Data store fast querying over combinations of key values

Question

I am looking for storing and retrieving elements of this nature:

K1 K2 K3 [...] Kn -> V

where the V field is a tiny string.

Querying should work good for any combination of keys, for instance {K1, K2, K10} or {K3, K7}.

By querying, I'm thinking about checking for equality or range inclusion (L <= Ki <= H).

In practice, there will be a maximum of 5 keys and the querying is going to be performed over all the keys or only one of them, but it would be nice to allow some space for future development.

At first I was looking at Redis, but I could not see it fit my use case.

[Update] VoltDB might look like a strong contender. What do you think ?

Answer 1

It's not entirely clear to me if you assume that there is some ordering between the keys or not. However, it seems that PostgreSQL's hstore could help you.

It allows you to store sets of (key,value)s in a column. So you could create a table with an hstore column for your keys and another column for your value.

Hstore directly provides you with the subset queries:

'K1=>1,K2=>2,K3=>3,...,Kn=>n'::hstore ?& ARRAY['K3','k7']

You could model your range queries by storing and querying values of the hstore keys. Resp., if you assume order you could have numbers as hstore keys and hstore value would be your key:

1=>k1,2=>k2,3=>k3,...

Then a range query for the 4-th key would be:

L <= (keys -> '4')::INT AND (keys -> '4')::INT <= H 

On the other hand, if the meaning of your range query is

(\forall i) L <= Ki <= H

Then I guess something could also be done.

The advantage ist that hstore supports GiST and GIN indexes so it could be reasonably fast.