[
https://issues.apache.org/jira/browse/FLINK-5429?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Flink Jira Bot updated FLINK-5429:
----------------------------------
Priority: Minor (was: Major)
> Code generate types between operators in Table API
> --------------------------------------------------
>
> Key: FLINK-5429
> URL: https://issues.apache.org/jira/browse/FLINK-5429
> Project: Flink
> Issue Type: New Feature
> Components: Table SQL / Legacy Planner
> Reporter: Timo Walther
> Priority: Minor
> Labels: auto-deprioritized-major
>
> Currently, the Table API uses the generic Row type for shipping records
> between operators in underlying DataSet and DataStream API. For efficiency
> reasons we should code generate those records. The final design is up for
> discussion but here are some ideas:
> A row like {{(a: INT NULL, b: INT NOT NULL, c: STRING)}} could look like
> {code}
> final class GeneratedRow$123 {
> public boolean a_isNull;
> public int a;
> public int b;
> public String c;
> }
> {code}
> Types could be generated using Janino in the pre-flight phase. The generated
> types should use primitive types wherever possible.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
