如何编写包含数组的数组字段的Avro Schema？

问题根源 #

当前写法的核心错误有两处：

1. 内层数组的类型定义多套了一层多余的type对象包装
2. default配置的层级不符合Avro语法规范

修正后完整Schema #

{
    "name": "SelfHealingStarter",
    "namespace": "SCP.Kafka.AvroSchemas",
    "doc": "Message with all the necessary information to run Self Healing process.",
    "type": "record",
    "fields": [
        {
            "name": "FiveMinutesAgoMeasurement",
            "type": "record",
            "doc": "Field with all five minutes ago measurement.",
            "fields": [
                {
                    "name": "equipments",
                    "doc": "List with all equipments measurement, each entry corresponds to a string array of single equipment measurement data.",
                    "type": {
                        "type": "array",
                        "items": {
                            "type": "array",
                            "items": "string",
                            "default": []
                        }
                    },
                    "default": []
                }
            ]
        }
    ]
}

核心调整说明 #

• 移除了内层数组外多余的type嵌套：外层数组的items属性直接接收内层数组的类型定义即可，不需要额外包裹{"type": XXX}结构
• 内层数组的default配置放在内层数组类型定义的同级，符合Avro类型属性的语法要求
• 修正后Schema对应的数据结构为二维字符串数组，可完美匹配「数组的数组」的需求，两层default配置分别对应外层数组为空、内层子数组为空的默认 fallback 值

内容的提问来源于stack exchange，提问作者Felipe Thomazi