天天看點

PostgreSQL 函數式索引使用注意 - 暨非immutable函數不适合索引的原因

标簽

PostgreSQL , 表達式 , 函數穩定性 , immutable

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E8%83%8C%E6%99%AF 背景

PostgreSQL支援表達式索引,但是表達式必須是immutable的,也即是當輸入參數不變時,結果是永恒不變的。

因為當表達式涉及的變量不變時,索引本身不會變化。

給個例子,如果我們有一張表存儲了商品價格,另一張表存儲了商品折扣 ,如果我們想通過折扣後的價格範圍搜尋符合價格區間的商品ID,可以使用索引嗎?

表達式索引,可以。但是前提是:輸入一個商品ID時,商品原價永恒不變。

否則原價發生變化就可能出現索引内容與實際不一緻的問題。

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E4%BE%8B%E5%AD%90 例子

create extension btree_gist;  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%95%86%E5%93%81%E8%A1%A8 商品表

create table t_item (id int8 primary key, price jsonb);  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E6%8A%98%E6%89%A3%E8%A1%A8 折扣表

create table t_item_discount (id int8, ts daterange, country text, discount float4);  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E8%8E%B7%E5%8F%96%E5%95%86%E5%93%81%E6%8A%98%E5%90%8E%E4%BB%B7%E6%A0%BC%E7%9A%84%E5%87%BD%E6%95%B0 擷取商品折後價格的函數

create or replace function get_price(int8,text,float4) returns float8 as $$  
  select (price->>$2)::float8*$3 from t_item where id=$1;  
$$ language sql strict immutable;  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%87%BD%E6%95%B0%E7%B4%A2%E5%BC%95immutable%E5%87%BD%E6%95%B0 函數索引,immutable函數

create index idx_t_item_discount_1 on t_item_discount using gist (ts, country, get_price(id,country,discount));  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%86%99%E5%85%A5%E5%95%86%E5%93%81 寫入商品

insert into t_item values (1, jsonb '{"global":200, "china":150}');  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%86%99%E5%85%A5%E6%8A%98%E6%89%A3 寫入折扣

insert into t_item_discount values (1, daterange('2018-01-01', '2018-01-10'), 'global', 0.4);  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%BC%BA%E5%88%B6%E7%B4%A2%E5%BC%95%E6%89%AB%E6%8F%8F 強制索引掃描

set enable_bitmapscan=off;  
set enable_seqscan=off;  
  
postgres=# explain select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and get_price(id,country,discount)<300 and country='china';  
                                                                QUERY PLAN                                                                   
-------------------------------------------------------------------------------------------------------------------------------------------  
 Index Scan using idx_t_item_discount_1 on t_item_discount  (cost=0.12..8.40 rows=1 width=90)  
   Index Cond: ((ts @> '2018-01-01'::date) AND (country = 'china'::text) AND (get_price(id, country, discount) < '300'::double precision))  
(2 rows)  
  
postgres=# explain select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and country='china' and get_price(id,country,discount)<300;  
                                                                QUERY PLAN                                                                   
-------------------------------------------------------------------------------------------------------------------------------------------  
 Index Scan using idx_t_item_discount_1 on t_item_discount  (cost=0.12..8.40 rows=1 width=90)  
   Index Cond: ((ts @> '2018-01-01'::date) AND (country = 'china'::text) AND (get_price(id, country, discount) < '300'::double precision))  
(2 rows)  
  
  
  
postgres=# select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and country='global' and get_price(id,country,discount)<300;  
 ctid  |    get_price     | id |           ts            | country | discount   
-------+------------------+----+-------------------------+---------+----------  
 (0,1) | 80.0000011920929 |  1 | [2018-01-01,2018-01-10) | global  |      0.4  
(1 row)  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E4%BD%86%E6%98%AF%E5%A6%82%E6%9E%9C%E5%8E%9F%E4%BB%B7%E5%8F%98%E5%8C%96%E7%B4%A2%E5%BC%95%E5%B9%B6%E4%B8%8D%E4%BC%9A%E6%9B%B4%E6%96%B0 但是如果原價變化,索引并不會更新

postgres=# update t_item set price = jsonb '{"global":2000, "china":1500}' where id=1;  
UPDATE 1  
           

下面的結果顯然是錯誤的

postgres=# select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and country='global' and get_price(id,country,discount)<300;  
 ctid  |    get_price     | id |           ts            | country | discount   
-------+------------------+----+-------------------------+---------+----------  
 (0,1) | 800.000011920929 |  1 | [2018-01-01,2018-01-10) | global  |      0.4  
(1 row)  
           
postgres=# update t_item_discount set discount = discount where id=1;  
UPDATE 1  
postgres=# select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and country='global' and get_price(id,country,discount)<300;  
 ctid  |    get_price     | id |           ts            | country | discount   
-------+------------------+----+-------------------------+---------+----------  
 (0,2) | 800.000011920929 |  1 | [2018-01-01,2018-01-10) | global  |      0.4  
(1 row)  
           

隻有當表達式字段内容發生變化時,相應的表達式才會變化

postgres=# update t_item_discount set discount=discount+0.0000001 where id=1;  
UPDATE 1  
  
postgres=# select ctid,get_price(id,country,discount),* from t_item_discount where ts @> '2018-01-01'::date and country='global' and get_price(id,country,discount)<300;  
 ctid | get_price | id | ts | country | discount   
------+-----------+----+----+---------+----------  
(0 rows)  
  
postgres=# select float4send(discount),* from t_item_discount ;  
 float4send | id |           ts            | country | discount   
------------+----+-------------------------+---------+----------  
 \x3eccccd0 |  1 | [2018-01-01,2018-01-10) | global  |      0.4  
(1 row)  
           

https://github.com/digoal/blog/blob/master/201807/20180703_01.md#%E5%8F%82%E8%80%83 參考

《PostgreSQL Oracle 相容性之 - PL/SQL DETERMINISTIC 與PG函數穩定性(immutable, stable, volatile)》 《PostgreSQL 函數穩定性與constraint_excluded分區表邏輯推理過濾的CASE》 《函數穩定性講解 - retalk PostgreSQL function's [ volatile|stable|immutable ]》 《函數穩定性講解 - 函數索引思考, pay attention to function index used in PostgreSQL》 《函數穩定性講解 - Thinking PostgreSQL Function's Volatility Categories》

繼續閱讀