下面的代码是 scala ，您可以将此转换为 python .

scala> df.show(false)
+--------------------------------+--------------------------------+
|2018-10-09                      |2018-11-30                      |
+--------------------------------+--------------------------------+
|[1.4861, 129.45, 1.7784, 1.1435]|[1.5116, 128.99, 1.6556, 1.1359]|
+--------------------------------+--------------------------------+

scala> val columns = df.columns

scala> 
val colExp = 
array(
    columns
    .map(c => 
        List(
            lit(c).as("date"),
            col(s"${c}.*")
        )
    )
    .map(l => struct(l:_*)):_*
)

scala> df.select(explode(colExp).as("data")).select("data.*").show(false)
+----------+------+------+------+------+
|date      |CAD   |JPY   |NZD   |USD   |
+----------+------+------+------+------+
|2018-10-09|1.4861|129.45|1.7784|1.1435|
|2018-11-30|1.5116|128.99|1.6556|1.1359|
+----------+------+------+------+------+

解释

val colExp = 
array(
    columns
    .map(c => List(lit(c).as("date"),col(s"${c}.*"))) // This one will create `2018-10-09` as `date`, `2018-10-09.*`
    .map(l => struct(l:_*)):_* // This will add above columns inside struct(_)
)

上面的代码会给你下面的输出。

array(
    struct(
        lit("2018-10-09").as(`date`), 
        col("2018-10-09.*")
    ), 
    struct(
        lit("2018-11-30").as(`date`), 
        col("2018-11-30.*")
    )
)

一旦你有了以上的输出，你就很容易做到 explode 数组提取（&E） struct 柱。