使用自动更正快速搜索（GIN INDEX 和 PG_TRGM 扩展名）

Question

我正在测试一个简单的搜索机制来处理 SMALL typos/misspellings。类似于自动更正机制。

我为此苦苦挣扎。所以我正在创建一个函数 (pl/pgsql) 来处理这个问题，我在 SUPABASE.IO、PostgreSQL 13.3（类似于 RDS）上运行。

我愿意：

将 returned 结果限制为仅高度相似的电子邮件地址，比如相似度 > 0.7；
使用 INDEX，因为实际的电子邮件列表将达到数千万，因此它必须 return 一秒钟之内。

DROP TABLE IF EXISTS email;
CREATE TABLE email (
  email_address TEXT NOT NULL UNIQUE,
  person_id UUID NOT NULL, 
  CONSTRAINT email_pk PRIMARY KEY (email_address)
);

DROP INDEX IF EXISTS email_address_trigram_idx;
CREATE INDEX email_address_trigram_idx ON email USING gin(email_address gin_trgm_ops);

INSERT INTO email(email_address, person_id) VALUES
  ('test100@gmail.com', uuid_generate_v4())
, ('100test@gmail.com', uuid_generate_v4())
, ('testoo1000@gmail.com', uuid_generate_v4())
, ('test1001@gmail.com', uuid_generate_v4())
, ('test100@gmial.com', uuid_generate_v4())
, ('test200@gmail.com', uuid_generate_v4())
, ('200test@gmail.com', uuid_generate_v4())
, ('testoo2000@gmail.com', uuid_generate_v4())
, ('test2002@gmail.com', uuid_generate_v4())
, ('test200@gmial.com', uuid_generate_v4())
, ('test300@gmail.com', uuid_generate_v4())
, ('300test@gmail.com', uuid_generate_v4())
, ('testoo3000@gmail.com', uuid_generate_v4())
, ('test3003@gmail.com', uuid_generate_v4())
, ('test300@gmial.com', uuid_generate_v4())
, ('test400@gmail.com', uuid_generate_v4())
, ('400test@gmail.com', uuid_generate_v4())
, ('testoo4000@gmail.com', uuid_generate_v4())
, ('test4004@gmail.com', uuid_generate_v4())
, ('test400@gmial.com', uuid_generate_v4())
, ('tset100@gmail.com', uuid_generate_v4())
, ('100tset@gmail.com', uuid_generate_v4())
, ('tsetoo1000@gmail.com', uuid_generate_v4())
, ('tset1001@gmail.com', uuid_generate_v4())
, ('tset100@gmial.com', uuid_generate_v4())
, ('tset200@gmail.com', uuid_generate_v4())
, ('200tset@gmail.com', uuid_generate_v4())
, ('tsetoo2000@gmail.com', uuid_generate_v4())
, ('tset2002@gmail.com', uuid_generate_v4())
, ('tset200@gmial.com', uuid_generate_v4())
, ('tset300@gmail.com', uuid_generate_v4())
, ('300tset@gmail.com', uuid_generate_v4())
, ('tsetoo3000@gmail.com', uuid_generate_v4())
, ('tset3003@gmail.com', uuid_generate_v4())
, ('tset300@gmial.com', uuid_generate_v4())
, ('tset400@gmail.com', uuid_generate_v4())
, ('400tset@gmail.com', uuid_generate_v4())
, ('tsetoo4000@gmail.com', uuid_generate_v4())
, ('tset4004@gmail.com', uuid_generate_v4())
, ('tset400@gmial.com', uuid_generate_v4())
, ('different_email@yahoo.com', uuid_generate_v4());

SET pg_trgm.similarity_threshold = 0.8; -- This doesn't seem to affect my queries

SELECT *, similarity('tesd100@gmail.com', email_address)
FROM email
WHERE email_address % 'tesd100@gmail.com';

我想要一种既能快速搜索又能容忍搜索中的一些小拼写错误的方法。

Answer 1

首先，您的 table 定义在 (email_address) 上创建了两个个唯一索引。不。删除 UNIQUE 约束，保留 PK:

CREATE TABLE email (
  email_address text PRIMARY KEY
, person_id uuid NOT NULL  -- bigint?
);

（也不确定为什么 person_id 需要 uuid。世界上几乎没有足够的人来证明超过 bigint。）

接下来，既然你想...

LIMIT the returned results to only the highly similar email addresses,

我建议最近邻搜索。为此目的创建一个 GiST 索引而不是 GIN:

CREATE INDEX email_address_trigram_gist_idx ON email USING gist (email_address gist_trgm_ops);

并使用这样的查询：

SELECT *, similarity('tesd100@gmail.com', email_address)
FROM   email
WHERE  email_address % 'tesd100@gmail.com'
ORDER  BY email_address <-> 'tesd100@gmail.com'  -- note the use of the operator <->
LIMIT  10;

引用 the manual:

This can be implemented quite efficiently by GiST indexes, but not by GIN indexes. It will usually beat the first formulation when only a small number of the closest matches is wanted.

在使用较小的 LIMIT 时，可能不需要将 pg_trgm.similarity_threshold 设置得非常高，因为此查询会首先为您提供最佳匹配。

使用自动更正快速搜索（GIN INDEX 和 PG_TRGM 扩展名）

Quick Search with autocorrect (GIN INDEX and PG_TRGM extension)

postgresql

plpgsql

fuzzy-search

autocorrect

supabase