cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benjamin Lerer (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-10271) ORDER BY should allow skipping equality-restricted clustering columns
Date Thu, 22 Dec 2016 15:35:58 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-10271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15770315#comment-15770315
] 

Benjamin Lerer commented on CASSANDRA-10271:
--------------------------------------------

Thanks for the patch and sorry for the delay.

Based on your unit tests, I think there is some misunderstanding in what should be allowed
or not.
Basically, we want to allow to skip a column in the {{ORDER BY}} clause if the clustering
column is restricted by an {{=}} or an {{IN}} with only one restriction.

So the following tests:
{code}
        assertRows(execute("SELECT * FROM %s WHERE a=? AND b<? ORDER BY c DESC", 0, 1),
                   row(0, 0, 2, 2),
                   row(0, 0, 1, 1),
                   row(0, 0, 0, 0));
{code}
and
{code}
        assertRows(execute("SELECT * FROM %s WHERE a=? AND b IN (?, ?) ORDER BY c ASC", 0,
1, 3),
                   row(0, 1, 0, 3),
                   row(0, 1, 1, 4),
                   row(0, 1, 2, 5));
{code}
are in fact invalid. In both cases an {{InvalidRequestException}} should be thrown.

If  a clustering column is restricted by an {{=}} or an {{IN}} with only one restriction {{StatementRestrictions::isColumnRestrictedByEq}}
will return {{true}}.
By consequence,
{code}
 if (!restrictions.clusteringColumnsFollowOrderByRestrictions(def.position()))
    checkTrue(i++ == def.position(),
                    "Order by currently only support the ordering of columns following their
declared order in the PRIMARY KEY");

{code}
could simply be replaced by:
{code}
 while (i != def.position())
 {
    checkTrue(restrictions.isColumnRestrictedByEq(clusteringColumns.get(i++)),
                    "Order by currently only support the ordering of columns following their
declared order in the PRIMARY KEY");
 }
 i++;
{code}
if I am not mistaken.

It will also be great if you could check the error messages using {{assertInvalidMessage}}
instead of {{assertInvalid}} (we had bad surprises in the past) and add the following test
for completeness:
{code}
          assertRows(execute("SELECT * FROM %s WHERE a=? AND b IN (?, ?) AND c = ? ORDER BY
b ASC", 0, 1, 3, 1),
                     row(0, 1, 1, 4));
{code}

> ORDER BY should allow skipping equality-restricted clustering columns
> ---------------------------------------------------------------------
>
>                 Key: CASSANDRA-10271
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-10271
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: CQL
>            Reporter: Tyler Hobbs
>            Assignee: Brett Snyder
>            Priority: Minor
>             Fix For: 3.x
>
>         Attachments: 10271-3.x.txt, cassandra-2.2-10271.txt
>
>
> Given a table like the following:
> {noformat}
> CREATE TABLE foo (a int, b int, c int, d int, PRIMARY KEY (a, b, c));
> {noformat}
> We should support a query like this:
> {noformat}
> SELECT * FROM foo WHERE a = 0 AND b = 0 ORDER BY c ASC;
> {noformat}
> Currently, this results in the following error:
> {noformat}
> [Invalid query] message="Order by currently only support the ordering of columns following
their declared order in the PRIMARY KEY"
> {noformat}
> However, since {{b}} is restricted by an equality restriction, we shouldn't require it
to be present in the {{ORDER BY}} clause.
> As a workaround, you can use this query instead:
> {noformat}
> SELECT * FROM foo WHERE a = 0 AND b = 0 ORDER BY b ASC, c ASC;
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message