Oracle SQL 中多集映射的意外结果答案

【问题标题】：Unexpected result of multiset mapping in Oracle SQLOracle SQL 中多集映射的意外结果
【发布时间】：2013-06-11 09:17:13
【问题描述】：

请帮助我确认下面解释的行为是错误，或者清楚地解释为什么它是正确的。我很可能误解了一些概念，但现在对我来说它看起来像一个错误。

下面的所有示例都尽可能简化以展示问题的核心。实际情况非常复杂，因此只能接受与查询构造原理相关的一般性答案和变通方法。
欢迎您在 cmets 中提出澄清问题，我会尽力回答。
感谢您的关注。 :)

问题

为什么在上一个示例（示例 5）(select count(1) ... 子查询中的 collection 实例从第一行映射到表的所有行，而预期结果是将每个 collection 实例映射到它自己的行？
同时collections中使用的cardinality(...)表达式选择正确。
如果以这种方式构造用于查询的from 或where 部分的集合，则存在相同的情况（示例中未涵盖）。

测试架构设置

(SQLFiddle)

create or replace type TabType0 as table of varchar2(100)
/

create table Table0( tab_str_field varchar2(100), tab_field TabType0)
nested table tab_field store as tab_field_table
/

insert into table0 (tab_str_field, tab_field) values (
   'A',
   cast(multiset(
     select 'A' from dual union all
     select 'B' from dual union all
     select 'C' from dual 
   ) as TabType0) 
)
/

insert into table0 (tab_str_field, tab_field) values (
   'B',
   cast(multiset(
     select 'B' from dual union all
     select 'C' from dual 
   ) as TabType0) 
)
/

insert into table0 (tab_str_field, tab_field) values (
   'C',
   cast(multiset(
     select 'A' from dual union all
     select 'B' from dual union all
     select 'C' from dual union all
     select 'D' from dual 
   ) as TabType0) 
)
/

insert into table0 (tab_str_field, tab_field) values (
   'D',
   cast(multiset(
     select 'A' from dual 
   ) as TabType0) 
)
/
select 'Initial table data' caption from dual
/
select * from table0
/

表格数据：

| TAB_STR_FIELD | TAB_FIELD |
-----------------------------
|             A |     A,B,C |
|             B |       B,C |
|             C |   A,B,C,D |
|             D |         A |

示例

示例 1 (SQLFiddle) - 使用嵌套表字段 - OK

select 'Work with nested table - OK' caption from dual
/
select 
  tab_field                               tab_field,

  -- cardinality
  cardinality(tab_field)                  tab_cardinality,

  -- select from table field of current row
  (select count(1) from table(tab_field)) tab_count,

  -- select from field of current row while joining 
  -- with another field of same row
  ( select column_value from table(tab_field) 
    where column_value = tab_str_field
  )                                       same_value
from table0
/

结果：

| TAB_FIELD | TAB_CARDINALITY | TAB_COUNT | SAME_VALUE |
--------------------------------------------------------
|     A,B,C |               3 |         3 |          A |
|       B,C |               2 |         2 |          B |
|   A,B,C,D |               4 |         4 |          C |
|         A |               1 |         1 |     (null) |

示例 2 (SQLFiddle) - 单独使用构造的源数据 - OK

select 'Work with constructed source data alone - OK' caption from dual
/
with table_data_from_set as (
  select
    'A' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'B' tab_str_field,
    cast(multiset(
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'C' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual union all
      select 'D' from dual
    ) as TabType0) tab_field
  from dual union all
  select
    'D' tab_str_field,
    cast(multiset(
      select 'A' from dual
    ) as TabType0) tab_field
  from dual
)
select
  tab_field                                tab_field,

  -- cardinality
  cardinality(tab_field)                   tab_cardinality,

  -- select from table field of current row
  (select count(1) from table(tab_field))  tab_count,

  -- select from field of current row while joining
  -- with another field of same row
  ( select column_value from table(tab_field)
    where column_value = tab_str_field
  )                                        same_value
from table_data_from_set
/

结果：

| TAB_FIELD | TAB_CARDINALITY | TAB_COUNT | SAME_VALUE |
--------------------------------------------------------
|     A,B,C |               3 |         3 |          A |
|       B,C |               2 |         2 |          B |
|   A,B,C,D |               4 |         4 |          C |
|         A |               1 |         1 |     (null) |

示例 3 (SQLFiddle) - 使用在 WITH 中构造的多重集连接表 - OK

select 'Join table with multisets constructed in WITH - OK' caption from dual
/
with table_data_from_set as (
  select
    'A' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'B' tab_str_field,
    cast(multiset(
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'C' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual union all
      select 'D' from dual
    ) as TabType0) tab_field
  from dual union all
  select
    'D' tab_str_field,
    cast(multiset(
      select 'A' from dual
    ) as TabType0) tab_field
  from dual
)
select
  table0.tab_field                                            table0_tab_field,
  table_data_from_set.tab_field                               set_tab_field,

  -- cardinality
  cardinality(table0.tab_field) table0_tab_cardinality,
  cardinality(table_data_from_set.tab_field)                  set_tab_cardinality,

  -- select from table field of current row
  (select count(1) from table(table_data_from_set.tab_field)) set_tab_count,

  -- select from field of current row while joining
  -- with another field of same row
  ( select column_value from table(table_data_from_set.tab_field)
    where column_value = table0.tab_str_field
  )                                                           same_value
from 
  table0, 
  table_data_from_set 
where 
  table_data_from_set.tab_str_field = table0.tab_str_field
/

结果：

| TABLE0_TAB_FIELD | SET_TAB_FIELD | TABLE0_TAB_CARDINALITY | SET_TAB_CARDINALITY | SET_TAB_COUNT | SAME_VALUE |
----------------------------------------------------------------------------------------------------------------
|            A,B,C |         A,B,C |                      3 |                   3 |             3 |          A |
|              B,C |           B,C |                      2 |                   2 |             2 |          B |
|          A,B,C,D |       A,B,C,D |                      4 |                   4 |             4 |          C |
|                A |             A |                      1 |                   1 |             1 |     (null) |

示例 4 (SQLFiddle) - 使用 WITH + 子查询构造的多集连接表 - OK

select 'Join table with multisets constructed in WITH and subquery - OK' caption from dual
/
with table_data_from_set as (
  select
    'A' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'B' tab_str_field,
    cast(multiset(
      select 'B' from dual union all
      select 'C' from dual
    ) as TabType0)  tab_field
  from dual union all
  select
    'C' tab_str_field,
    cast(multiset(
      select 'A' from dual union all
      select 'B' from dual union all
      select 'C' from dual union all
      select 'D' from dual
    ) as TabType0) tab_field
  from dual union all
  select
    'D' tab_str_field,
    cast(multiset(
      select 'A' from dual
    ) as TabType0) tab_field
  from dual
)
select
  table0_tab_field                            table0_tab_field,
  set_tab_field                               set_tab_field,

  -- cardinality
  cardinality(table0_tab_field)               table0_tab_cardinality,
  cardinality(set_tab_field)                  set_tab_cardinality,

  -- select from table field of current row
  (select count(1) from table(set_tab_field)) set_tab_count,

  -- select from field of current row while joining
  -- with another field of same row
  ( select column_value from table(set_tab_field)
    where column_value = table0_tab_str_field
  )                                           same_value
from (
  select 
    table0.tab_str_field              table0_tab_str_field,
    table0.tab_field                  table0_tab_field,
    table_data_from_set.tab_str_field set_tab_str_field,
    table_data_from_set.tab_field     set_tab_field
  from 
    table0, 
    table_data_from_set 
  where 
    table_data_from_set.tab_str_field = table0.tab_str_field
)
/

结果：

| TABLE0_TAB_FIELD | SET_TAB_FIELD | TABLE0_TAB_CARDINALITY | SET_TAB_CARDINALITY | SET_TAB_COUNT | SAME_VALUE |
----------------------------------------------------------------------------------------------------------------
|            A,B,C |         A,B,C |                      3 |                   3 |             3 |          A |
|              B,C |           B,C |                      2 |                   2 |             2 |          B |
|          A,B,C,D |       A,B,C,D |                      4 |                   4 |             4 |          C |
|                A |             A |                      1 |                   1 |             1 |     (null) |

示例 5 (SQLFiddle) - 使用动态构建的多集连接表 - FAILED

select 'Join table with multisets constructed on the fly - FAIL (set_tab_count wrong)' caption from dual
/
with string_set as (
  select 'A' str_field from dual union all
  select 'B' str_field from dual union all
  select 'C' str_field from dual union all
  select 'D' str_field from dual union all
  select 'E' str_field from dual 
)
select
  table0_tab_field                            table0_tab_field,
  set_tab_field                               set_tab_field,

  -- cardinality
  cardinality(table0_tab_field)               table0_tab_cardinality,
  cardinality(set_tab_field)                  set_tab_cardinality,

  -- select from table field of current row
  (select count(1) from table(set_tab_field)) set_tab_count,

  -- select from field of current row while joining
  -- with another field of same row
  ( select column_value from table(set_tab_field)
    where column_value = table0_tab_str_field
  )                                            same_value
from (
  select 
    table0.tab_str_field     table0_tab_str_field,
    table0.tab_field         table0_tab_field,
    ( 
      cast(multiset(
        
        select 
          string_set.str_field 
        from 
          string_set, 
          table(table0.tab_field) tab_table
        where 
          string_set.str_field = tab_table.column_value
        
      ) as TabType0)
    )                        set_tab_field
  from 
    table0 
)  
/

结果（set_tab_count 列中的所有值都相同 - 错误！）：

| TABLE0_TAB_FIELD | SET_TAB_FIELD | TABLE0_TAB_CARDINALITY | SET_TAB_CARDINALITY | SET_TAB_COUNT | SAME_VALUE |
----------------------------------------------------------------------------------------------------------------
|            A,B,C |         A,B,C |                      3 |                   3 |             3 |          A |
|              B,C |           B,C |                      2 |                   2 |             3 |          B |
|          A,B,C,D |       A,B,C,D |                      4 |                   4 |             3 |          C |
|                A |             A |                      1 |                   1 |             3 |     (null) |

Oracle 版本信息

实例 1

BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production
PL/SQL Release 11.2.0.3.0 - Production
CORE    11.2.0.3.0  Production
TNS for IBM/AIX RISC System/6000: Version 11.2.0.3.0 - Production
NLSRTL Version 11.2.0.3.0 - Production

实例 2

BANNER 
--------------------------------------------------------------------------------
Oracle Database 11g Express Edition Release 11.2.0.2.0 - Production 
PL/SQL Release 11.2.0.2.0 - Production 
CORE    11.2.0.2.0  Production 
TNS for 32-bit Windows: Version 11.2.0.2.0 - Production 
NLSRTL Version 11.2.0.2.0 - Production

SQLFiddle 将所有查询放在一起。

【问题讨论】：

这太令人困惑了。你没有很好地阐明你想要做什么，预期的结果是什么，以及到底出了什么问题。我所知道的是，您在上次测试中对set_tab_count 的结果不是您所期望的。我不知道这个值应该代表什么，而且我没有通读所有这些代码。我很欣赏这可能是您可以简化可重现的测试用例的最简单方法，但如果是这种情况，您需要对您的代码进行一些解释，因为我们不会对所有内容进行挑剔。
@DavidMarx 你需要什么样的解释才能理解以上所有内容？前 4 个示例演示了在连接中使用集合的不同方法的相同结果。但是第五个结果打破了前 4 个例子中可以看到的所有逻辑。我只是想了解示例 5 中的逻辑或得到合理的解释为什么它是一个错误。
在 OTN 论坛 (forums.oracle.com/thread/2550938) 上回答我的问题时建议了另外 2 个解决方法，但比@jonearles 提供的更具体

标签： sql oracle collections oracle11g multiset

【解决方案1】：

这是一个错误。在最后一个示例中向第二个内联视图添加/*+ NO_MERGE */ 提示将生成预期结果。有关示例，请参阅this SQL Fiddle。无论查询如何，该提示都不应更改结果。您还可以进行其他一些看似无关的更改来生成正确的结果，例如删除一些列，或在中间添加一个未使用的 ROWNUM。

Oracle 正在重写您的查询以优化它，但是做错了。您可能可以通过跟踪查询获得更多信息，但我怀疑您能否真正解决问题。暂时解决它并向 Oracle 提交服务请求，以便他们可以创建错误并最终修复它。

【讨论】：

干得好！非常感谢您发现解决方法！我在更符合我需求的示例上尝试了两种解决方法（例如：sqlfiddle.com/#!4/12159/11），发现两者都有效。