Design of the indextype

The indextype implemented here, called sbtree, operates similar to btree indexes. It supports the evaluation of three user-defined operators

• gt(Greater Than)

• lt(Less Than)

• eq(EQuals)

These operators can operate on the operands of VARCHAR2 datatype.

The index data consists of records of the form <key, rid> where key is the value of the indexed column and rid is the row identifier of the corresponding row. To simplify the implementation of the indextype, the index data will be stored in a regular table. Thus, the index manipulation routines merely translate operations on the SBtree into operations on the table storing the index data. When a user creates a SBtree index, a table is created consisting of the indexed column and a rowid column. Inserts into the base table will cause appropriate insertions into the index table. Deletes and updates are handled similarly. When the SBtree is queried based on a user-defined operator (one of gt, lt and eq), an appropriate query will be issued against the index table to retrieve all the satisfying rows.