[
https://issues.apache.org/jira/browse/FLINK-16048?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17159788#comment-17159788
]
Kurt Young commented on FLINK-16048:
------------------------------------
Fields declared in DDL should be the ground truth. We also need to check
whether the avro schema is compatible with the schema comes from DDL. For
example, it's ok that avro has more fields than user declared. And we have
multiple choices if user declared non-exist field in DDL, we can throw
exception or fill it with *NULL*.
> Support read/write confluent schema registry avro data from Kafka
> ------------------------------------------------------------------
>
> Key: FLINK-16048
> URL: https://issues.apache.org/jira/browse/FLINK-16048
> Project: Flink
> Issue Type: Improvement
> Components: Formats (JSON, Avro, Parquet, ORC, SequenceFile), Table
> SQL / Ecosystem
> Affects Versions: 1.11.0
> Reporter: Leonard Xu
> Assignee: Danny Chen
> Priority: Major
> Labels: pull-request-available, usability
> Fix For: 1.12.0
>
>
> *The background*
> I found SQL Kafka connector can not consume avro data that was serialized by
> `KafkaAvroSerializer` and only can consume Row data with avro schema because
> we use `AvroRowDeserializationSchema/AvroRowSerializationSchema` to se/de
> data in `AvroRowFormatFactory`.
> I think we should support this because `KafkaAvroSerializer` is very common
> in Kafka.
> and someone met same question in stackoverflow[1].
> [[1]https://stackoverflow.com/questions/56452571/caused-by-org-apache-avro-avroruntimeexception-malformed-data-length-is-negat/56478259|https://stackoverflow.com/questions/56452571/caused-by-org-apache-avro-avroruntimeexception-malformed-data-length-is-negat/56478259]
> *The format details*
> _The factory identifier (or format id)_
> There are 2 candidates now ~
> - {{avro-sr}}: the pattern borrowed from KSQL {{JSON_SR}} format [1]
> - {{avro-confluent}}: the pattern borrowed from Clickhouse {{AvroConfluent}}
> [2]
> Personally i would prefer {{avro-sr}} because it is more concise and the
> confluent is a company name which i think is not that suitable for a format
> name.
> _The format attributes_
> || Options || required || Remark ||
> | schema-string | true | avro schema string used for (de)serialization |
> | schema-registry.url | true | URL to connect to schema registry service |
> | schema-registry.subject | false | Subject name to write to the Schema
> Registry service, required for sink |
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
