We have a top level domain object / @Entity called Audit.

For simplicity, let's say Audit.java has 4 fields:

public class Audit{
Date date; // Will just be queried on as MM/DD/yyyy
String event;
String user;
String desc;
}

In this example, we are creating 3 different Cassandra 2.0 tables to this one entity.

AuditsByDate - Partition key is the MM/dd/yyyy date of the audit
AuditsByEvent - Partition key is the audit event name
AuditsByUser - Partition key is the audit user


Each of these tables has the same 1 row written every time an audit is written. We just want to be able to query these identical rows different ways as needed.


My newbie Cassandra / Hibernate OGM question is:

Is there a way for us to have one single OGM -annotated class that can interact with all three of these highly related tables? Is there any way to use Hibernte OGM 5.x to do something basically like this?

@Entity
@Table("AuditsByDate")
@Table("AuditsByEvent")
@Table("AuditsByUser") // Mapping it to 3 tables causes each table to have a row written in each table for every insert operation.
public class Audit{

@PartitionKey(table="AuditsByDate") // Specifying each of these partition keys lets us select from the right sub-table correctly based on which partition key was passed in
@Column(name="date")
Date date; // Will just be queried on as MM/DD/yyyy

@PartitionKey(table="AuditsByEvent") // Specifying each of these partition keys lets us select from the right sub-table correctly based on which partition key was passed in
@Column(name="event")
String event;

@PartitionKey(table="AuditsByUser") // Specifying each of these partition keys lets us select from the right sub-table correctly based on which partition key was passed in
@Column(name="user")
String user;

@Column(name="desc")
String desc;
}


Or would we need a separate OGM -annotated .java class for each of our 3 tables?

Thanks in advance!!!

Unfortunately today, Hibernate OGM does not support easily what you want.
From what I understand, you use these tables as some kind of secondary indexes w/ a materialized view on each entry. And depending on the query you want to perform, you use one or the other of the tables.

We want to have that kind of capability in OGM at some point: declaratively denormalize how the data is stored and have the query engine use the best denormalized form, but we are not there yet. In the mean time indeed, you would need to create three entities which is inconvenient.

I'm tempted to say though that your use case might be solvable by Cassandra secondary indexes, am I missing something?