[
https://issues.apache.org/jira/browse/AVRO-2788?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17073190#comment-17073190
]
Tianyu Lang commented on AVRO-2788:
-----------------------------------
Hmmm. I thought I updated the patch so that no Proto change is needed. Wonder
why Jira did not get my latest patch.
Instead of adding a new "emptyArray" field, I just created a test case that
checks if "fooArray" has a default value of an empty array. I will ensure the
correct patch will be submitted.
Also, in the "How to contribute" Wiki
[https://cwiki.apache.org/confluence/display/AVRO/How+To+Contribute], we are
still advising people to submit diffs this way. Should we update the Wiki so
people just use pull requests instead?
> Generated Avro schema from Protobuf is missing default values for repeated
> fields
> ---------------------------------------------------------------------------------
>
> Key: AVRO-2788
> URL: https://issues.apache.org/jira/browse/AVRO-2788
> Project: Apache Avro
> Issue Type: Improvement
> Components: java
> Affects Versions: 1.9.2
> Reporter: Tianyu Lang
> Assignee: Tianyu Lang
> Priority: Major
> Attachments: AVRO-2788.patch
>
>
> Avro schemas generated from Protobuf schemas by *ProtobufData.java* are
> missing default values for repeated (array) fields.
> This will break compatibility when Avro is used as a transport format between
> 2 services that use Protobuf internally.
>
> For example:
> A publisher converts Protobuf to Avro, then sends the message through Kafka
> to a consumer. The consumer then converts Avro back into Protobuf, then does
> all the processing with Protobuf.
>
> A compatibility issue will occur when a new repeated Protobuf field is added
> to the consumer Protobuf schema. The corresponding Avro schema generated from
> {code:java}
> Schema schema = ProtobufData.get().getSchema(MyProtobufClass.class);
> {code}
> will not assign default values to the newly added repeated field. Because
> the publisher is still on the schema without the newly added array field,
> deserialization on the consumer side will fail since there is no default
> values to fill in.
>
>
>
> I discussed this with [~cutting] on the mailing list and it makes sense to
> just add default values for Protobuf repeated fields.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
