For CASSANDRA-7814 we want to add a DESCRIBE INDEX idxname cqlsh command as well as making the INDEX or TABLE redundant, and just be able to do DESCRIBE ksname|tablename|idxname.
At the moment we can retrieve index metadata by looping on all tables and then for each table by looping on all columns. Whilst this is OK, it may cause performance issues for large schemas.
It would be convenient to map the index metadata by index name and store it in a dictionary at the table or keyspace level or both as we've done in this branch:
We would need this for Cassandra 2.1 if possible.
There's a Cassandra bug which causes this test to fail intermittently on C* 1.2 and 2.0. Once it has been addressed, I'll push this ticket to resolved.
The C* bug has been fixed and has been patched into their various branches. I've tested to confirm that the failures no longer exist.