textsearch_intl

textsearch_intl : Multilingual full text search with ICU segmentation, normalization and n-gram matching

Overview

Packages

Source

pig build pkg textsearch_intl;		# build rpm/deb

Install

Make sure PGDG and PIGSTY repo available:

pig repo add pgsql -u   # add both repo and update cache

Install this extension with pig:

pig install textsearch_intl;		# install via package name, for the active PG version

pig install textsearch_intl -v 18;   # install for PG 18
pig install textsearch_intl -v 17;   # install for PG 17
pig install textsearch_intl -v 16;   # install for PG 16
pig install textsearch_intl -v 15;   # install for PG 15
pig install textsearch_intl -v 14;   # install for PG 14

Create this extension with:

CREATE EXTENSION textsearch_intl CASCADE; -- requires unaccent

Usage

Sources:

• README.md
• textsearch_intl–1.0.sql
• textsearch_intl.control
• META.json
• LICENSE
• Rename commit
• PGXN textsearch_intl 1.0.1

textsearch_intl combines ICU word segmentation, normalization and n-gram matching for multilingual text search. It requires PostgreSQL 14 or later, a UTF8 database, ICU and unaccent; upstream CI covers PostgreSQL 14–18.

Name and Version

The project renamed icu_parser to textsearch_intl on 2026-10-09 and restarted versioning. The renamed PGXN distribution 1.0.1 was published on 2026-10-22. Its SQL/control version remains 1.0; the distribution number is not a SQL upgrade target. There is no direct SQL upgrade script from the old name: inspect dependent objects and plan their recreation before migrating.

Core Workflow

CREATE EXTENSION textsearch_intl CASCADE;
SELECT alias, token FROM ts_debug('intl_simple', '東京駅前ラーメン店 ร้านกาแฟ Café');
SELECT to_tsvector('intl_search', 'Café') @@ phraseto_tsquery('intl_search', 'cafe');
SELECT to_tsvector('intl_search_ngram', '행복한편의점')
       @@ phraseto_tsquery('intl_search_ngram', '편의점');
SELECT intl_icu_version();

Search Configurations

• intl_simple uses the intl parser and preserves accents and character widths.
• intl_search combines intl_normalize, unaccent and intl_join for normalized whole-word search. Word joining is off by default.
• intl_search_ngram uses intl_runs and intl_ngram to match inside Korean/CJK tokens and unspaced Thai, Lao, Khmer and Burmese runs. N-grams can overmatch and increase index size.

Operational Boundaries

Install with a superuser; the extension is relocatable, is not marked trusted and needs no preload. Prefer phraseto_tsquery() when word order matters. intl_icu_version() reports the loaded ICU version; rebuild affected indexes after an ICU major-version change. ICU segmentation is not Japanese morphological analysis, and partial CJK prefix queries remain unreliable.