avro架构中的可选阵列

Question

我想知道是否有可能有一个可选的数组。 让我们假设一个架构是这样的：

{ 
    "type": "record", 
    "name": "test_avro", 
    "fields" : [ 
     {"name": "test_field_1", "type": "long"}, 
     {"name": "subrecord", "type": [{ 
     "type": "record", 
     "name": "subrecord_type", 
      "fields":[{"name":"field_1", "type":"long"}] 
      },"null"] 
    }, 
    {"name": "simple_array", 
    "type":{ 
     "type": "array", 
     "items": "string" 
     } 
    } 
    ] 
} 

尝试写没有“simple_array”将导致在datafilewriter一个NPE的Avro的记录。 对于子记录它只是罚款，但是当我尝试了数组定义为可选：

{"name": "simple_array", 
"type":[{ 
    "type": "array", 
    "items": "string" 
    }, "null"] 

它不会导致NPE，但运行时异常：

AvroRuntimeException: Not an array schema: [{"type":"array","items":"string"},"null"] 

感谢。

Answer 1

我想你想要的这里是工会零和数组：

{ 
    "type":"record", 
    "name":"test_avro", 
    "fields":[{ 
      "name":"test_field_1", 
      "type":"long" 
     }, 
     { 
      "name":"subrecord", 
      "type":[{ 
        "type":"record", 
        "name":"subrecord_type", 
        "fields":[{ 
          "name":"field_1", 
          "type":"long" 
         } 
        ] 
       }, 
       "null" 
      ] 
     }, 
     { 
      "name":"simple_array", 
      "type":["null", 
       { 
        "type":"array", 
        "items":"string" 
       } 
      ], 
      "default":null 
     } 
    ] 
} 

当我使用与Python示例数据上面的架构，这里的结果（schema_string是上面的JSON字符串）：

>>> from avro import io, datafile, schema 
>>> from json import dumps 
>>> 
>>> sample_data = {'test_field_1':12L} 
>>> rec_schema = schema.parse(schema_string) 
>>> rec_writer = io.DatumWriter(rec_schema) 
>>> rec_reader = io.DatumReader() 
>>> 
>>> # write avro file 
... df_writer = datafile.DataFileWriter(open("/tmp/foo", 'wb'), rec_writer, writers_schema=rec_schema) 
>>> df_writer.append(sample_data) 
>>> df_writer.close() 
>>> 
>>> # read avro file 
... df_reader = datafile.DataFileReader(open('/tmp/foo', 'rb'), rec_reader) 
>>> print dumps(df_reader.next()) 
{"simple_array": null, "test_field_1": 12, "subrecord": null}