0

I'm trying to use the Splitter app to split a JSON array e.g. [{...},{...}] into multiple messages {...} {...}. With input contentType=application/json (per the docs) Spring Cloud is surfacing an exception from Jackson: 

com.fasterxml.jackson.databind.JsonMappingException: Can not deserialize instance of java.lang.String out of START_ARRAY token at [Source: [B@163b1945; line: 1, column: 1]

Unit tests showed I have the correct pattern for the split to work:

splitter.expression=#jsonPath(payload,'$.[*]')

This worked for me in Spring XD 1.3. How should Spring Cloud (or Splitter) be configured to handle this case? Input and output are both Kafka strings (no headers).

Kevin Niemann
  • 51
  • 8
  • I can also add that using input.contentType=text/plain produces this somewhat arcane exception: Exception thrown when sending a message with key='null' and payload='{-1, 5, 13, 99, 111, 114, 114, 101, 108, 97, 116, 105, 111, 110, 73, 100, 0, 0, 0, 38, 34, 102, 55, ...' – Kevin Niemann Feb 14 '17 at 21:57
  • Are you sure you are using data flow stream in this case? or, do you mean just a Spring Cloud stream app that receives data from kafka topic? – Ilayaperumal Gopinathan Feb 15 '17 at 18:43

1 Answers1

0

If the messages to the splitter come from non Spring Cloud Stream applications, then you would need to set --spring.cloud.stream.bindings.<inputChannelName>.consumer.headerMode=raw. While the application/json contentType messages have jackson exception, this would at least get the text/plain contentType messages evaluated against the expression appropriately.

Ilayaperumal Gopinathan
  • 4,099
  • 1
  • 13
  • 12
  • On the send I am now getting `failed to send Message to channel 'output'; nested exception is java.lang.IllegalArgumentException: payload must not be null`. I turned on the debug logs, there is definitely something in the payload: – Kevin Niemann Feb 16 '17 at 00:50
  • `DEBUG 1311 --- [afka-listener-1] o.s.integration.channel.DirectChannel : preSend on channel 'output', message: GenericMessage [payload={"some": "data"}, headers={sequenceNumber=1, kafka_offset=2663, sequenceSize=1, correlationId=f9c64983-5908-af0c-3b2c-bb9c4bb03e86, id=07 3d1d21-8474-bdf0-46c8-e4a71e2117b2, kafka_receivedPartitionId=0, kafka_receivedTopic=MyTopic, contentType=te xt/plain, timestamp=1487206118408}]` – Kevin Niemann Feb 16 '17 at 00:51
  • Could I be seeing this issue? http://stackoverflow.com/questions/41781351/spring-cloud-dataflow-type-conversion-not-working-in-processor-component – Kevin Niemann Feb 16 '17 at 01:04
  • May be. But, do you have any debug log on why message conversion fails when sending output? Have you tried setting any explicit output contentType? Also, is that a typo on the contentType=te xt/plain above? – Ilayaperumal Gopinathan Feb 16 '17 at 04:51
  • Yes I'm using "text/plain" on both input and output when I get the null payload. What level would I see the message conversion fail on output? I've set app.*.logging.level.org.springframework=DEBUG – Kevin Niemann Feb 16 '17 at 16:33