DBX indexing internals

Pagesize determines the size of each page in an index

Cachesize determines the number of pages maintained in a cache when
writing or reading.

Pages are one of the following types, defined as a value in the page:

Value Type      Description

    1 root      Root page
    2 internal  Internal page
    4 leaf      Leaf node
    8 bucket    Bucket node for string key
   16 overflow  Overflow page (not used)
   32 pribucket Primary key bucket
   64 secbucket Secondary key bucket
  128 numbucket Numeric key bucket

Indexed key terms are stored in a single buckets.

For the main keys (id, ac, sv) the key defined as a "hybrid key", a
combination of the key as a string and a possible list of duplicates
with te same key.

The index data is stored in a bucket, with the
following fields:

 int      dbno: 0..n number of the database source file
 int      dups: 0 for a unique identifier, or number of duplicates
long    offset: file position in sequence file
long refoffset: file position in reference file (if any - e.g. GCG)

with information about duplicates (entries with the same identifier)
stored in numbuckets

long    offset: file position in sequence file
long refoffset: file position in reference file (if any - e.g. GCG)
 int      dbno: 0..n number of the database source file

For the secondary keys (key, des, org) the keys are stored in a secbucket