Parameterized Path 的例子

Improve the planner's ability to use nested loops with inner index scans (Tom Lane)



The new "parameterized path" mechanism allows inner index scans to use values from relations that are more than one join level up from the scan. This can greatly improve performance in situations where semantic restrictions (such as outer joins) limit the allowed join orderings.

http://www.postgresql.org/docs/current/static/release-9-2.html

数据准备：

postgres=# create table tst01(id integer);

CREATE TABLE

postgres=# 



postgres=# insert into tst01 values(generate_series(1,100000));

INSERT 0 100000

postgres=# 



postgres=# create index idx_tst01_id on tst01(id);

CREATE INDEX

postgres=# 

运行：

postgres=# prepare s(int) as select * from tst01 t where id < $1;

PREPARE

postgres=# explain execute s(2);

                                   QUERY PLAN                                    

---------------------------------------------------------------------------------

 Index Only Scan using idx_tst01_id on tst01 t  (cost=0.00..8.38 rows=1 width=4)

   Index Cond: (id < 2)

(2 rows)



postgres=# explain execute s(10000);

                                      QUERY PLAN                                       

---------------------------------------------------------------------------------------

 Index Only Scan using idx_tst01_id on tst01 t  (cost=0.00..337.64 rows=10130 width=4)

   Index Cond: (id < 10000)

(2 rows)



postgres=# explain execute s(1000000);

                          QUERY PLAN                           

---------------------------------------------------------------

 Seq Scan on tst01 t  (cost=0.00..1693.00 rows=100000 width=4)

   Filter: (id < 1000000)

(2 rows)



postgres=# explain execute s(100000);

                          QUERY PLAN                           

---------------------------------------------------------------

 Seq Scan on tst01 t  (cost=0.00..1693.00 rows=100000 width=4)

   Filter: (id < 100000)

(2 rows)



postgres=# 

这是一个小例子，而且还是一个有些特殊的例子。

对比一下在PostgreSQL9.1.0中的表现：

postgres=# create table tst01(id integer);

CREATE TABLE

postgres=# insert into tst01 values(generate_series(1,100000));

INSERT 0 100000

postgres=# create index idx_tst01_id on tst01(id);

CREATE INDEX

postgres=# prepare s(int) as select * from tst01 t where id < $1;

PREPARE

postgres=# explain execute s(2);

                                   QUERY PLAN                                   

 

--------------------------------------------------------------------------------

-

 Bitmap Heap Scan on tst01 t  (cost=626.59..1486.25 rows=33333 width=4)

   Recheck Cond: (id < $1)

   ->  Bitmap Index Scan on idx_tst01_id  (cost=0.00..618.26 rows=33333 width=0)

         Index Cond: (id < $1)

(4 rows)



postgres=# explain execute s(10000);

                                   QUERY PLAN                                   

 

--------------------------------------------------------------------------------

-

 Bitmap Heap Scan on tst01 t  (cost=626.59..1486.25 rows=33333 width=4)

   Recheck Cond: (id < $1)

   ->  Bitmap Index Scan on idx_tst01_id  (cost=0.00..618.26 rows=33333 width=0)

         Index Cond: (id < $1)

(4 rows)



postgres=# 

可以看到，在9.1里，是不区分状况，执行计划固定。