[CASSANDRA-7711] composite column not sliced when using IN clause on (other) composite columns

• Type: Bug
• Status: Resolved
• Priority: Low
• Resolution: Fixed
• Fix Version/s: 2.0.11, 2.1.1
• Component/s: Legacy/CQL
• Labels: cql

Environment

cassandra 2.0.9

• Severity: Low
• Since Version: 2.0.9
• Assignee: Benjamin Lerer
• Reporter: Frens Jan Rumph

Authors

Benjamin Lerer

Reviewers

Tom Hobbs

• Created: 07/Aug/14 10:19
• Updated: 16/Apr/19 09:31
• Resolved: 03/Sep/14 18:06

Description

Hi,

I'm storing data points in cassandra keyed by a number of values and a timestamp. I'd want to use IN clauses to select points and sliced by time. The in clauses work, but I can't get it to work in combination with the slicing: all values are returned / the range in the where clause seems to be ignored.

A dumbed down abstract version of my layout and some sample data:

create table tbl (
  a text,
  b text,
  c int,
  d int,
  primary key ((a), b, c)
);

insert into tbl (a,b,c,d) values ('a1', 'b1', 1, 1);
insert into tbl (a,b,c,d) values ('a1', 'b1', 2, 2);
insert into tbl (a,b,c,d) values ('a1', 'b2', 1, 1);
insert into tbl (a,b,c,d) values ('a1', 'b2', 2, 2);
insert into tbl (a,b,c,d) values ('a2', 'b1', 1, 1);
insert into tbl (a,b,c,d) values ('a2', 'b1', 2, 2);
insert into tbl (a,b,c,d) values ('a3', 'b2', 1, 1);
insert into tbl (a,b,c,d) values ('a3', 'b2', 2, 2);

So the table contains:

 a  | b  | c | d
----+----+---+---
 a1 | b1 | 1 | 1
 a1 | b1 | 2 | 2
 a1 | b2 | 1 | 1
 a1 | b2 | 2 | 2
 a2 | b1 | 1 | 1
 a2 | b1 | 2 | 2
 a3 | b2 | 1 | 1
 a3 | b2 | 2 | 2

When performing select * from tbl where a in ('a1', 'a2') and (b) in (('b1'), ('b2')) and c > 1; I get:

 a  | b  | c | d
----+----+---+---
 a1 | b1 | 1 | 1
 a1 | b1 | 2 | 2
 a1 | b2 | 1 | 1
 a1 | b2 | 2 | 2
 a2 | b1 | 1 | 1
 a2 | b1 | 2 | 2

But I expected:

 a  | b  | c | d
----+----+---+---
 a1 | b1 | 2 | 2
 a1 | b2 | 2 | 2
 a2 | b1 | 2 | 2

Am I doing something wrong? Or is c > 1 incorrectly ignored?

select * from tbl where a in ('a1', 'a2') and b='b1' and c > 1; does correctly produce:

a  | b  | c | d
----+----+---+---
 a1 | b1 | 2 | 2
 a2 | b1 | 2 | 2

So I expect this behaviour to relate to the interworking of the IN clause on the clustering column b and the > predicate on column c.

Cheers,
Frens Jan