[
https://issues.apache.org/jira/browse/CALCITE-3339?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16979561#comment-16979561
]
Julian Hyde edited comment on CALCITE-3339 at 11/21/19 7:37 PM:
----------------------------------------------------------------
Reviewing your [PR 1599|https://github.com/apache/calcite/pull/1599]:
* It would be good to isolate 'Add DESCRIPTOR' in its own change, including
the DESCRIPTOR operator in SqlStdOperatorTable, but not linked to the TUMBLE
function.
* Add a test to SqlParserTest (yes, I know that you accomplished this without
changing the parser, and that's great, but it impinges on the parser's area of
functionality).
* Needs more (user and developer) documentation about what a descriptor is.
* The return type of the DESCRIPTOR function should be something special. I
think SqlTypeName.COLUMN_LIST will do.
* Also add a validator test that calls a UDF with a DESCRIPTOR argument. You
may need a special type that identifies an argument of a UDF as a descriptor;
if so, add class org.apache.calcite.runtime.Descriptor.
* It's sufficient for this change to do parsing and validation. Don't need to
do SqlToRel or execution.
Also, it's fine if you carry on developing this in the same branch as TUMBLE,
if you wish. We can re-order and re-organize the commits later.
was (Author: julianhyde):
Reviewing your [PR 1599|https://github.com/apache/calcite/pull/1599]:
* It would be good to isolate 'Add DESCRIPTOR' in its own change, including
the DESCRIPTOR operator in SqlStdOperatorTable, but not linked to the TUMBLE
function.
* Add a test to SqlParserTest (yes, I know that you accomplished this without
changing the parser, and that's great, but it impinges on the parser's area of
functionality).
* Needs more (user and developer) documentation about what a descriptor is.
* The return type of the DESCRIPTOR function should be something special. I
think SqlTypeName.COLUMN_LIST will do.
* Also add a validator test that calls a UDF with a DESCRIPTOR argument. You
may need a special type that identifies an argument of a UDF as a descriptor;
if so, add class org.apache.calcite.runtime.Descriptor.
* It's sufficient for this change to do parsing and validation. Don't need to
do SqlToRel or execution.
> DESCRIPTOR as a SQL operator in SqlStdOperatorTable
> ---------------------------------------------------
>
> Key: CALCITE-3339
> URL: https://issues.apache.org/jira/browse/CALCITE-3339
> Project: Calcite
> Issue Type: Sub-task
> Reporter: Rui Wang
> Assignee: Rui Wang
> Priority: Major
> Labels: pull-request-available
> Time Spent: 10m
> Remaining Estimate: 0h
>
> For query:
> SELECT *
> FROM TABLE(TUMBLE_TVF(
> TABLE ORDERS,
> DESCRIPTOR(ROWTIME),
> INTERVAL '10' MINUTE))
> TABLE ORDERS is converted to SqlPrefixOperator, but DESCRIPTOR(ROWTIME) has
> no mapping in SqlStdOperatorTable.
> There are two options:
> 1. There is a SqlColumnListConstructor which serves the same(similar) purpose
> to specific a list of column.
> 2. We create a new operator for DESCRIPTOR.
> Reuse existing code is always good so we can start from option one and see if
> it works.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
