PostgreSQL Extension for Full-Text Search in Any Language
summary
A PostgreSQL parser extension called Chamkho PG implements a customizable tokenizer (word lists, regex replacements, cluster rules) that handles Thai and many other languages including mixed-script tokens and stores rules as plain text so no recompilation is required. Core is port to Rust and benchmarks on Hetzner show roughly 2× faster than an alternate Rust algorithm. An ICUx backup parser is available, and planned work includes a C interface and storing rules in DB tables for managed deployments.
AI-written summary · about
chapters (10)
- 0:00Introduction and speaker intro
- 1:02Speaker background and motivation
- 2:09PostgreSQL full-text search limitations
- 4:00Related work and existing extensions
- 5:46Introducing Chamkho PG extension
- 7:13Why use Chamkho and performance
- 8:15Customization with word lists and rules
- 10:03Backup plan with ICU integration
- 10:25Future work and compatibility
- 11:09Q&A on stemming and performance
related talks
43mFull-Text-Search Explained from A to Z with French Food
34mFull-Text Search in PostgreSQL
28mElasticsearch-Quality Full-Text Search in Postgres via Tantivy
41mPostgreSQL is (probably) all you need
31mPostgres without SQL: Natural language queries using GPT-3 & Rust
53mModern PostgreSQL for AI...
59mFull-text search - seek and ye shall find
33mPostgres, the Best Tool You're Already Using