Question

给出具有以下格式的数据框：

{
    "field1": "value1",
    "field2": "value2",
    "elements": [{
        "id": "1",
        "name": "a"
    },
    {
        "id": "2",
        "name": "b"
    },
    {
        "id": "3",
        "name": "c"
    }]
}

我们可以像这样平整列：

val exploded = df.withColumn("elements", explode($"elements"))
exploded.show()
 >> +--------+------+------+
 >> |elements|field1|field2|
 >> +--------+------+------+
 >> |   [1,a]|value1|value2|
 >> |   [2,b]|value1|value2|
 >> |   [3,c]|value1|value2|
 >> +--------+------+------+
val flattened = exploded.select("elements.*", "field1", "field2")
flattened.show()
 >> +---+----+------+------+
 >> | id|name|field1|field2|
 >> +---+----+------+------+
 >> |  1|   a|value1|value2|
 >> |  2|   b|value1|value2|
 >> |  3|   c|value1|value2|
 >> +---+----+------+------+

是否有一种无需明确指定其余列即可获取展平数据帧的方法？这样的东西（尽管这不起作用）？

val flattened = exploded.select("elements.*", "*")

Answer 1

是的，您可以查询exploded的列，然后选择除elements之外的所有列：

import org.apache.spark.sql.functions.col

val colsToSelect = exploded.columns.filterNot(c => c == "elements").map(col)

val flattened = exploded.select(($"elements.*" +:colsToSelect):_*)

如何扁平化DataFrame中的列

1 个答案: