python - NotImplementedError apache 梁 python

Question

我正在使用 apache 梁将 json 写入 gcs。但是遇到如下错误

NotImplementedError: offset: 0, whence: 0, position: 50547, last: 50547 [while running 'Writing new data to gcs/write data gcs/Write/WriteImpl/WriteBundles/WriteBundles']

不知道为什么会出现这个错误。相同的代码如下：

class WriteDataGCS(beam.PTransform):
        """
        To write data to GCS
        """

        def __init__(self, bucket):
            """
            Initiate the bucket as a class field

            :type bucket:string
            :param bucket: query to be run for data
            """
            self.bucket = bucket

        def expand(self, pcoll):
            """
            PTransform Method run when called on Class Name

            :type pcoll: PCollection
            :param pcoll: A pcollection
            """
            (pcoll | "print intermediate" >> beam.Map(print_row))
            return (pcoll | "write data gcs" >> beam.io.WriteToText(self.bucket, coder=JsonCoder(), file_name_suffix=".json"))

class JsonCoder:
    """
    This class represents dump and load operations performed on json
    """
    def encode(self,data):
        """
        Encodes the json data.

        :type data: string
        :param data: Data to be encoded
        """
        # logger.info("JSON DATA for encoding - {}".format(data))
        return json.dumps(data,default=str)

    def decode(self,data):
        """
        Decodes the json data.

        :type data: string
        :param data: Data to be decoded
        """
        # logger.info("JSON DATA for decoding - {}".format(data))
        return json.loads(data)

标签： pythonpython-2.7apache-beamapache-beam-io

Answer 1

的coder参数WriteToText需要一个apache_beam.coders.Coder实例。您可以尝试JsonCoder从基Coder类继承，但我认为您也可以使用 a 将数据转换为字符串Map：

        def expand(self, pcoll):
            """
            PTransform Method run when called on Class Name

            :type pcoll: PCollection
            :param pcoll: A pcollection
            """
            return (pcoll
              | "print intermediate" >> beam.Map(print_row))
              | "to_json" >> beam.Map(lambda x: json.dumps(x, default=str)))
              | "write data gcs" >> beam.io.WriteToText(self.bucket, file_name_suffix=".json"))