I have a spark parquet table on disk with the following schema :
id1 : STRING weight: array (nullable = true) |-- element: struct (containsNull = true)