CODEWITHSUNDEEP
xmljsonnosqlredis
user1302575
user1302575

Reputation:

Indexing JSON or XML

What is the widely used technique in indexing a JSON or XML string when storing it on a NoSQL database like Redis?

I know database like MongoDB provides this already, but I want to understand how to implement this when I will be using a Key-value store, like Redis or Voldemort.

Such that:

Upvotes: 1

Views: 538

Answers (1)

Didier Spezia
Didier Spezia

Reputation: 73246

With a pure key/value store, you are supposed to maintain an additional set of keys to simulate a secondary index.

For instance you can store:

user:1   ->   { id:1, firstname:Bilbo, lastname:Baggins, race:hobbit }
user:2   ->   { id:2, firstname:Peregrin, lastname:Took, race:hobbit }

and then:

firstname:Bilbo   -   > [1]
lastname:Baggins     -> [1]
firstname:Peregrin   -> [2]    
lastname:Took        -> [2]
race:hobbit          -> [1,2]

To find hobbit users, get the value of race:hobbit, and then for each returned id, get user:id.

Of course with a pure key/value store (memcached for instance), it is difficult to manage indexes with low cardinalities (ie. lots of entries for a given value).

With Redis, it is easier due to its support of set and hash datatypes. Also, intersection/union of redis sets is a convenient way to implement queries with logical AND/OR expressions.

See also the following questions:

how to have relations many to many in redis

Porting from SQLite to Redis

Project Voldemort also supports list storage that can be used for a similar purpose (provided there are not too many entries per value).

Upvotes: 1

Related Questions