[
https://issues.apache.org/jira/browse/FLINK-20175?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17239289#comment-17239289
]
Dawid Wysakowicz commented on FLINK-20175:
------------------------------------------
I am currently checking if it is feasible. I am running a backported version on
my azure:
https://dev.azure.com/wysakowiczdawid/Flink/_build/results?buildId=578&view=results
Unfortunately a backport is not as easy as simply cherry picking.
> Avro Confluent Registry SQL format does not support adding nullable columns
> ---------------------------------------------------------------------------
>
> Key: FLINK-20175
> URL: https://issues.apache.org/jira/browse/FLINK-20175
> Project: Flink
> Issue Type: Bug
> Components: Formats (JSON, Avro, Parquet, ORC, SequenceFile), Table
> SQL / Ecosystem
> Affects Versions: 1.12.0, 1.11.2
> Reporter: Dawid Wysakowicz
> Assignee: Dawid Wysakowicz
> Priority: Critical
> Fix For: 1.12.0
>
>
> The {{AvroSchemaConverter#convertToSchema}} generates a union with "null" for
> nullable logical types, but it does not set the default value to null. In
> turn it makes it impossible to generate a backwards compatible schema from a
> DDL statement.
> Example:
> 1. Create a table: {{CREATE TABLE t (id INT NOT NULL) WITH (/* avro confluent
> format*/)}}
> 2. Create a new table over the same topic or alter the old table with
> {{CREATE TABLE newT(id INT NOT NULL, optionalDescription STRING) WITH (/*
> avro confluent format */)}}
> 3. When reading from {{newT}} records inserted into {{t}} it will fail,
> because the {{optionalDescription}} has no default value.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
