ИСТИНА

We present a prototype of new access method, heavily based on GIN and optimized for efficient indexing of nested structures like hstore and json(b). Introducing of the nested hstore and jsonb in PostgreSQL brought new challenge to the developers, namely, an efficient indexing of hierarchical keys. Those keys are consist of duplicated strings, which made index to be uselessly huge if store key-value pairs independently. We propose to replace btree data structure, which used in GIN to index keys, by digital tree. To do this in 'right way', we would like to experiment with hybrid access method based on of SP-GiST and GIN. This is a first step in making GIN more flexible to support richer set of queries. In principle, one could be able to use other than btree data structure to index not just keys, but also the posting lists.