在下面的示例中,我为每个to_str()
子类型编写了一个set()
函数和一个pls_integer
过程。除类型外,功能和程序几乎相同。
如何在不放弃子类型提供的约束的情况下,为新的子类型再次编写另一个to_str()
和set()
的需要?
回到varchar2
喜欢
procedure set(list in varchar2, prefix in varchar2)
然后将其称为
set(to_str(list), 'foos:')
听起来不太好,我仍然需要为每个子类型提供to_str()
。
我对所有不同的提案持开放态度,因为我是甲骨文新手,几乎每天都有新的Oracle功能让我感到惊讶。
我正在运行11.2.0.1.0。
create table so1table (
id number,
data varchar(20)
);
create or replace package so1 as
subtype foo_t is pls_integer range 0 .. 4 not null;
type foolist is table of foo_t;
procedure set(id_ in number, list in foolist default foolist(1));
subtype bar_t is pls_integer range 5 .. 10 not null;
type barlist is table of bar_t;
procedure set(id_ in number, list in barlist default barlist(5));
end;
/
show errors
create or replace package body so1 as
/* Do I have always to implement these very similar functions/procedures for
every single type ? */
function to_str(list in foolist) return varchar2 as
str varchar2(32767);
begin
for i in list.first .. list.last loop
str := str || ' ' || list(i);
end loop;
return str;
end;
function to_str(list in barlist) return varchar2 as
str varchar2(32767);
begin
for i in list.first .. list.last loop
str := str || ' ' || list(i);
end loop;
return str;
end;
procedure set(id_ in number, list in foolist default foolist(1)) as
values_ constant varchar2(32767) := 'foos:' || to_str(list);
begin
insert into so1table (id, data) values (id_, values_);
end;
procedure set(id_ in number, list in barlist default barlist(5)) as
values_ constant varchar2(32767) := 'bars:' || to_str(list);
begin
insert into so1table (id, data) values (id_, values_);
end;
end;
/
show errors
begin
so1.set(1, so1.foolist(0, 3));
so1.set(2, so1.barlist(5, 7, 10));
end;
/
SQLPLUS> select * from so1table;
ID DATA
---------- --------------------
1 foos: 0 3
2 bars: 5 7 10
答案 0 :(得分:3)
create table so1table (
id number,
data varchar(20)
);
create or replace type parent_type as object
(
v_number number,
--Prefix probably belongs with a list, not an individual value.
--For simplicity, I'm not adding another level to the objects.
v_prefix varchar2(10)
) not instantiable not final;
/
create or replace type parentlist as table of parent_type;
/
create or replace type foo_type under parent_type
(
constructor function foo_type(v_number number) return self as result
);
/
--The data must be stored as a NUMBER, since ADTs don't support
--PL/SQL specific data types. The type safety is enforced by the
--conversion in the constructor.
create or replace type body foo_type is
constructor function foo_type(v_number number) return self as result
as
subtype foo_subtype is pls_integer range 0 .. 4 not null;
new_number foo_subtype := v_number;
begin
self.v_number := new_number;
self.v_prefix := 'foos:';
return;
end;
end;
/
create or replace type foolist as table of foo_type;
/
create or replace type bar_type under parent_type
(
constructor function bar_type(v_number number) return self as result
);
/
create or replace type body bar_type is
constructor function bar_type(v_number number) return self as result
as
subtype bar_subtype is pls_integer range 5 .. 10 not null;
new_number bar_subtype := v_number;
begin
self.v_number := new_number;
self.v_prefix := 'bars:';
return;
end;
end;
/
create or replace type barlist as table of bar_type;
/
create or replace package so1 as
procedure set(id_ in number, list in parentlist);
end;
/
create or replace package body so1 as
function to_str(list in parentlist) return varchar2 as
v_value VARCHAR2(32767);
begin
for i in list.first .. list.last loop
if i = 1 then
v_value := list(i).v_prefix;
end if;
v_value := v_value || ' ' || list(i).v_number;
end loop;
return v_value;
end to_str;
procedure set(id_ in number, list in parentlist) as
values_ constant varchar2(32767) := to_str(list);
begin
insert into so1table (id, data) values (id_, values_);
end set;
end so1;
/
begin
--You probably don't want to mix foos and bars, but it is allowed.
so1.set(1, parentlist(foo_type(0), foo_type(3)));
so1.set(2, parentlist(bar_type(5), bar_type(7), bar_type(10)));
--These would generate "ORA-06502: PL/SQL: numeric or value error"
--so1.set(1, parentlist(foo_type(5)));
--so1.set(1, parentlist(bar_type(4)));
end;
/
select * from so1table;
答案 1 :(得分:1)
这可能无法解答您的问题,但为什么不将数据放在常规表中,然后使用wm_concat
聚合函数连接它们,如图所示?
即,
> select * from myTable;
ID Category Value
--- --------- ------
1 foo 0
2 foo 3
3 bar 5
4 bar 7
5 bar 10
> select Category||'s: '||replace(wm_concat(Value),',',' ') Data
from myTable
group by Category;
Data
-------------
bars: 5 7 10
foos: 0 3
wm_concat
与类型无关,因此您无需重载您的函数。此外,还有其他methods可以使用;分析函数方法看起来不错,但我没有11g用于测试!
(编辑否则,我认为您可以使用Oracle的对象模型实现您正在寻找的东西;特别是多态性。但是,这超出了我...所以也许别人可以插话。)
答案 2 :(得分:-1)
以下答案实际上是你如何在postgresql(和plpgsql)中进行的,我也不知道oracle子类型,但我认为它们足够相似,至少它会引导你到你的答案。
create function add (anynonarray,anynonarray) returning anynonarray
as 'begin return $1 + $2; end';
我知道我破坏了语法,但无论如何它应该显示我想用它显示的内容。
这个想法是它将“anynonarray”or any of the alternatives替换为调用参数的类型。一个限制是上面例子中的所有“anynonarray”都属于同一类型。
文档将此称为多态。