[
https://issues.apache.org/jira/browse/FLINK-6864?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16168056#comment-16168056
]
ASF GitHub Bot commented on FLINK-6864:
---------------------------------------
Github user greghogan commented on a diff in the pull request:
https://github.com/apache/flink/pull/4574#discussion_r139184023
--- Diff: docs/dev/types_serialization.md ---
@@ -115,6 +115,8 @@ conditions are fulfilled:
or have a public getter- and a setter- method that follows the Java beans
naming conventions for getters and setters.
+Note that when a data type can't be recognized as a POJO type, it will be
handled as GenericType.
--- End diff --
"... it will be processed as a GenericType and serialized with Kryo"? Is it
generally understood that use of Kryo can have a significant impact on
performance? It's not just serialization which is slower. It looks like if the
class does not implement `NormalizedKey` then objects are deserialized for each
comparison.
> Remove confusing "invalid POJO type" messages from TypeExtractor
> ----------------------------------------------------------------
>
> Key: FLINK-6864
> URL: https://issues.apache.org/jira/browse/FLINK-6864
> Project: Flink
> Issue Type: Improvement
> Components: Documentation, Type Serialization System
> Reporter: Tzu-Li (Gordon) Tai
> Assignee: Fang Yong
>
> When a user's type cannot be treated as a POJO, the {{TypeExtractor}} will
> log warnings such as ".. must have a default constructor to be used as a
> POJO.", " ... is not a valid POJO type because not all fields are valid POJO
> fields." in the {{analyzePojo}} method.
> These messages are often conceived as misleading for the user to think that
> the job should have failed, whereas in fact in these cases Flink just
> fallsback to Kryo and treat then as generic types. We should remove these
> messages, and at the same time improve the type serialization docs at [1] to
> explicitly inform what it means when Flink does / does not recognizes a user
> type as a POJO.
> [1]
> https://ci.apache.org/projects/flink/flink-docs-release-1.3/dev/types_serialization.html#rules-for-pojo-types
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)