[ https://issues.apache.org/jira/browse/CASSANDRA-8270?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
T Jake Luciani updated CASSANDRA-8270: -------------------------------------- Description: For some queries with clustering keys it would be helpful to allow server side to avoid processing things you will simply filter out. Example schema: {code} create TABLE foo (a text, b int, c int, d int, primary key(a, b, c)); insert into foo (a, b, c, d ) VALUES ( 'key', 2, 3, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 2, 4, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 3, 3, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 3, 4, 4); {code} {code} select count(*) from foo where a = 'key' and b = 2 and c > 3; --return 1 select count(*) from foo where a = 'key' and b > 2 and c > 3; --error select count(*) from foo where a = 'key' and c > 3; --error {code} The first query is only possible because our slices only allow a fixed prefix but if we extended slices to include slices we could effectively request for: {noformat} b(2,*) c(3,*) b(*,*) c(3,*) {noformat} was: For some queries with clustering keys it would be helpful to allow server side to avoid processing things you will simply filter out. Example schema: {code} create TABLE foo (a text, b int, c int, d int, primary key(a, b, c)); insert into foo (a, b, c, d ) VALUES ( 'key', 2, 3, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 2, 4, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 3, 3, 4); insert into foo (a, b, c, d ) VALUES ( 'key', 3, 4, 4); {code} {code} select count(*) from foo where a = 'key' and b = 2 and c > 3; --return 1 select count(*) from foo where a = 'key' and b > 2 and c > 3; --error select count(*) from foo where a = 'key' and c > 3; --error {code} The first query is only possible because our slices only allow a fixed prefix but if we extended slices to include slices we could effectively request for: {noformat} b(2,*) c (3,*) b (*,*) c (3,*) {noformat} > Allow sub slices for composites > ------------------------------- > > Key: CASSANDRA-8270 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8270 > Project: Cassandra > Issue Type: Improvement > Reporter: T Jake Luciani > Priority: Minor > > For some queries with clustering keys it would be helpful to allow server > side to avoid processing things you will simply filter out. > Example schema: > {code} > create TABLE foo (a text, b int, c int, d int, primary key(a, b, c)); > insert into foo (a, b, c, d ) VALUES ( 'key', 2, 3, 4); > insert into foo (a, b, c, d ) VALUES ( 'key', 2, 4, 4); > insert into foo (a, b, c, d ) VALUES ( 'key', 3, 3, 4); > insert into foo (a, b, c, d ) VALUES ( 'key', 3, 4, 4); > {code} > {code} > select count(*) from foo where a = 'key' and b = 2 and c > 3; --return 1 > select count(*) from foo where a = 'key' and b > 2 and c > 3; --error > select count(*) from foo where a = 'key' and c > 3; --error > {code} > The first query is only possible because our slices only allow a fixed prefix > but if we extended slices to include slices we could effectively request for: > {noformat} > b(2,*) c(3,*) > b(*,*) c(3,*) > {noformat} -- This message was sent by Atlassian JIRA (v6.3.4#6332)