[
https://issues.apache.org/jira/browse/CASSANDRA-19699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17854307#comment-17854307
]
Benjamin Lerer commented on CASSANDRA-19699:
--------------------------------------------
This behaviour is not a bug. It has been there for 9 years (CASSANDRA-728) and
if my understanding is correct is an inheritance from Thrift. So, I doubt that
it is something that we can change without taking the risk of breaking some
existing code out there.
>From my point of view the behaviour is correct and returning something
>different would be quite surprising for the user as it will not match any more
>what was requested with the CQL query.
> Tuple clustering slice queries return non-contiguous results for mixed
> asc/desc clustering columns.
> ---------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-19699
> URL: https://issues.apache.org/jira/browse/CASSANDRA-19699
> Project: Cassandra
> Issue Type: Bug
> Components: CQL/Interpreter
> Reporter: Blake Eggleston
> Priority: Normal
>
> Given the table:
> {{CREATE TABLE ks.tbl (k int, c1 int, c2 int, v int, PRIMARY KEY (k, c1, c2))
> WITH CLUSTERING ORDER BY (c1 ASC, c2 DESC)}}
> (Note the different clustering orders of c1 and c2)
> If you run a slice query with a clustering tuple like:
> {{SELECT * FROM ks.tbl WHERE k=0 AND (c1, c2) >= (2, 3)}}
> C* will return a non contiguous set of rows that would match the following 2
> queries:
> {{SELECT * FROM ks.tbl WHERE k=0 AND c1=2 AND c2 >=3}}
> {{SELECT * FROM ks.tbl WHERE k=0 AND c1>2.}}
> Whether a query like {{(c1, c2) >= (2, 3)}} makes sense for mixed ASC/DESC
> clustering columns is debateable, but there is a use case for paging type
> queries like this, so I’d propose that the tuple slice behavior be changed to
> simply use the direction of the first clustering column. In other words,
> {{(c1, c2) >= (2, 3) }}would start at the row where {{c1=2 and c2=3}} and
> continue iterating forwards, regardless of the direction of subsequent
> clustering columns.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
