[
https://issues.apache.org/jira/browse/FLINK-19038?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17185585#comment-17185585
]
Nicholas Jiang edited comment on FLINK-19038 at 8/29/20, 8:50 AM:
------------------------------------------------------------------
[~dwysakowicz], I have confusion as follows:
{code:java}
private SortQueryOperation validateAndGetChildSort(QueryOperation child,
PostResolverFactory postResolverFactory) {
final SortQueryOperation previousSort;
if (child instanceof SortQueryOperation) {
previousSort = (SortQueryOperation) child;
} else {
previousSort = (SortQueryOperation) createSort(Collections.emptyList(),
child, postResolverFactory);
}
if ((previousSort).getFetch() != -1) {
throw new ValidationException("FETCH is already defined.");
}
return previousSort;
}
{code}
Why does method validateAndGetChildSort check the fetch of previousSort? In my
opinion, this validation makes no sense if user desire to continuously fetch or
limit. in SQL perspective, Flink SQL supports continuously fetch or limit,
therefore I thought that the Table API need also support.
was (Author: nicholasjiang):
[~twalthr],[~dwysakowicz],[~lzljs3620320]
{code:java}
private SortQueryOperation validateAndGetChildSort(QueryOperation child,
PostResolverFactory postResolverFactory) {
final SortQueryOperation previousSort;
if (child instanceof SortQueryOperation) {
previousSort = (SortQueryOperation) child;
} else {
previousSort = (SortQueryOperation) createSort(Collections.emptyList(),
child, postResolverFactory);
}
if ((previousSort).getFetch() != -1) {
throw new ValidationException("FETCH is already defined.");
}
return previousSort;
}
{code}
Why does method validateAndGetChildSort check the fetch of previousSort? In my
opinion, this validation makes no sense if user desire to continuously fetch or
limit. in SQL perspective, Flink SQL supports continuously fetch or limit,
therefore I thought that the Table API need also support.
> It doesn't support to call Table.limit() continuously
> -----------------------------------------------------
>
> Key: FLINK-19038
> URL: https://issues.apache.org/jira/browse/FLINK-19038
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / API
> Affects Versions: 1.12.0
> Reporter: Dian Fu
> Assignee: Nicholas Jiang
> Priority: Major
> Fix For: 1.12.0
>
>
> For example, table.limit(3).limit(2) will failed with "FETCH is already
> defined."
> {code}
> org.apache.flink.table.api.ValidationException: FETCH is already defined.
> at
> org.apache.flink.table.operations.utils.SortOperationFactory.validateAndGetChildSort(SortOperationFactory.java:125)
> at
> org.apache.flink.table.operations.utils.SortOperationFactory.createLimitWithFetch(SortOperationFactory.java:105)
> at
> org.apache.flink.table.operations.utils.OperationTreeBuilder.limitWithFetch(OperationTreeBuilder.java:418)
> {code}
> However, as we support to call table.limit() without specifying the order, I
> guess this should be a valid usage and should be allowed.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
