On 2017-07-10 07:13 (-0700), Siddharth Prakash Singh <sps...@gmail.com> wrote: > I am planning to build a user activity timeline. Users on our system > generates different kind of activity. For example - Search some product, > Calling our sales team, Marking favourite etc. > Now I would like to generate timeline based on these activities. Timeline > could be for all events, filtered on specific set of events, filtered on > time interval, filtered on specific set of events between time intervals. > Composite column keys looks like a viable solution. > > Any other thoughts here? >
You probably want to take advantage of multiple/compound clustering keys, at least one of which being a timeuuid to give yourself ordering, and one giving you a 'type' of event. CREATE TABLE whatever ( product_id uuid , event_type text, event_id timeuuid, event_action text, event_data text, PRIMARY KEY(product_id, event_id, event_type, event_action, event_data)); This will let you do "SELECT * FROM whatever WHERE product_id=?" and get all of the events, sorted by time, then by type, then you can have another unique "action", and finally a data field where you can shove your blob of whatever it is. This would let you do time slices by specifying "event_id >= X and event_id < Y", but you'd need (want) to filter event_type client side. Alternatively, PRIMARY KEY(product_id, event_type, event_id, event_action, event_data) would let you do event_type=X and event_id >= Y and event_id < Z, which is all events of a given type within a slice. "product_id" may not be the natural partition key, feel free to use a compound partition key as well (may be "PRIMARY KEY((product_id, office_id), event_type, event_id, event_action, event_data)" to make a partition-per-office, as a silly example. --------------------------------------------------------------------- To unsubscribe, e-mail: user-unsubscr...@cassandra.apache.org For additional commands, e-mail: user-h...@cassandra.apache.org
