-- kham_pg--0.6.0.sql
--
-- Creates the kham text search parser, a pass-through dictionary, and a
-- ready-to-use text search configuration for Thai documents.
--
-- Token types produced by the parser:
--   1  thai     Thai word token
--   2  latin    Latin script token
--   3  number   Numeric token
--   4  punct    Punctuation
--   5  emoji    Emoji token
--   6  unknown  Unknown / OOV token
--   7  named    Named entity token (person, place, organisation)

-- Guard against accidental direct \i load
\echo Use "CREATE EXTENSION kham_pg" to load this file. \quit

-- ── C function registrations ─────────────────────────────────────────────────
-- PostgreSQL resolves these to symbols in MODULE_PATHNAME (= $libdir/kham_pg).
-- Signatures match what ts_parse.c passes:
--   startfunc  (internal, int4)                        → internal
--   gettoken   (internal, internal, internal)          → int4
--   endfunc    (internal)                              → void
--   lextypes   (internal)                              → internal
--   headline   (internal, internal, tsquery)           → internal
--   dict_lexize (internal, internal, internal, internal) → internal

CREATE FUNCTION kham_start(internal, int4)
    RETURNS internal
    AS 'MODULE_PATHNAME', 'kham_start'
    LANGUAGE c STRICT;

CREATE FUNCTION kham_gettoken(internal, internal, internal)
    RETURNS internal
    AS 'MODULE_PATHNAME', 'kham_gettoken'
    LANGUAGE c STRICT;

CREATE FUNCTION kham_end(internal)
    RETURNS void
    AS 'MODULE_PATHNAME', 'kham_end'
    LANGUAGE c STRICT;

CREATE FUNCTION kham_lextypes(internal)
    RETURNS internal
    AS 'MODULE_PATHNAME', 'kham_lextypes'
    LANGUAGE c STRICT;

CREATE FUNCTION kham_headline(internal, internal, tsquery)
    RETURNS internal
    AS 'MODULE_PATHNAME', 'kham_headline'
    LANGUAGE c STRICT;

CREATE FUNCTION kham_dict_lexize(internal, internal, internal, internal)
    RETURNS internal
    AS 'MODULE_PATHNAME', 'kham_dict_lexize'
    LANGUAGE c;  -- NOT STRICT: dict state (arg0) is NULL when no INIT is provided

-- ── Parser ───────────────────────────────────────────────────────────────────

CREATE TEXT SEARCH PARSER kham (
    START    = kham_start,
    GETTOKEN = kham_gettoken,
    END      = kham_end,
    LEXTYPES = kham_lextypes,
    HEADLINE = kham_headline
);

-- ── Dictionary template — soundex + RTGS expansion ─────────────────────────
-- kham_fts_template: custom dictionary that expands Thai / Named tokens into
--   [word, lk82_soundex_code, rtgs_romanization]
-- at the same tsvector position, enabling phonetic and Latin-script search.

CREATE TEXT SEARCH TEMPLATE kham_fts_template (
    LEXIZE = kham_dict_lexize
);

CREATE TEXT SEARCH DICTIONARY kham_fts_dict (
    TEMPLATE = kham_fts_template
);

-- ── Dictionary ───────────────────────────────────────────────────────────────
-- Simple pass-through: lowercases Latin/Number tokens; Thai is returned as-is
-- (Thai script is not case-folded by the simple template).

CREATE TEXT SEARCH DICTIONARY kham_dict (
    TEMPLATE = simple
);

-- ── Configuration ────────────────────────────────────────────────────────────

CREATE TEXT SEARCH CONFIGURATION kham (
    PARSER = kham
);

-- Map substantive token types through kham_dict.
-- Punctuation and emoji are omitted intentionally — no MAPPING means PG
-- discards those token types during indexing.

-- Thai and Named tokens go through kham_fts_dict which expands each token
-- to [word, lk82_soundex_code, rtgs_romanization] at the same position.
ALTER TEXT SEARCH CONFIGURATION kham
    ADD MAPPING FOR thai    WITH kham_fts_dict;

ALTER TEXT SEARCH CONFIGURATION kham
    ADD MAPPING FOR named   WITH kham_fts_dict;

-- Latin, Number, Unknown: simple lowercase pass-through.
ALTER TEXT SEARCH CONFIGURATION kham
    ADD MAPPING FOR latin   WITH kham_dict;

ALTER TEXT SEARCH CONFIGURATION kham
    ADD MAPPING FOR number  WITH kham_dict;

ALTER TEXT SEARCH CONFIGURATION kham
    ADD MAPPING FOR unknown WITH kham_dict;