CODEWITHSUNDEEP
cassandracassandra-cli
Wild Goat
Wild Goat

Reputation: 3579

Cassandra table structure suggestion and way of query

I am trying to create a following hierarchy: UserId as rowKey, Hourly time series as columns and inside each hourly column I want to have a user specific information such as hourly activity.

{
   UserId:long
   {
      Timestamp:datetime{
         pageview: integer,
         clicks:integer
      }
   }

I've read that it is possible to achieve it using supercolumns but at the same time it was mentioned that supercolumns are outdated right now. If it is true, any alternatives I can use?

Could you please provide me CQL / Java thrift example how should I create and insert such type of structure in Cassandra?

Thanks!

Upvotes: 0

Views: 208

Answers (2)

Ananth
Ananth

Reputation: 991

If your information is subjected to a particular used and accessed together . For example,if you are at anytime , require both clicks and pageview, i would suggest you to use it as a json store

CREATE TABLE user_click_by_hour(
userid long,
time_stamp timestamp,
val text,
PRIMARY KEY(userid,time_stamp)
)

val is a json object containing clicks, pageview and etc.

Advantage

1.You need not worry about altering the table for adding extra column, which add a null value for each and every previous entry

  1. If this data is designated to grow, you are bound to save a lot of space as there is one less column metadata in each node

Upvotes: 0

Dinusha
Dinusha

Reputation: 716

You can user composite primary key for this, I add a table creation CQL query for the table. And you can use counter column for clicks. 

CREATE TABLE user_click_by_hour(
userid long,
time_stamp timestamp,
clicks int,
pageview int,
PRIMARY KEY(userid,time_stamp)

)

Upvotes: 1

Related Questions