将多列中的逗号分隔值转换为行

Question

我正在尝试从表中创建一个视图，该视图会将多列中的逗号分隔值转换为 Oracle 中的行。

我已经成功地为一栏做到了这一点。但我无法为两三列做到这一点。

我使用以下脚本成功运行了一列。

Create VIEW MULTITESET AS 
SELECT rownum AS ID1,Tagging.COMMENTS,Tagging.category,Tagging.STATUS,
        trim(regexp_substr(Tagging.OBJ_ID, '[^,]+', 1, lines.column_value)) OBJ_ID
    FROM Tagging,
      TABLE (CAST (MULTISET
      (SELECT LEVEL FROM dual
              CONNECT BY instr(Tagging.OBJ_ID, ',', 1, LEVEL - 1) > 0
      ) AS sys.odciNumberList ) ) lines
    ORDER BY id, lines.column_value

现在我需要对名为 OBJname 的第二列执行相同的操作，就像对列 OBJ_ID 一样。

所以我尝试了一些愚蠢的方法，例如以下不起作用。

Create VIEW MULTITESET AS 
SELECT rownum AS ID1,Tagging.COMMENTS,Tagging.category,Tagging.STATUS,
        trim(regexp_substr(Tagging.OBJ_ID, '[^,]+', 1, lines.column_value)) OBJ_ID
    FROM Tagging,
      TABLE (CAST (MULTISET
      (SELECT LEVEL FROM dual
              CONNECT BY instr(Tagging.OBJ_ID, ',', 1, LEVEL - 1) > 0
      ) AS sys.odciNumberList ) ) lines
    ORDER BY id, lines.column_value ,
     trim(regexp_substr(Tagging.OBJname , '[^,]+', 1, lines.column_value)) OBJname 
    FROM Tagging,
      TABLE (CAST (MULTISET
      (SELECT LEVEL FROM dual
              CONNECT BY instr(Tagging.OBJname , ',', 1, LEVEL - 1) > 0
      ) AS sys.odciNumberList ) ) lines
    ORDER BY id, lines.column_value

简而言之，我希望将下图中的 table-A 转换为 table-B。我该怎么做？

我的实际视图查询是：

SELECT rownum AS TRACKID2, LEAPFROG_TAGGING.ID, LEAPFROG_TAGGING.CREATED_DATE,
  LEAPFROG_TAGGING.CREATED_BY, LEAPFROG_TAGGING.COMMENTS,
  leapfrog_tagging.category, LEAPFROG_TAGGING.STATUS, LEAPFROG_TAGGING.OBJ_NAME,
  trim(regexp_substr(LEAPFROG_TAGGING.OBJ_ID, '[^,]+', 1, lines.column_value)) OBJ_ID
FROM LEAPFROG_TAGGING,
TABLE (CAST (MULTISET (
    SELECT LEVEL FROM dual
    CONNECT BY instr(LEAPFROG_TAGGING.OBJ_ID, ',', 1, LEVEL - 1) > 0
  ) AS sys.odciNumberList ) ) lines
ORDER BY id, lines.column_value;

Answer 1

假设逗号分隔值的数量始终在两列中匹配（因此每个 ID 恰好有一个名称），并且您使用的是 11gR2，您可以使用recursive subquery factoring：

create view multiteset as
with t (obj_id, obj_name, status, other_columns, rn, orig_obj_id, orig_obj_name) as (
  select regexp_substr(obj_id, '[^,]+', 1, 1),
    regexp_substr(obj_name, '[^,]+', 1, 1),
    status, other_columns, 1, obj_id, obj_name
  from tagging
  union all
  select regexp_substr(orig_obj_id, '[^,]+', 1, rn + 1),
    regexp_substr(orig_obj_name, '[^,]+', 1, rn + 1),
    status, other_columns, rn + 1, orig_obj_id, orig_obj_name
  from t
  where rn < regexp_count(orig_obj_id, '[^,]+')
)
select obj_id, obj_name, status, other_columns
from t;

select * from multiteset;

OBJ_ID                       OBJ_NAME STATUS OTHER_
---------------------------- -------- ------ ------
1                            a        open   value1
1                            a        closed value2
4                            hj       na     value3
2                            s        open   value1
2                            s        closed value2
5                            hj       na     value3
3                            d        open   value1
6                            hj       na     value3
4                            f        open   value1
7                            hj       na     value3

10 rows selected. 

您的问题中的列名有点不一致，因此您必须根据真实姓名进行调整。

锚成员从每个列表中获取第一个元素，并保留原始列表的副本，以及查看我们在哪个元素上的计数器（从 1 开始）。

然后递归成员在每个列表中查找下一个 (rn +) 元素，除了计数器之外，其余的值保持不变。

然后，针对 CTE 的最终查询仅获取生成的值，忽略 CTE 正在跟踪的临时值（rn、orig_obj_id 等）

根据您在 cmets 中添加的新原始视图定义，您可能真的想要：

create view multiteset as
with t (id, created_date, created_by, comments, category, status,
  obj_name, obj_id, rn, orig_obj_name, orig_obj_id)
as (
  select id, created_date, created_by, comments, category, status,
    regexp_substr(obj_name, '[^,]+', 1, 1),
    regexp_substr(obj_id, '[^,]+', 1, 1),
    1, obj_name, obj_id
  from leapfrog_tagging
  union all
  select id, created_date, created_by, comments, category, status,
    regexp_substr(orig_obj_name, '[^,]+', 1, rn + 1),
    regexp_substr(orig_obj_id, '[^,]+', 1, rn + 1),
    rn + 1, orig_obj_name, orig_obj_id
  from t
  where rn < regexp_count(orig_obj_id, '[^,]+')
)
select id, created_date, created_by, comments, category, status,
  obj_name, obj_id
from t;

还要注意没有order by 子句；这在视图定义中没有真正意义，应该在您查询视图时应用：

select * from multiteset order by id, obj_id;

如果您想保留位置标记，以便视图结果也可以按此排序，则在最终选择列表中包含 rn（或更有意义的名称！），例如：

...
select id, rn as trackid2, created_date, created_by, comments, category, status,
  obj_name, obj_id
from t;

这会让你：

select * from multiteset order by id, trackid2;