[
https://issues.apache.org/jira/browse/FLINK-24001?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17412517#comment-17412517
]
JING ZHANG commented on FLINK-24001:
------------------------------------
This pr aims to support evaluating individual window table-valued function. For
example, run the following SQL.
{code:java}
SELECT * FROM TABLE(
TUMBLE(TABLE Bid, DESCRIPTOR(bidtime), INTERVAL '10' MINUTES));
{code}
I have two points which need discuss. Any suggestions would be appreciated,
thanks a lot. cc [~twalthr] [~jark] [~lzljs3620320] [~godfrey] [~MartijnVisser]
# When to send data to downstream ?
## There are two options for row-semantic window table function(tumbling, hop,
cumulate window)
### Option1: send the data directly when receive input data
### Option2: buffer the input data into state, send the data after the window
is passed
## There is only one option for set-semantic window table function (session
window): buffer the input data into state, send the data after the window is
passed
# If we choose option2 (buffer the input data into state, send the data after
the window is passed) for row-semantic window, I have two minor problems in
implementation.
## I wanna to use timer to trigger window. Timer only works on KeyedStream.
However row-semantic window table function operator does not require input
stream are KeyedStream.
## I wanna to store the data into state. If input stream is not keyedStream,
I have to use Operator state.
> Support evaluating individual window table-valued function in runtime
> ---------------------------------------------------------------------
>
> Key: FLINK-24001
> URL: https://issues.apache.org/jira/browse/FLINK-24001
> Project: Flink
> Issue Type: Sub-task
> Components: Table SQL / Runtime
> Affects Versions: 1.15.0
> Reporter: JING ZHANG
> Priority: Major
>
> {{Currently, window table-valued function has to be used with other window
> operation, such as window aggregate, window topN and window join. }}
> {{In the ticket, we aim to support evaluating individual window table-valued
> function in runtime, which means, introduce an operator to handle this.}}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
