如何消除子类型依赖？

Question

在下面的示例中，我为每个 pls_integer 子类型编写了一个 to_str() 函数和一个 set() 过程。除了类型之外，功能和过程几乎相同。

如何消除为新子类型编写另一个 to_str() 和 set() 的需要，而不放弃子类型提供的约束？

回到 varchar2 like

procedure set(list in varchar2, prefix in varchar2)

然后调用它

set(to_str(list), 'foos:')

听起来不是一个好主意，我仍然需要为每个子类型提供 to_str() 。

我对各种不同的建议持开放态度，因为我是 Oracle 新手，新的 Oracle 功能几乎每天都让我感到惊讶。

我运行的是 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

Answer 1

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;

Answer 2

这可能无法回答您的问题，但为什么不将数据放入常规表中，然后使用 wm_concat 聚合函数将它们连接起来，如您所示？

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，因此无需重载函数。此外，还有其他方法可以使用；解析函数法看起来不错，但是我没有11g来测试！

（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目标；特别是多态性。但是，这超出了我的范围......所以也许其他人可以帮腔。）

Answer 3

下面的答案实际上是关于你如何在 postgresql （和 plpgsql）中做到这一点，我也不知道 oracle 子类型，但我认为它们足够相似，至少它会引导你找到你的答案。

create function add (anynonarray,anynonarray) returning anynonarray
as 'begin return $1 + $2; end';

我知道我搞砸了语法，但无论如何它应该显示我想用它显示的内容。

我们的想法是，它将替代“anynonarray”或任何替代方案 与调用参数的类型。一个限制是上例中的所有“anynonarray”都属于同一类型。

该文档将此称为多态性。