resolved
Anonymous
Unregistered
Would it be possible to implement this column family and query/insert such via CQL?
post {
id: 1
title: "Hello Dennis"
body: "Post body..."
timestamp '2012-03-18 17:41:55'
tags: ["hello", "world", "testing"]
}
"tags" would also need to be indexable - to the best of my knowledge the above isn't possible when interfacing via CQL, I've read up on composite columns but information is disturbing thin. If anyone could help direct me it'd be much appreciated.
Cheers!
You'd have to split the tags out into a separate columnfamily.
https://issues.apache.org/jira/browse/CASSANDRA-3647 is open to support lists, maps, and sets directly as column values.
Anonymous
Unregistered
So essentially I would be normalising the data as follows:
posts {
id uuid
...
}
postTags {
id binary, // hash / checksum
tag string
}
postTagsMap {
postID uuid,
tagID binary
}Makes sense really as it allows indexing on tags and tag/post ID maps, although it feels very RDBMS (I don't mind).
Better would be to denormalize just one level:
CREATE TABLE post_tags (
post_id uuid,
tag string,
PRIMARY KEY (post_id, tag)
);
(See http://www.datastax.com/dev/blog/schema-in-cassandra-1-1 for discussion on composite PKs.)
For Cassandra 1.0 you'd need to manually spell out the wide row nature by specifying the comparator and validator instead.
Anonymous
Unregistered
Thanks for the help, I compile from source so I'm using 1.1.0-beta and the latest JDBC.
I'm assuming the composite primary key will have to be created via JDBC rather than ./cqlsh as it doesn't yet support CQL 2+?
Anonymous
Unregistered
Thanks for the heads up!
