[CASSANDRA-14808] Support ORDER BY with 2ndary Indexes

• Type: Bug
• Status: Open
• Priority: Normal
• Resolution: Unresolved
• Component/s: Legacy/CQL

Environment

Apache Cassandra 3.11.3

• Severity: Normal
• Assignee: Unassigned
• Reporter: Yan Yao
• Created: 10/Oct/18 06:01
• Updated: 18/Dec/19 23:16

Description

 Suppose we have a generic table:

CREATE TABLE base_table(
    partition1 uuid,
    ...
    partitionN uuid,
    static_column text static,
    clustering1 uuid,
    ...
    clusteringM uuid,
    regular text,
    list_text list,
    set_text set,
    map_int_text map<int, text>,
    PRIMARY KEY((partition1, ..., partitionN), clustering1, ... , clusteringN)
);

 

And create an index on regular text column, the schema of the index table will be:

CREATE TABLE regular_idx(
    regular text,
    partitionColumns blob,
    clustering1 uuid,
    ...
    clusteringM uuid,
    PRIMARY KEY((regular), partitionColumns, clustering1, ..., clusteringM)
);

 

Then it's possible to execute queries like:

SELECT * FROM base_table WHERE regular = <some_value_0> AND partition1 = <some_value_1> AND ... AND partitionN = <some_value_N> ORDER BY clustering1;

 

However, CQL3 would check if the secondary index is used WITH order by during prepare a select statement, and throw an exception at once for queries like above. 

Could we support ORDER BY with 2ndary Indexes at least for above query pattern?