[
https://issues.apache.org/jira/browse/CALCITE-4486?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17382410#comment-17382410
]
duan xiong commented on CALCITE-4486:
-------------------------------------
According to jooq's
[unique-predicate|https://www.jooq.org/doc/3.0/manual/sql-building/conditional-expressions/unique-predicate/],
we can implement the UNIQUE predicate by a convert to EXISTS in
SubQueryRemoveRule rewriteUnique method.
For example:
UNIQUE (SELECT PUBLISHED_IN FROM BOOK WHERE AUTHOR_ID = 3)
{code:java}
NOT EXISTS (
SELECT 1 FROM (
SELECT PUBLISHED_IN
FROM BOOK
WHERE AUTHOR_ID = 3
) T
WHERE (T.PUBLISHED_IN) IS NOT NULL
GROUP BY T.PUBLISHED_IN
HAVING COUNT(*) > 1
){code}
But When have more than 1 field in the subquery, How to handle it?
Because we can group by like this:
{code:java}
GROUP BY $1,$2{code}
But how to add the filter condition(connect each field condition by AND
operator?) like this:
{code:java}
WHERE ($1) IS NOT NULL and ($2) IS NOT NULL{code}
I need some advice about this or confirm this design is right.
> UNIQUE predicate
> ----------------
>
> Key: CALCITE-4486
> URL: https://issues.apache.org/jira/browse/CALCITE-4486
> Project: Calcite
> Issue Type: Bug
> Reporter: Julian Hyde
> Assignee: duan xiong
> Priority: Major
>
> Implement the UNIQUE predicate, as specified by the SQL standard.
> Example:
> {code:java}
> UNIQUE (SELECT publishedIn FROM Book
> WHERE authorId = 3)
> NOT UNIQUE (SELECT publishedIn FROM BOOK
> WHERE authorId = 3)
> {code}
> {{UNIQUE}} returns {{FALSE}} if the query returns two or more rows that do
> not include NULL values and are equal. Like {{EXISTS}}, it never returns
> {{UNKNOWN}}.
> More details at
> [JOOQ|https://www.jooq.org/doc/3.0/manual/sql-building/conditional-expressions/unique-predicate/].
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
