Apache NiFi中使用Jolt Transform拆分记录求助

数据老陈

2026-8-18

用Jolt Transform实现数组按索引拆分生成多条记录

输入数据

{"resourceid": "d6315d4d7f0c","timestamp": [166406,166404,166504],"Key": ["mem","net","diskspace"],"data": [89,90,91]}

期望输出

[{"resourceid": "d6315d4d7f0c","timestamp": 166406,"Key": "mem","data": 89},{"resourceid": "d6315d4d7f0c","timestamp": 166404,"Key": "net","data": 90},{"resourceid": "d6315d4d7f0c","timestamp": 166504,"Key": "diskspace","data": 91}]

Jolt Transform Spec

通过shift和cardinality组合操作实现需求，具体配置如下：

[
  {
    "operation": "shift",
    "spec": {
      "timestamp": {
        "*": {
          "@(2,resourceid)": "[&1].resourceid",
          "@": "[&1].timestamp",
          "@(2,Key[&1])": "[&1].Key",
          "@(2,data[&1])": "[&1].data"
        }
      }
    }
  },
  {
    "operation": "cardinality",
    "spec": {
      "*": {
        "resourceid": "ONE",
        "timestamp": "ONE",
        "Key": "ONE",
        "data": "ONE"
      }
    }
  }
]

逻辑说明

shift操作：遍历timestamp数组的每个元素（*匹配数组索引），通过@(2,resourceid)向上两层获取根节点的resourceid值，映射到对应索引的对象中；@直接取当前timestamp元素值；@(2,Key[&1])和@(2,data[&1])分别提取同索引的Key、data数组元素，对应到每个对象的字段。
cardinality操作：强制每个字段为单值类型，避免shift操作生成数组形式的字段值，保证输出结构符合预期。