Oracle的PL/SQL提供了Pipelined Table Functions特性用於把多行資料返回到呼叫者，可以有效的提升效能。
在PostgreSQL中，可以透過在函式中利用SETOF或者RETURN NEXT來實現。

**Oracle**
建立資料表，插入資料
```
TEST-orcl@DESKTOP-V430TU3>drop table t_piperow;
drop table t_piperow
           *
ERROR at line 1:
ORA-00942: table or view does not exist

TEST-orcl@DESKTOP-V430TU3>create table t_piperow(id int,c1 timestamp ,c2 varchar2(20),c3 number);

Table created.

TEST-orcl@DESKTOP-V430TU3>
TEST-orcl@DESKTOP-V430TU3>insert into t_piperow(id,c1,c2,c3)
  2    select rownum,sysdate,'test'||rownum,123455.55 from dba_objects where rownum <= 500;

500 rows created.

TEST-orcl@DESKTOP-V430TU3>commit;

Commit complete.

TEST-orcl@DESKTOP-V430TU3>
```
建立Type
```
TEST-orcl@DESKTOP-V430TU3>CREATE OR REPLACE TYPE rec_t_piperow AS OBJECT(id int,c1 timestamp ,c2 varchar2(20),c3 number);
  2  /

Type created.

TEST-orcl@DESKTOP-V430TU3>CREATE OR REPLACE TYPE type_t_piperow AS TABLE OF rec_t_piperow;
  2  /

Type created.

```
函式實現
```
TEST-orcl@DESKTOP-V430TU3>CREATE OR REPLACE FUNCTION func_piperow_demo1 RETURN type_t_piperow PIPELINED IS
  2  BEGIN
  3    FOR rec in (select * from t_piperow) LOOP
  4      PIPE ROW (rec_t_piperow(rec.id,rec.c1,rec.c2,rec.c3));
  5    END LOOP;
  6    RETURN;
  7  END;
  8  /

Function created.

```
查詢資料
```
TEST-orcl@DESKTOP-V430TU3>column c3 format 9999999999999.9999
TEST-orcl@DESKTOP-V430TU3>select * from table(func_piperow_demo1()) where rownum < 5;

ID C1                   C2                                    C3
---------- -------------------- -------------------- -------------------
         1 31-OCT-19 10.50.38.0 test1                        123455.5500
           00000 AM

2 31-OCT-19 10.50.38.0 test2                        123455.5500
           00000 AM

3 31-OCT-19 10.50.38.0 test3                        123455.5500
           00000 AM

4 31-OCT-19 10.50.38.0 test4                        123455.5500
           00000 AM
```

**PostgreSQL**
下面來看看PG的實現，建立表，插入資料
```
[local]:5432 pg12@testdb=# drop table if exists t_piperow;
DROP TABLE
Time: 5.255 ms
[local]:5432 pg12@testdb=# create table t_piperow(id int,c1 timestamp ,c2 varchar(20),c3 float);
CREATE TABLE
Time: 4.711 ms
[local]:5432 pg12@testdb=# 
[local]:5432 pg12@testdb=# insert into t_piperow(id,c1,c2,c3) 
pg12@testdb-#   select x,now(),'test'||x,123455.55 from generate_series(1,500) x;
INSERT 0 500
Time: 10.183 ms
[local]:5432 pg12@testdb=# 
[local]:5432 pg12@testdb=#

```
函式實現
第一種方式，使用SQL：
```
[local]:5432 pg12@testdb=# CREATE OR REPLACE FUNCTION func_piperow_demo1() RETURNS SETOF PUBLIC.t_piperow
pg12@testdb-# AS
pg12@testdb-# $$
pg12@testdb$#   SELECT * FROM t_piperow;
pg12@testdb$# $$
pg12@testdb-# LANGUAGE SQL;
CREATE FUNCTION
Time: 1.341 ms
[local]:5432 pg12@testdb=#

[local]:5432 pg12@testdb=# select func_piperow_demo1() limit 5;
                func_piperow_demo1                
--------------------------------------------------
 (1,"2019-10-31 11:09:27.222996",test1,123455.55)
 (2,"2019-10-31 11:09:27.222996",test2,123455.55)
 (3,"2019-10-31 11:09:27.222996",test3,123455.55)
 (4,"2019-10-31 11:09:27.222996",test4,123455.55)
 (5,"2019-10-31 11:09:27.222996",test5,123455.55)
(5 rows)

Time: 1.038 ms
[local]:5432 pg12@testdb=#

```
第二種方式，使用PL/pgSQL，RETURN QUERY
```
[local]:5432 pg12@testdb=# CREATE OR REPLACE FUNCTION func_piperow_demo2() RETURNS SETOF PUBLIC.t_piperow
pg12@testdb-# AS
pg12@testdb-# $$
pg12@testdb$# BEGIN
pg12@testdb$#   RETURN QUERY SELECT * FROM t_piperow;
pg12@testdb$# END;
pg12@testdb$# $$
pg12@testdb-# LANGUAGE PLPGSQL;
CREATE FUNCTION
Time: 3.850 ms
[local]:5432 pg12@testdb=# select func_piperow_demo2() limit 5;
                func_piperow_demo2                
--------------------------------------------------
 (1,"2019-10-31 11:09:27.222996",test1,123455.55)
 (2,"2019-10-31 11:09:27.222996",test2,123455.55)
 (3,"2019-10-31 11:09:27.222996",test3,123455.55)
 (4,"2019-10-31 11:09:27.222996",test4,123455.55)
 (5,"2019-10-31 11:09:27.222996",test5,123455.55)
(5 rows)

Time: 5.645 ms
[local]:5432 pg12@testdb=#

```
第三種方式，使用PL/pgSQL，RETURN NEXT
```
[local]:5432 pg12@testdb=# select func_piperow_demo3() limit 5;
                func_piperow_demo3                
--------------------------------------------------
 (1,"2019-10-31 11:09:27.222996",test1,123455.55)
 (2,"2019-10-31 11:09:27.222996",test2,123455.55)
 (3,"2019-10-31 11:09:27.222996",test3,123455.55)
 (4,"2019-10-31 11:09:27.222996",test4,123455.55)
 (5,"2019-10-31 11:09:27.222996",test5,123455.55)
(5 rows)

Time: 5.588 ms
[local]:5432 pg12@testdb=#

```

**參考資料**
[PostgreSQL Oracle 相容性 之 TABLE、PIPELINED函式](https://yq.aliyun.com/articles/228257)
[Pipelined in Oracle as well in PostgreSQL](http://manojadinesh.blogspot.com/2011/11/pipelined-in-oracle-as-well-in.html)

Oracle vs PostgreSQL Develop（19） - PIPE ROW

相關文章