TIN Full-Text Search Extension Released for Postgres and Neki Databases

Introduction of TIN Extension

A new full-text search extension, TIN (Text INdex), has been launched for Postgres and Neki databases. This release aims to provide comprehensive full-text search functionalities that were previously unmet by existing solutions within the Postgres ecosystem.

Key Features and Capabilities

TIN supports a range of advanced search features, including boolean expressions, phrase queries, and span queries. It also incorporates fuzzy, wildcard, and regular-expression matching for terms, along with case and accent folding. The extension is designed to handle COUNT(*) queries and BM25-scored top-k queries efficiently.

Addressing Existing Gaps

The developers state that while other text-search indexes exist for Postgres, none fully met the requirements for robust full-text search, especially concerning concurrent operations. TIN is built to support joins, complex WHERE clauses across different column types, continuous updates, replication, backups, and correct transaction visibility, even with ongoing data modifications.

Application in Various Use Cases

TIN can be applied in various scenarios, such as e-commerce platforms for product searches, legal discovery platforms for document retrieval, and photo tagging services for counting specific tags. The extension is designed to maintain performance even when applications are simultaneously inserting, updating, and deleting documents.

Performance and Availability

Benchmarks were conducted to evaluate TIN's performance across different query types and workloads, including those with concurrent write operations. The extension is immediately available as a General Availability (GA) release for all Postgres and Neki databases.