with MATERIALIZE inline cardinality 联合使用 literals vs bind variables

jnrjian2026-04-11 13:26

Applies To

All Users

Summary

NOTE: In the images and/or the document content below, the user information and data used represents fictitious data from the Oracle sample schema(s) or Public Documentation delivered with an Oracle database product. Any similarity to actual persons, living or dead, is purely coincidental and not intended in any manner.

If similar subqueries appear more than once in a sql, it is worth trying to rewrite the query using WITH subquery.
There are some things we should consider in order to get the benefit from it.

Solution

Subquery Factoring

Subquery Factoring or WITH clause enables you to assign a name to a subquery block that can then be referenced in multiple places in the query.
This subquery will be transformed as either a temporary table (materialization) or an inline view

Things to Consider when Deciding to Use Subquery Factoring

  • Subquery Factoring can simplify the query statement making the format of the query clearer and more readable.
  • As it eliminates the repeated scan of the same table, it may reduce physical I/O and logical reads.
    This is more noticeable when retrieving a large amount of data.
  • It causes some recursive calls and "db block gets"(reads in current mode), and generates some redo data because it creates a temporary table during execution.
  • It runs faster only if the total I/O is less than the plan that does not use materialization

Examples of Subquery Factoring

Example 1
SELECT ...
FROM A, B, C
WHERE ...
AND A.a in (select col from T1, T2, T3 where ...)
AND B.b in (select col from T1, T2, T3 where ...)
AND C.c in (select col from T1, T2, T3 where ...);

can be written as:
WITH S AS (select col from T1, T2, T3 where ...)
SELECT ...
FROM A, B, C
WHERE ...
AND A.a in (select col from S)
AND B.b in (select col from S)
AND C.c in (select col from S);

Example 2
SELECT ...
FROM (select ... from T1 ...) S1,
(select ... from T2 ...) S2,
(select ... from T3 ...) S3
WHERE ... ;

can be written as:
WITH S1 AS (select ... from T1 ...),
S2 AS (select ... from T2 ...),
S3 AS (select ... from T3 ...)
SELECT ...
FROM S1, S2, S3
WHERE ... ;

Influencing the Plan with Hints and Parameters

The decision whether to transfer the subquery into an inline view or a temporary table is done before the optimization and is determined by certain rules. However the decision can be influenced by including a hint in the WITH clause or by setting the parameter "_with_subquery".

Hints

  • /*+ MATERIALIZE */ : The subquery is materialized into a temporary table
    WITH S AS (select /*+ MATERIALIZE */ col from T1, T2, T3 where ...)
    SELECT ...
    FROM A, B, C
    WHERE ...
    AND A.a in (select col from S)
    AND B.b in (select col from S)
    AND C.c in (select col from S);

  • /*+ INLINE */ : The subquery is inlined and not materialized
    WITH S AS (select /*+ INLINE */ col from T1, T2, T3 where ...)
    SELECT ...
    FROM A, B, C
    WHERE ...
    AND A.a in (select col from S)
    AND B.b in (select col from S)
    AND C.c in (select col from S);

Optimizer Parameter

The parameter "_WITH_SUBQUERY" can have the following values:

  • optimizer : Let the optimizer choose (default)
  • materialize: Always materialize the WITH subquery
  • inline : Always inline.

The parameter can be set in pfile/spfile, at system level or at session level.

Demonstration

This test was done in SH demo schema.

Create 2 tables PRODUCTS2 and PRODUCTS3
connect sh/sh

create table products2 as select * from products;
create table products3 as select * from products;

Query 1 - Without Subquery Factoring

select P.prod_name, S.sale_count
from products P, (select prod_id, count(*) sale_count from sales group by prod_id) S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products2 P, (select prod_id, count(*) sale_count from sales group by prod_id) S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products3 P, (select prod_id, count(*) sale_count from sales group by prod_id) S
where P.prod_id = S.prod_id;

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
| 0 | SELECT STATEMENT | | 216 | 13536 | 862 (97)| 00:00:01 | | |
| 1 | UNION-ALL | | | | | | | |
|* 2 | HASH JOIN | | 72 | 4032 | 287 (90)| 00:00:01 | | |
| 3 | TABLE ACCESS FULL | PRODUCTS | 72 | 2160 | 2 (0)| 00:00:01 | | |
| 4 | VIEW | | 72 | 1872 | 285 (90)| 00:00:01 | | |
| 5 | HASH GROUP BY | | 72 | 288 | 285 (90)| 00:00:01 | | |
| 6 | PARTITION RANGE ALL | | 918K| 3589K| 29 (0)| 00:00:01 | 1 | 28 |
| 7 | BITMAP CONVERSION COUNT | | 918K| 3589K| 29 (0)| 00:00:01 | | |
| 8 | BITMAP INDEX FAST FULL SCAN| SALES_PROD_BIX | | | | | 1 | 28 |
|* 9 | HASH JOIN | | 72 | 4752 | 287 (90)| 00:00:01 | | |
| 10 | TABLE ACCESS FULL | PRODUCTS2 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 11 | VIEW | | 72 | 1872 | 285 (90)| 00:00:01 | | |
| 12 | HASH GROUP BY | | 72 | 288 | 285 (90)| 00:00:01 | | |
| 13 | PARTITION RANGE ALL | | 918K| 3589K| 29 (0)| 00:00:01 | 1 | 28 |
| 14 | BITMAP CONVERSION COUNT | | 918K| 3589K| 29 (0)| 00:00:01 | | |
| 15 | BITMAP INDEX FAST FULL SCAN| SALES_PROD_BIX | | | | | 1 | 28 |
|* 16 | HASH JOIN | | 72 | 4752 | 287 (90)| 00:00:01 | | |
| 17 | TABLE ACCESS FULL | PRODUCTS3 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 18 | VIEW | | 72 | 1872 | 285 (90)| 00:00:01 | | |
| 19 | HASH GROUP BY | | 72 | 288 | 285 (90)| 00:00:01 | | |
| 20 | PARTITION RANGE ALL | | 918K| 3589K| 29 (0)| 00:00:01 | 1 | 28 |
| 21 | BITMAP CONVERSION COUNT | | 918K| 3589K| 29 (0)| 00:00:01 | | |
| 22 | BITMAP INDEX FAST FULL SCAN| SALES_PROD_BIX | | | | | 1 | 28 |

Predicate Information (identified by operation id):

2 - access("P"."PROD_ID"="S"."PROD_ID")
9 - access("P"."PROD_ID"="S"."PROD_ID")
16 - access("P"."PROD_ID"="S"."PROD_ID")

Statistics

0 recursive calls
0 db block gets
251 consistent gets
0 physical reads
0 redo size
9191 bytes sent via SQL*Net to client
573 bytes received via SQL*Net from client
16 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
216 rows processed

Query 1 - With Subquery Factoring

set autotrace on

with V_SALE
as (select prod_id, count(*) sale_count from sales group by prod_id)
select P.prod_name, S.sale_count
from products P, V_SALE S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products2 P, V_SALE S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products3 P, V_SALE S
where P.prod_id = S.prod_id;

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
| 0 | SELECT STATEMENT | | 216 | 13536 | 14 (72)| 00:00:01 | | |
| 1 | TEMP TABLE TRANSFORMATION | | | | | | | |
| 2 | LOAD AS SELECT | SYS_TEMP_0FD9D6608_1A3773F | | | | | | |
| 3 | HASH GROUP BY | | 72 | 288 | 285 (90)| 00:00:01 | | |
| 4 | PARTITION RANGE ALL | | 918K| 3589K| 29 (0)| 00:00:01 | 1 | 28 |
| 5 | BITMAP CONVERSION COUNT | | 918K| 3589K| 29 (0)| 00:00:01 | | |
| 6 | BITMAP INDEX FAST FULL SCAN| SALES_PROD_BIX | | | | | 1 | 28 |
| 7 | UNION-ALL | | | | | | | |
|* 8 | HASH JOIN | | 72 | 4032 | 5 (20)| 00:00:01 | | |
| 9 | TABLE ACCESS FULL | PRODUCTS | 72 | 2160 | 2 (0)| 00:00:01 | | |
| 10 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 11 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6608_1A3773F | 72 | 288 | 2 (0)| 00:00:01 | | |
|* 12 | HASH JOIN | | 72 | 4752 | 5 (20)| 00:00:01 | | |
| 13 | TABLE ACCESS FULL | PRODUCTS2 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 14 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 15 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6608_1A3773F | 72 | 288 | 2 (0)| 00:00:01 | | |
|* 16 | HASH JOIN | | 72 | 4752 | 5 (20)| 00:00:01 | | |
| 17 | TABLE ACCESS FULL | PRODUCTS3 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 18 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 19 | TABLE ACCESS FULL | SYS_TEMP_0FD9D6608_1A3773F | 72 | 288 | 2 (0)| 00:00:01 | | |

Predicate Information (identified by operation id):

8 - access("P"."PROD_ID"="S"."PROD_ID")
12 - access("P"."PROD_ID"="S"."PROD_ID")
16 - access("P"."PROD_ID"="S"."PROD_ID")

Statistics

2 recursive calls
8 db block gets
114 consistent gets
1 physical reads
804 redo size
9191 bytes sent via SQL*Net to client
573 bytes received via SQL*Net from client
16 SQL*Net roundtrips to/from client
0 sorts (memory)
0 sorts (disk)
216 rows processed

Query 2 - Without Subquery Factoring

select P.prod_name, S.sale_count
from products P, (select prod_id, count(*) sale_count from sales where amount_sold <= 10 group by prod_id) S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products2 P, (select prod_id, count(*) sale_count from sales where amount_sold <= 10 group by prod_id) S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products3 P, (select prod_id, count(*) sale_count from sales where amount_sold <= 10 group by prod_id) S
where P.prod_id = S.prod_id;

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
| 0 | SELECT STATEMENT | | 216 | 13536 | 871 (78)| 00:00:01 | | |
| 1 | UNION-ALL | | | | | | | |
|* 2 | HASH JOIN | | 72 | 4032 | 290 (34)| 00:00:01 | | |
| 3 | TABLE ACCESS FULL | PRODUCTS | 72 | 2160 | 2 (0)| 00:00:01 | | |
| 4 | VIEW | | 72 | 1872 | 288 (34)| 00:00:01 | | |
| 5 | HASH GROUP BY | | 72 | 648 | 288 (34)| 00:00:01 | | |
| 6 | PARTITION RANGE ALL| | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 7 | TABLE ACCESS FULL | SALES | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 8 | HASH JOIN | | 72 | 4752 | 290 (34)| 00:00:01 | | |
| 9 | TABLE ACCESS FULL | PRODUCTS2 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 10 | VIEW | | 72 | 1872 | 288 (34)| 00:00:01 | | |
| 11 | HASH GROUP BY | | 72 | 648 | 288 (34)| 00:00:01 | | |
| 12 | PARTITION RANGE ALL| | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 13 | TABLE ACCESS FULL | SALES | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 14 | HASH JOIN | | 72 | 4752 | 290 (34)| 00:00:01 | | |
| 15 | TABLE ACCESS FULL | PRODUCTS3 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 16 | VIEW | | 72 | 1872 | 288 (34)| 00:00:01 | | |
| 17 | HASH GROUP BY | | 72 | 648 | 288 (34)| 00:00:01 | | |
| 18 | PARTITION RANGE ALL| | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 19 | TABLE ACCESS FULL | SALES | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |

Predicate Information (identified by operation id):

2 - access("P"."PROD_ID"="S"."PROD_ID")
7 - filter("AMOUNT_SOLD"<=10)
8 - access("P"."PROD_ID"="S"."PROD_ID")
13 - filter("AMOUNT_SOLD"<=10)
14 - access("P"."PROD_ID"="S"."PROD_ID")
19 - filter("AMOUNT_SOLD"<=10)

Statistics

0 recursive calls
0 db block gets
4919 consistent gets
4857 physical reads
0 redo size
1675 bytes sent via SQL*Net to client
430 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
30 rows processed

Query 2 - With Subquery Factoring

with V_SALE
as (select prod_id, count(*) sale_count from sales where amount_sold <= 10 group by prod_id)
select P.prod_name, S.sale_count
from products P, V_SALE S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products2 P, V_SALE S
where P.prod_id = S.prod_id
union all
select P.prod_name, S.sale_count
from products3 P, V_SALE S
where P.prod_id = S.prod_id;

| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time | Pstart| Pstop |
| 0 | SELECT STATEMENT | | 216 | 13536 | 14 (72)| 00:00:01 | | |
| 1 | TEMP TABLE TRANSFORMATION | | | | | | | |
| 2 | LOAD AS SELECT | SYS_TEMP_0FD9D660A_1A3773F | | | | | | |
| 3 | HASH GROUP BY | | 72 | 648 | 288 (34)| 00:00:01 | | |
| 4 | PARTITION RANGE ALL | | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
|* 5 | TABLE ACCESS FULL | SALES | 2118 | 19062 | 287 (34)| 00:00:01 | 1 | 28 |
| 6 | UNION-ALL | | | | | | | |
|* 7 | HASH JOIN | | 72 | 4032 | 5 (20)| 00:00:01 | | |
| 8 | TABLE ACCESS FULL | PRODUCTS | 72 | 2160 | 2 (0)| 00:00:01 | | |
| 9 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 10 | TABLE ACCESS FULL | SYS_TEMP_0FD9D660A_1A3773F | 72 | 648 | 2 (0)| 00:00:01 | | |
|* 11 | HASH JOIN | | 72 | 4752 | 5 (20)| 00:00:01 | | |
| 12 | TABLE ACCESS FULL | PRODUCTS2 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 13 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 14 | TABLE ACCESS FULL | SYS_TEMP_0FD9D660A_1A3773F | 72 | 648 | 2 (0)| 00:00:01 | | |
|* 15 | HASH JOIN | | 72 | 4752 | 5 (20)| 00:00:01 | | |
| 16 | TABLE ACCESS FULL | PRODUCTS3 | 72 | 2880 | 2 (0)| 00:00:01 | | |
| 17 | VIEW | | 72 | 1872 | 2 (0)| 00:00:01 | | |
| 18 | TABLE ACCESS FULL | SYS_TEMP_0FD9D660A_1A3773F | 72 | 648 | 2 (0)| 00:00:01 | | |

Predicate Information (identified by operation id):

5 - filter("AMOUNT_SOLD"<=10)
7 - access("P"."PROD_ID"="S"."PROD_ID")
11 - access("P"."PROD_ID"="S"."PROD_ID")
15 - access("P"."PROD_ID"="S"."PROD_ID")

Statistics

2 recursive calls
8 db block gets
1657 consistent gets
1620 physical reads
804 redo size
1675 bytes sent via SQL*Net to client
430 bytes received via SQL*Net from client
3 SQL*Net roundtrips to/from client
1 sorts (memory)
0 sorts (disk)
30 rows processed

Documentation

Oracle® Database SQL Language Reference
12c Release 1 (12.1)
subquery_factoring_clause

Restriction

Distributed transactions are not supported for temporary tables.
Query having WITH clause will be transformed as either a temporary table (materialization) or an inline view.

Note: The following fixes should be applied if query would be transformed as a temporary table (materialization):

<Document 9399589.8> "WITH" subqueries cannot be materialized inside a global transaction
9706532.8 "WITH" subqueries cannot be materialized inside a global transaction (or PLSQL RPC over DBlink)

如果使用变量，Oracle不知道with as 返回结果，可以使用

SELECT /*+ cardinality(mem,200000) */ 明确告知Oracle返回多少行，

相关推荐
忧郁的紫菜8 小时前
WCF+JSON+实体对象与WebService+DataSet效率大比拼
数据库·oracle·json
标致的自行车11 小时前
Entity Framework之问题收集
jvm·数据库·oracle
spider_xcxc1 天前
Redis 数据库高质量实践指南(一)
运维·数据库·redis·oracle·云计算
落叶-IT1 天前
Java异常处理深度实战教程：异常传播的失败场景分析
数据库·oracle
水木流年追梦1 天前
agent面试必备31- AI Agent 核心进阶：工具路由（Tool Routing）
数据库·人工智能·oracle·面试·职场和发展·embedding
weixin_397574092 天前
框架v5多数据源让AI查遍企业数据
数据库·人工智能·oracle
oradh2 天前
Oracle RAC集群环境下正常启停数据库
数据库·oracle·rac集群环境正常启停数据库·rac启停数据库
澈2072 天前
【无标题】QT入门第十二天：数据库编程（下）模型视图与数据展示 | 零基础学QT
数据库·qt·oracle
在路上~~~~2 天前
EBS 启用系统标准的debug
运维·oracle
落叶-IT3 天前
Java异常处理深度实战教程：彻底掌握异常传播机制，规避线上隐性故障
java·数据库·oracle
热门推荐
01GitHub 镜像站点022026 年 AI 编程工具终极横评：Cursor vs Claude Code vs Copilot vs Windsurf03AI科技热点日报 | 2026年07月01日04【AI】2026 年具身智能模型和世界模型总结052026 AI 编程工具终极实战指南：Cursor vs Claude Code vs Copilot，开发者该怎么选？062026 年 AI 大模型 & AI 编程工具实战全总结072026 AI 编程工具选型横评：Cursor / Claude Code / Trae / Copilot 到底选谁（建议收藏·避坑版）082026 国产 AI 大模型横评：DeepSeek、通义千问、Kimi、文心一言、星火、豆包谁更能打？092026年6月AI大模型全景报告：GPT-5.6、Claude Opus 4.8、Gemini 3.5，中美AI三足鼎立谁主沉浮？10Agnes AI 免费 API 接入指南：文本、生图、生视频，一套接口全免费