cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-13013) Potential regression from CASSANDRA-6377
Date Fri, 27 Jan 2017 09:35:24 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-13013?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15842446#comment-15842446
] 

Sylvain Lebresne commented on CASSANDRA-13013:
----------------------------------------------

The attached patch does properly revert things to the pre-CASSANDRA-6377 state for 3.X/trunk,
but I'll note that the comment doesn't correspond exactly to the condition anymore (both because
it doesn't mention the regular restriction conditions, but also because the comment suggests
the "static compact" condition should take precedence on any other) and I'd like to suggest
we simplify this code by:
# creating a new {{StatementRestrictions.queriesFullPartitions()}} method (that in practice
would be {{!hasClusteringRestriction && !hasRegularRestrictions}}).
# make the condition be {{queriesFullPartitions() || isStaticCompactTable}}, which is more
readable.

But those are trivial changes and I'm +1 on the patch in general so I'm fine if you do those
changes on commit.



> Potential regression from CASSANDRA-6377
> ----------------------------------------
>
>                 Key: CASSANDRA-13013
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-13013
>             Project: Cassandra
>          Issue Type: Bug
>          Components: CQL
>            Reporter: Sylvain Lebresne
>            Assignee: Benjamin Lerer
>
> As noted by [~thobbs] in CASSANDRA-12768, in 3.0 (and prior) [we return static results|https://github.com/apache/cassandra/blob/cassandra-3.0/src/java/org/apache/cassandra/cql3/statements/SelectStatement.java#L753-L754]
for a partition if it is the only results when the query is a 2ndary index query. This doesn't
seem to happen anymore in 3.X and that was removed by CASSANDRA-6377 (see the [merge commit|https://github.com/apache/cassandra/commit/8c83c8edab4f1c23c382bb0ac076cab44d5f4dda#diff-75ebe654dcf6c8c474f787abaf47bb68L705],
but that removal is actually part of the original [trunk patch for CASSANDRA-6377|https://github.com/blerer/cassandra/commit/e22a311a6d379a9f81668b7995501962ba705380#diff-75ebe654dcf6c8c474f787abaf47bb68L705]).
> The removal looks intentional but it's unclear to [~thobbs] and myself why it's not a
potentially breaking change, and even if it's a legit change, why it was done in 3.X (then
trunk) but not 3.0?
> [~blerer], can you enlighten us?



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

Mime
View raw message