:

    https://github.com/apache/spark/pull/21073#discussion_r186570491
  
    --- Diff: 
sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/collectionOperations.scala
 ---
    @@ -116,6 +117,169 @@ case class MapValues(child: Expression)
       override def prettyName: String = "map_values"
     }
     
    +/**
    + * Returns the union of all the given maps.
    + */
    +@ExpressionDescription(
    +usage = "_FUNC_(map, ...) - Returns the union of all the given maps",
    +examples = """
    +    Examples:
    +      > SELECT _FUNC_(map(1, 'a', 2, 'b'), map(2, 'c', 3, 'd'));
    +       [[1 -> "a"], [2 -> "c"], [3 -> "d"]
    +  """, since = "2.4.0")
    +case class MapConcat(children: Seq[Expression]) extends Expression {
    +
    +  override def checkInputDataTypes(): TypeCheckResult = {
    +    // check key types and value types separately to allow 
valueContainsNull to vary
    +    if (children.exists(!_.dataType.isInstanceOf[MapType])) {
    +      TypeCheckResult.TypeCheckFailure(
    +        s"The given input of function $prettyName should all be of type 
map, " +
    +          "but they are " + 
children.map(_.dataType.simpleString).mkString("[", ", ", "]"))
    +    } else if (children.map(_.dataType.asInstanceOf[MapType].keyType)
    +      .exists(_.isInstanceOf[MapType])) {
    +      // map_concat needs to pick a winner when multiple maps contain the 
same key. map_concat
    +      // can do that only if it can detect when two keys are the same. 
SPARK-9415 states "map type
    +      // should not support equality, hash". As a result, map_concat does 
not support a map type
    +      // as a key
    +      TypeCheckResult.TypeCheckFailure(
    +        s"The given input maps of function $prettyName cannot have a map 
type as a key")
    +    } else if 
(children.map(_.dataType.asInstanceOf[MapType].keyType).distinct.length > 1) {
    +      TypeCheckResult.TypeCheckFailure(
    +        s"The given input maps of function $prettyName should all be the 
same type, " +
    +          "but they are " + 
children.map(_.dataType.simpleString).mkString("[", ", ", "]"))
    +    } else if 
(children.map(_.dataType.asInstanceOf[MapType].valueType).distinct.length > 1) {
    +      TypeCheckResult.TypeCheckFailure(
    +        s"The given input maps of function $prettyName should all be the 
same type, " +
    +          "but they are " + 
children.map(_.dataType.simpleString).mkString("[", ", ", "]"))
    +    } else {
    +      TypeCheckResult.TypeCheckSuccess
    +    }
    +  }
    +
    +  override def dataType: MapType = {
    +    MapType(
    +      keyType = children.headOption
    +        
.map(_.dataType.asInstanceOf[MapType].keyType).getOrElse(StringType),
    +      valueType = children.headOption
    +        
.map(_.dataType.asInstanceOf[MapType].valueType).getOrElse(StringType),
    +      valueContainsNull = children.map { c =>
    +        c.dataType.asInstanceOf[MapType]
    +      }.exists(_.valueContainsNull)
    +    )
    +  }
    +
    +  override def nullable: Boolean = children.exists(_.nullable)
    +
    +  override def eval(input: InternalRow): Any = {
    +    val union = new util.LinkedHashMap[Any, Any]()
    +    children.map(_.eval(input)).foreach { raw =>
    +      if (raw == null) {
    +        return null
    +      }
    +      val map = raw.asInstanceOf[MapData]
    +      map.foreach(dataType.keyType, dataType.valueType, (k, v) =>
    +        union.put(k, v)
    +      )
    +    }
    +    val (keyArray, valueArray) = union.entrySet().toArray().map { e =>
    --- End diff --
    
    I would imagine bad things would happen before you got this far (even Map's 
size method returns an Int).


---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to