About data compression

Question

The lastest version of Berkeley DB supports data compression with its set_bt_compression method.
I created a database, using default data compression method provided by Berkeley DB. Like this:
DB *dbp;
db_create(&dbp, inenv, 0);
dbp->set_flags( dbp, DB_DUPSORT );
dbp->set_bt_compress(dbp, NULL, NULL);
Then i insert key, data.
For keys, they are random char arrays,
For data, they are char array with the same content.
Now the problem is: the compressed database file is the same size of the one that i didn't use the compress method.
Can someone tell me why? THX

Answer 1

Hi,
This is likely because the default compression function does not have much to work with.
Specifying NULL for both compression and decompression functions in the DB->set_bt_compress method call implies using the default compression/decompression functions in BDB. Berkeley DB's default compression function performs prefix compression on all keys and prefix compression on data values for duplicate keys.
You haven't specified a prefix or key comparison function (DB->set_bt_prefix, DB->set_bt_compare), hence a default lexical comparison function is used as the prefix function. Given that your keys are random char arrays, the default lexical comparison function may not perform very well in identifying efficient (large-sized) prefixes for the keys.
Also, as the keys are truly random, it's unlikely that you'll have duplicates, so there's likely nothing to compress on data values for duplicate keys.
Even if the compression function does compress any keys' prefixes or prefixes for duplicate's data items, if the compressed items (and uncompressed ones) still require to be stored on the same number of database pages as in the case without compression, you'll not see any difference in database file size.
Regards,
Andrei