AWS Glue读取JSON时trace数组格式异常，如何还原原始格式？

我获取了一份ADS-B数据的JSON文件，上传至S3后使用AWS Glue读取，遇到以下问题：

原始JSON内容 #

{
  "icao": "f3b100",
  "noRegData": true,
  "timestamp": 1690848000,
  "trace": [
    [
      51213.77,
      39.090519,
      -85.870361,
      4500,
      95,
      359.4,
      5,
      0,
      {
        "type": "tisb_icao",
        "track": 359.4,
        "geom_rate": 0,
        "nic": 0,
        "rc": 0,
        "nac_p": 0,
        "nac_v": 0,
        "sil": 0,
        "sil_type": "unknown",
        "alert": 0,
        "spi": 0
      },
      "tisb_icao",
      null,
      0,
      null,
      null,
      "f9adebf3"
    ]
  ]
}

AWS Glue解析后的trace格式 #

[
  [
    {
      "double": 51213.77,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": 39.090519,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": -85.870361,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": 4500,
      "string": null,
      "struct": null
    },
    {
      "double": 95,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": 359.4,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": 5,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": 0,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": null,
      "string": null,
      "struct": {
        "type": "tisb_icao",
        "alt_geom": null,
        "track": 359.4,
        "geom_rate": 0,
        "nic": 0,
        "rc": 0,
        "nac_p": 0,
        "nac_v": 0,
        "sil": 0,
        "sil_type": "unknown",
        "alert": 0,
        "spi": 0
      }
    },
    {
      "double": null,
      "int": null,
      "string": "tisb_icao",
      "struct": null
    },
    {
      "double": null,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": 0,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": null,
      "string": null,
      "struct": null
    },
    {
      "double": null,
      "int": null,
      "string": "f9adebf3",
      "struct": null
    }
  ]
]

解析对应的Schema #

array<array<struct<double:double,int:int,string:string,struct:struct<type:string,track:double,geom_rate:int,nic:int,rc:int,nac_p:int,nac_v:int,sil:int,sil_type:string,alert:int,spi:int>>>>

疑问 #

• 这种将原始混合类型数组转换为带类型标记结构体的行为，是AWS Glue的预期行为吗？
• 如何配置AWS Glue，才能正确读取并保留原始trace数组的格式？

内容的提问来源于stack exchange，提问作者Hongbo Miao