PostgreSQL pg_trgm 模糊搜索完全指南

SELECT show_trgm('iPhone');
-- 结果: {"  i"," ip","hon","iph","ne ","one","pho"}

CREATE EXTENSION IF NOT EXISTS pg_trgm;

CREATE TABLE products (
  tenant_id uuid,
  id integer,
  name text,
  description text,
  PRIMARY KEY (tenant_id, id)
);

-- GiST 索引：平衡型，更新快，占用空间小
CREATE INDEX trgm_idx_products_name 
ON products USING gist (name gist_trgm_ops);

-- GIN 索引：查询更快，但更新慢，占用更多空间
CREATE INDEX trgm_gin_idx_products_name 
ON products USING gin (name gin_trgm_ops);

-- 插入产品（注意第二条有拼写错误 "iPhne"）
INSERT INTO products (tenant_id, id, name, description) VALUES
  ('d1c06023-3421-4fbb-9dd1-c96e42d2fd02', 1, 'iPhone 13 Pro', 'Latest Apple smartphone'),
  ('d1c06023-3421-4fbb-9dd1-c96e42d2fd02', 2, 'iPhne 13', 'Budget Apple smartphone'),
  ('d1c06023-3421-4fbb-9dd1-c96e42d2fd02', 3, 'Samsung Galaxy S21', 'Android flagship phone');

-- 返回 0 到 1 之间的数字（1 = 完全相同）
SELECT similarity('iPhone', 'iPhne');
-- 结果: 0.5454545

-- 查找与 'iPhone' 相似的产品名
SELECT name, similarity(name, 'iPhone') AS sim
FROM products
WHERE name % 'iPhone'  -- 相似度超过阈值
ORDER BY sim DESC;

-- 查找包含 'phone' 的产品
SELECT name, similarity(name, 'phone') AS sim
FROM products
WHERE name %> 'phone'
ORDER BY sim DESC;

SET pg_trgm.similarity_threshold = 0.5;

SELECT name
FROM products
WHERE name % 'iPhone';

SELECT name, word_similarity('iPhone', name) as sim
FROM products
ORDER BY sim DESC;

SELECT name, strict_word_similarity('iPhone', name) as sim
FROM products
ORDER BY sim DESC;

SELECT similarity('iPhone 13 Pro', 'ipone');  -- 结果约 0.25

BEGIN;
SET LOCAL pg_trgm.similarity_threshold = 0.2;

SELECT name, similarity(name, 'ipone') AS score
FROM products
WHERE tenant_id = 'd1c06023-3421-4fbb-9dd1-c96e42d2fd02'
  AND name % 'ipone'
ORDER BY score DESC
LIMIT 5;

COMMIT;  -- 或 ROLLBACK，阈值会自动恢复

SELECT name, similarity(name, 'ipone') AS score
FROM products
WHERE tenant_id = 'd1c06023-3421-4fbb-9dd1-c96e42d2fd02'
  AND similarity(name, 'ipone') > 0.15  -- 自定义阈值
ORDER BY score DESC
LIMIT 5;

SELECT name, word_similarity('iPh', name) AS score
FROM products
WHERE tenant_id = 'd1c06023-3421-4fbb-9dd1-c96e42d2fd02'
  AND 'iPh' <% name  -- word_similarity 操作符
ORDER BY score DESC
LIMIT 10;

SELECT name
FROM products
WHERE tenant_id = 'd1c06023-3421-4fbb-9dd1-c96e42d2fd02'
  AND name ILIKE 'iPh%'  -- 大小写不敏感
ORDER BY name
LIMIT 10;

SELECT p1.name, p2.name, similarity(p1.name, p2.name) AS sim
FROM products p1
JOIN products p2 ON p1.id < p2.id
WHERE p1.tenant_id = p2.tenant_id
  AND p1.name % p2.name
  AND similarity(p1.name, p2.name) > 0.7
ORDER BY sim DESC;

WHERE LOWER(name) % LOWER('iphone')

CREATE INDEX trgm_gist_idx ON table_name USING gist (column_name gist_trgm_ops);

CREATE INDEX trgm_gin_idx ON table_name USING gin (column_name gin_trgm_ops);