hadoop - 从AVSC创建Hive表，其中包含对先前定义的架构的引用作为一种类型

Question

我正在寻找一种通过Hive获取以下AVSC文件内容并外部化嵌套模式“RENTALRECORDTYPE”的方法，以实现模式重用。

{
    "type": "record",
    "name": "EMPLOYEE",
    "namespace": "",
    "doc": "EMPLOYEE is a person that works here",
    "fields": [
        {
            "name": "RENTALRECORD",
            "type": {
                "type": "record",
                "name": "RENTALRECORDTYPE",
                "namespace": "",
                "doc": "Rental record is a record that is kept on every item rented",
                "fields": [
                    {
                        "name": "due_date",
                        "doc": "The date when item is due",
                        "type": "int"
                    } 
                ]
            }
        },
        {
            "name": "hire_date",
            "doc": "Employee date of hire",
            "type": "int"
        }
    ]
}

这种定义架构的方法很好用。我可以发出以下HiveQL语句，并且表已成功创建。

CREATE EXTERNAL TABLE employee
STORED AS AVRO
LOCATION '/user/dtom/store/data/employee'
TBLPROPERTIES ('avro.schema.url'='/user/dtom/store/schema/employee.avsc');

但是，我希望能够引用现有架构，而不是在多个架构中复制记录定义。例如，将生成两个AVSC文件，而不是单个模式文件。即rentalrecord.avsc和employee.avsc。

rentalrecord.avsc

{
    "type": "record",
    "name": "RENTALRECORD",
    "namespace": "",
    "doc": "A record that is kept for every rental",
    "fields": [
        {
            "name": "due_date",
            "doc": "The date on which the rental is due back to the store",
            "type": "int"
        }
    ]
}

employee.avsc

{
    "type": "record",
    "name": "EMPLOYEE",
    "namespace": "",
    "doc": "EMPLOYEE is a person that works for the VIDEO STORE",
    "fields": [
        {
            "name": "rentalrecord",
            "doc": "A rental record is a record on every rental",
            "type": "RENTALRECORD"
        },
        {
            "name": "hire_date",
            "doc": "Employee date of hire",
            "type": "int"
        }
    ]
}

在上述情况下，我们希望能够外部化 RENTALRECORD 模式定义，并能够在 employee.avsc 和其他地方重用它。

尝试使用以下两个HiveQL语句导入架构时，它将失败…

CREATE EXTERNAL TABLE rentalrecord
STORED AS AVRO
LOCATION '/user/dtom/store/data/rentalrecord'
TBLPROPERTIES ('avro.schema.url'='/user/dtom/store/schema /rentalrecord.avsc');

CREATE EXTERNAL TABLE employee
STORED AS AVRO
LOCATION '/user/dtom/store/data/employee'
TBLPROPERTIES ('avro.schema.url'='/user/dtom/store/schema/employee.avsc');

Rentalrecord.avsc已成功导入，但是employee.avsc在第一个字段定义上失败。类型为“RENTALRECORD”的字段。 Hive输出以下错误……

FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.DDLTask. java.lang.RuntimeException: MetaException(message:org.apache.hadoop.hive.serde2.SerDeException Encountered exception determining schema. Returning signal schema to indicate problem: "RENTALRECORD" is not a defined name. The type of the "rentalrecord" field must be a defined name or a {"type": ...} expression.)

我的研究告诉我，Avro文件确实支持这种形式的模式重用。所以我丢失了某些东西，或者这是Hive不支持的东西。

任何帮助将不胜感激。

Answer 1

我已经定义了带有所有引用的AVDL，然后使用带有idl2schemata选项的avro工具jar文件来生成avsc。生成的avsc像 hive 一样吸引人!!