Querying on the basis of some attribute of composite primary key

Question

Hello,
I am looking for suggestion about querying just on the basis of one attribute of the composite primary key.
TO exemplify,
I have a 6 attributes example A1,A2,A3,A4,A5,A6.
A1,A2 and A3 together, serve as composite primary key.Now,because of the need of the project,I want to do querying on the basis of any A1,A2 or A3.One way I could think of is to have secondary indices on each of A1,A2,A3.
Can some one explain me roughly how to go about it?
I am a new user of BerkeleyDB Java Edition,hence not sure what would be a good way to do it.I understand one way to do it,would be to keep A1,A2,A3 in the Key class and A1,A2,A3,A4,A5,A6 in the value class as well.Then create secondary indices on the A1,A2 and A3 individually.
Can someone suggest a more efficient way?
Thanks,
Will appreciate any suggestions.
Prateek

Answer 1

Exactly as you said. Create secondary indices on each attribute you want to index off of. I don't use the Java interface, but what you pretty much need to do is form your secondary key like so:
skey: pkey_individual_attribute
skeysize: pkey_individual_attribute_sizeHow to do this is documented in the BDB Java API docs:
For example:
    class MyKeyCreator implements SecondaryKeyCreator {
        public boolean createSecondaryKey(SecondaryDatabase secondary,
                                            DatabaseEntry key,
                                            DatabaseEntry data,
                                            DatabaseEntry result)
            throws DatabaseException {
            // DO HERE: Extract the secondary key from the primary key and
            // data, and set the secondary key into the result parameter.
            return true;
    SecondaryConfig secConfig = new SecondaryConfig();
    secConfig.setKeyCreator(new MyKeyCreator());
    // Now pass secConfig to Environment.openSecondaryDatabaseThe extractor function used to construct the secondary index is passed the primary key and primary data - therefore all the data is available to you with no need to duplicate the key within the data itself. While the standard example is to use some part of the primary data to form a secondary key - there's absolutely nothing against using only a part of the primary key to form a secondary key instead. The only thing you have to do is slice up said primary key and construct the "result" parameter to be a single attribute. The backend API already knows which composite key this secondary entry will be associated with and as such will implicitly form the data (or as you called it "value") section of the index (which will be the composite primary key passed to it).
The primary key/data should consist of the composite A1,A2,A3 with only A4,A5,A6 as data.
The secondary->get() call (within the Java API) takes a key and provides back the primary key and primary data (basically the same as the db->pget() call in the C API). Since you've already indexed individual attributes, based off of the composite key, into their own respective databases - you just query from one of your secondary indexes with whatever specific attribute as the key. You then use the filled in primary key and primary data to work off of.