[jira] [Commented] (FLINK-15803) Expose the new type inference for aggregate functions

["Timo Walther (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-15803?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17172110#comment-17172110 ] 

Timo Walther commented on FLINK-15803:
--------------------------------------

Fixed in 1.12.0:
{code}
c55665e9a250ba326185e52f4e147a0098cd1a58
[FLINK-15803][table] Replace type information with serializer in StateDataViewStore

ae3ccb9767f6dc9b4b0780617773af450da56e9f
[FLINK-15803][table-runtime-blink] Add an external serializer that can handle all data types

e00f7e9c8aff3ef6262654e3b6e1deabf6b0b2b7
[FLINK-15803][table-common] Update DataView classes to the new type system

25b18fd32d178336095ac3ed83516ce5ed09714f
[FLINK-15803][table-common] Integrate data views into data type extractor

c31fdc695e232aaf8f2f692d854bdd29a8e76230
[FLINK-15803][table] Support DataViews in FLIP-65 aggregate functions

This enables support for ListView and MapView in FLIP-65 aggregate
functions. There are two major differences to the previous design:

1. We use the internal serialization format for checkpoints/savepoints
of data views. Data structure converters are used at the edges which
means that we support all data types and conversion classes.

2. The data types of data views are expressed in a declarative way. We
don't instantiate the accumulators anymore. Instead users can use
annotations or override getTypeInference.

Old aggregate functions are still supported when using
TableEnvironment.registerFunction.

{code}

> Expose the new type inference for aggregate functions
> -----------------------------------------------------
>
>                 Key: FLINK-15803
>                 URL: https://issues.apache.org/jira/browse/FLINK-15803
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Table SQL / API, Table SQL / Planner
>            Reporter: Timo Walther
>            Assignee: Timo Walther
>            Priority: Major
>              Labels: pull-request-available
>
> This will allow to use aggregate functions with the new type inference. It requires different changes through the stack. This should be implemented after we have support for structured types. It also includes a long-term solution for the concept of DataViews with the new type system.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)