Abstract

Tsearch2, a contributed extension in PostgreSQL 7.4, provides fast indexed access for full text search queries. This talk will discuss Tsearch2's history and technology as well as how to install, configure, and optimize performance of the software. Tsearch2 supports relevance ranking, multilingual word stemming, HTML parsing, headline generation, boolean queries, stop words, and synonyms.

Tsearch2, written by Oleg Bartunov and Teodor Sigaev, utilizes advanced PostgreSQL features such as custom data types, operators, stored procedures, and GiST (Generalized Search Tree) indexes. A GiST index is a template indexing structure that allows domain experts to adhere to an interface and index custom data types. In particular, a custom data type, tsvector, is used to store a vector of word stems. A custom GiST index provides indexed access for the tsvector data type.