在SQL GROUP BY中聚合集合中的不同值

I have a table with a collection type in it. I want to SELECT from the table and GROUP BY some column. I want the results for each group to include a collection that contains the distinct union of all the individual collections in the group.

我有一个包含集合类型的表。我想从表和GROUP BY中选择一些列。我希望每个组的结果都包含一个集合，该集合包含组中所有单个集合的不同并集。

For example,

例如，

CREATE OR REPLACE TYPE my_varchar2_list AS TABLE OF VARCHAR2 (80);

WITH test_data (id, a_list) AS
       (SELECT 1,
               NEW my_varchar2_list ('A', 'B', 'C')
        FROM   DUAL
        UNION ALL
        SELECT 1,
               NEW my_varchar2_list ('C', 'D', 'E')
        FROM   DUAL)
SELECT id,
       ... some magic syntax here...
FROM   test_data
GROUP BY id

Desired results:

1,  ('A','B','C','D','E')

What I am looking for in an answer

I am only looking for expressions that I can plug into the "magic syntax goes here" placeholder above. I know I can accomplish the aggregate by joining TABLE(a_list) to the main table or otherwise restructuring my query (or, of course, using PL/SQL). Please don't bother to provide answers along those lines: they don't answer my question.

我只是寻找表达式，我可以插入上面的“magic syntax goes here”占位符。我知道我可以通过将TABLE（a_list）连接到主表或以其他方式重构我的查询（或者当然，使用PL / SQL）来完成聚合。请不要在这些方面提供答案：他们不回答我的问题。

1 个解决方案

#1

Adapted from my answer to another question.

改编自我对另一个问题的回答。

Oracle Setup:

Oracle安装程序：

CREATE OR REPLACE TYPE VARCHAR2s_Table IS TABLE OF VARCHAR2(100);
/

Create a user-defined aggregation type:

创建用户定义的聚合类型：

CREATE OR REPLACE TYPE Varchar2sTableUnion AS OBJECT(
  list VARCHAR2s_Table,

  STATIC FUNCTION ODCIAggregateInitialize(
    ctx         IN OUT Varchar2sTableUnion
  ) RETURN NUMBER,

  MEMBER FUNCTION ODCIAggregateIterate(
    self        IN OUT Varchar2sTableUnion,
    value       IN     VARCHAR2s_Table
  ) RETURN NUMBER,

  MEMBER FUNCTION ODCIAggregateTerminate(
    self        IN OUT Varchar2sTableUnion,
    returnValue    OUT VARCHAR2s_Table,
    flags       IN     NUMBER
  ) RETURN NUMBER,

  MEMBER FUNCTION ODCIAggregateMerge(
    self        IN OUT Varchar2sTableUnion,
    ctx         IN OUT Varchar2sTableUnion
  ) RETURN NUMBER
);
/

CREATE OR REPLACE TYPE BODY Varchar2sTableUnion
IS
  STATIC FUNCTION ODCIAggregateInitialize(
    ctx         IN OUT Varchar2sTableUnion
  ) RETURN NUMBER
  IS
  BEGIN
    ctx := Varchar2sTableUnion( NULL );
    RETURN ODCIConst.SUCCESS;
  END;

  MEMBER FUNCTION ODCIAggregateIterate(
    self        IN OUT Varchar2sTableUnion,
    value       IN     VARCHAR2s_Table
  ) RETURN NUMBER
  IS
  BEGIN
    IF value IS NULL THEN
      NULL;
    ELSIF self.list IS NULL THEN
      self.list := value;
    ELSE
      self.list := self.list MULTISET UNION DISTINCT value;
    END IF;
    RETURN ODCIConst.SUCCESS;
  END;

  MEMBER FUNCTION ODCIAggregateTerminate(
    self        IN OUT Varchar2sTableUnion,
    returnValue    OUT VARCHAR2s_Table,
    flags       IN     NUMBER
  ) RETURN NUMBER
  IS
  BEGIN
    returnValue := self.list;
    RETURN ODCIConst.SUCCESS;
  END;

  MEMBER FUNCTION ODCIAggregateMerge(
    self        IN OUT Varchar2sTableUnion,
    ctx         IN OUT Varchar2sTableUnion
  ) RETURN NUMBER
  IS
  BEGIN
    IF self.list IS NULL THEN
      self.list := ctx.list;
    ELSIF ctx.list IS NULL THEN
      NULL;
    ELSE
      self.list := self.list MULTISET UNION DISTINCT ctx.list;
    END IF;
    RETURN ODCIConst.SUCCESS;
  END;
END;
/

Create a user-defined aggregation function:

创建用户定义的聚合函数：

CREATE FUNCTION MULTISET_UNION( list VARCHAR2s_Table )
RETURN VARCHAR2s_Table
PARALLEL_ENABLE AGGREGATE USING Varchar2sTableUnion;
/

Query:

查询：

Then you can just use it to perform the aggregation in your query:

然后，您可以使用它在查询中执行聚合：

WITH test_data (id, a_list) AS
       (SELECT 1,
               varchar2s_table ('A', 'B', 'C')
        FROM   DUAL
        UNION ALL
        SELECT 1,
               varchar2s_table ('C', 'D', 'E')
        FROM   DUAL)
SELECT id,
       MULTISET_UNION( a_list )
FROM   test_data
GROUP BY id

Output:

输出：

ID MULTISET_UNION(A_LIST)
-- -------------------------------------------
 1 SCHEMA.VARCHAR2S_TABLE('A','B','C','D','E')

#1