Has anyone looked into CDB_File:
Looking it over, I see that it has 24 bytes of overhead per record. It stores a record with a 4 byte key length, 4 byte data length, variable length key and variable length data. Each record is pointed to by a hash bucket which contains a hash key and a pointer.
It looks like a very well designed and optimized database for its purpose, but the purpose appears to be for databases with larger variable length records and variable length keys. There is no cost for the code, but the author wants people to use CDB unmodified.
It doesn't look like it would work for our purposes with that extra 24 bytes per record.
-- sidney
