Postgres 不使用带有 LEFT JOIN 或 OR in where 的索引

Question

我遇到了一些与我有些相关的问题（例如1、2）。但他们都没有真正回答这个问题。

基本上我的问题是 Postgres 拒绝在某些看似简单的查询中使用索引。我设法将我的复杂查询归结为以下一些示例：

-- SETUP

CREATE TEMP TABLE test ( 
    id BIGSERIAL PRIMARY KEY,
    DATA INT 
);

INSERT INTO test(DATA)
SELECT RANDOM() FROM GENERATE_SERIES(1, 100000);

-- just in case the stats are outdated
VACUUM ANALYSE test;

-- QUERIES

-- correct but does a full table scan
EXPLAIN SELECT
    id,
    DATA
FROM
    test
WHERE
    (FALSE
    OR id IN (
    SELECT
        UNNEST(ARRAY[100]) 
    ));



-- simpler than above but still does a table scan
EXPLAIN SELECT
    id,
    DATA
FROM
    test
WHERE
    FALSE
    OR id IN (SELECT 100);

解释的输出：

QUERY PLAN                                                
----------------------------------------------------------
Seq Scan on test  (cost=0.01..1791.01 rows=50000 width=12)
  Filter: (hashed SubPlan 1)                              
  SubPlan 1                                               
    ->  Result  (cost=0.00..0.01 rows=1 width=4)          

-- uses PRIMARY KEY index correctly but limited to 1 ID
-- NOTE that it uses "=" instead of "IN"
EXPLAIN SELECT
    id,
    DATA
FROM
    test
WHERE
    FALSE
    OR id = (SELECT 100 LIMIT 1);

QUERY PLAN                                                           
---------------------------------------------------------------------
Index Scan using test_pkey on test  (cost=0.30..8.32 rows=1 width=12)
  Index Cond: (id = $0)                                              
  InitPlan 1 (returns $0)                                            
    ->  Limit  (cost=0.00..0.01 rows=1 width=4)                      
          ->  Result  (cost=0.00..0.01 rows=1 width=4)               

正确使用 PRIMARY KEY 索引，但我需要 LEFT JOIN 而不是 INNER JOIN。 -- 见下一个例子。

EXPLAIN
WITH ids AS (
SELECT
    UNNEST(ARRAY[100]) AS ids )
SELECT
    id,
    DATA
FROM
    test
INNER JOIN ids ON
    id = ids.ids;

QUERY PLAN                                                                 
---------------------------------------------------------------------------
Nested Loop  (cost=0.29..8.34 rows=1 width=12)                             
  ->  ProjectSet  (cost=0.00..0.02 rows=1 width=4)                         
        ->  Result  (cost=0.00..0.01 rows=1 width=0)                       
  ->  Index Scan using test_pkey on test  (cost=0.29..8.31 rows=1 width=12)
        Index Cond: (id = (unnest('{100}'::integer[])))                    

我需要一个左连接，但这会进行全表扫描

EXPLAIN WITH ids AS (
  SELECT 100 AS id 
)
SELECT
    test.id,
    DATA
FROM
    test
LEFT JOIN ids ON
    test.id = ids.id;

QUERY PLAN                                                 
-----------------------------------------------------------
Seq Scan on test  (cost=0.00..1541.00 rows=100000 width=12)

所以我的问题：

1. 为什么FALSE OR ... IN ... 阻止使用索引，而FALSE OR ... = ... 没问题？
2. 为什么在我的示例中 INNER JOIN 使用索引而 LEFT JOIN 没有？
3. 是否有某种方法可以在不使用 UNION 或复制整个查询的情况下重写第一个查询？ （我的实际查询要复杂得多）

Answer 1

1. 将指定的选择移到表格附近，例如：

   SELECT
     id, DATA
   FROM
     test, (SELECT UNNEST(ARRAY[100]) as id1) as sel
   WHERE
     (FALSE OR id = sel.id1);
   

2. 使用 UNION ALL 的中断或子句如下：

   SELECT
     id, DATA
   FROM
     test
   WHERE
     FALSE -- first or predicate
   UNION ALL
   SELECT
     id, DATA
   FROM
     test
   WHERE id in (select UNNEST(ARRAY[100]);