[jira] [Commented] (CASSANDRA-5304) Support 2ndary indexed columns in UPDATE and DELETE
[ https://issues.apache.org/jira/browse/CASSANDRA-5304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13646478#comment-13646478 ] Sylvain Lebresne commented on CASSANDRA-5304: - bq. Bulk-modification is really best left to things like Hadoop As a side note, that kind of modification-by-2ndary index could also be used for updating/deleting selected CQL3 rows within one partition (within a "wide row"), in which case Hadoop is definitively overkill (and likely way more inefficient), and at least some have suggested that this was common-ish (CASSANDRA-5527). Not to suggest that we re-open this necessarily, I still somewhat stand by "maybe it's better to let user do it client-side and be aware of what that involves", but just wanted to add some color. > Support 2ndary indexed columns in UPDATE and DELETE > --- > > Key: CASSANDRA-5304 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5304 > Project: Cassandra > Issue Type: Wish > Components: Core >Affects Versions: 1.2.2 >Reporter: Joachim Haagen Skeie >Priority: Minor > > I have a Column Family with the following index: > CREATE INDEX live_stat_is_calculated ON live_statistics (iscalculated) > Then, I would like to delete records based on this index via CQL3 query: > delete from live_statistics where iscalculated = true; > But Cassandra returns the following error: > PRIMARY KEY part iscalculated found in SET part -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5304) Support 2ndary indexed columns in UPDATE and DELETE
[ https://issues.apache.org/jira/browse/CASSANDRA-5304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13611614#comment-13611614 ] Jonathan Ellis commented on CASSANDRA-5304: --- Bulk-modification is really best left to things like Hadoop. It's very, very rare that this is useful for the kind of live, massively concurrent application that C* supports. > Support 2ndary indexed columns in UPDATE and DELETE > --- > > Key: CASSANDRA-5304 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5304 > Project: Cassandra > Issue Type: Wish > Components: Core >Affects Versions: 1.2.2 >Reporter: Joachim Haagen Skeie >Priority: Minor > > I have a Column Family with the following index: > CREATE INDEX live_stat_is_calculated ON live_statistics (iscalculated) > Then, I would like to delete records based on this index via CQL3 query: > delete from live_statistics where iscalculated = true; > But Cassandra returns the following error: > PRIMARY KEY part iscalculated found in SET part -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-5304) Support 2ndary indexed columns in UPDATE and DELETE
[ https://issues.apache.org/jira/browse/CASSANDRA-5304?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13592381#comment-13592381 ] Sylvain Lebresne commented on CASSANDRA-5304: - Sorry, I'm changing the title of this back because I told Joachim to create that ticket and I wasn't thinking of the error message. Namely, we don't support 2ndary indexed columns in UPDATE and DELETE where clauses, and I believe that's a legit feature request. Do we want to implement it is another question. We could do it slightly more efficiently than what use can do today (by reading first, then updating/deleting second) as we could push that to the replica directly (hence saving round-trips, and not just between client and server, but internally too). That being said, it would hairy to implement and while it would be faster than doing it client side, it wouldn't be order of magnitude faster either, so maybe it's better to let user do it client-side and be aware of what that involves. Nonetheless, even if we close as won't fix for those reasons, I wanted to have the issue here as public record, because I suspect this is a feature request that might come back. As for the error message itself, I'm happy to ninja-fix it, except that from reading the code I have no clue how a delete query could have sent you this error message. The error message you should have gotten is "Non PRIMARY KEY iscalculated found in where clause", which is at least true (if not very explanatory). Are you sure a delete give you that error message? > Support 2ndary indexed columns in UPDATE and DELETE > --- > > Key: CASSANDRA-5304 > URL: https://issues.apache.org/jira/browse/CASSANDRA-5304 > Project: Cassandra > Issue Type: Wish > Components: Core >Affects Versions: 1.2.2 >Reporter: Joachim Haagen Skeie >Priority: Minor > > I have a Column Family with the following index: > CREATE INDEX live_stat_is_calculated ON live_statistics (iscalculated) > Then, I would like to delete records based on this index via CQL3 query: > delete from live_statistics where iscalculated = true; > But Cassandra returns the following error: > PRIMARY KEY part iscalculated found in SET part -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira