结构化流式处理，用于将 JSON 保存到 HDFS

Question

我的结构化Spark Streaming程序是从Kafka读取JSON数据并以 JSON 格式写入 HDFS。 我可以将JSON保存到HDFS，但是它保存 JSON 字符串：

 "jsontostructs(CAST(value AS STRING))"
key as below: {"jsontostructs(CAST(value AS STRING))":{"age":42,"name":"John"}}.

如何仅保存

{"age":42,"name":"John"}?


StructType schema = kafkaPrimerRow.schema();
//Read json from kafka. JSON is: {"age":42,"name":"John"}
Dataset<Row> df = spark
                    .readStream()
                    .format("kafka")
                    .option("kafka.bootstrap.servers", input_bootstrap_server)
                    .option("subscribe", topics[0])
                    .load();


    //Save Stream to HDFS
    StreamingQuery ds = df             
.select(functions.from_json(col("value").cast(DataTypes.StringType),schema)) 
.writeStream()
.format("json")
.outputMode(OutputMode.Append())
.option("path", destPath)
.option("checkpointLocation", checkpoint)
.start();

Answer 1

下面的 .select("data.*"( 做到了。

StreamingQuery ds = df
                        .select(functions.from_json(col("value").cast(DataTypes.StringType),schema).as("data"))
                        .select("data.*")
                        .writeStream()
                        .format("json")
                        .outputMode(OutputMode.Append())
                        .option("path", destPath)
                        .option("checkpointLocation", checkpoint)
                        .start();