自参照表的SQL查询来行

Question

我有这样的Oracle 11g查询：

SELECT RPAD(' ', 2 * (T.ID_LEVEL - 1)) || T.IDE IDE, T.ID_LEVEL, T.CODE, T.FK_IDE 
FROM TEST_DYNAMIC T 
START WITH T.FK_IDE = 0 
CONNECT BY NOCYCLE PRIOR T.IDE = T.FK_IDE 
ORDER BY T.IDE,T.FK_IDE; 

返回此数据：

IDE |IDE_LEVEL |CODE |FK_IDE | 
-----|-----------|------|--------| 
1 |   1|A01 |  0| 
2 |   2|A01 |  1| 
    3 |   3|A01 |  2| 
    4|   4|A01 |  3| 
5 |   2|A02 |  1| 
6 |   2|A03 |  1| 
    7 |   3|A01 |  6| 
8 |   1|A02 |  0| 

正如你所看到的，是从自我refrencing表得到的数据，其中IDE_LEVEL列是包含此值（1，2，3，4）作为PK的主表的外键，而IDE列是来自TEST_DYNAMIC表的自动增量PK。

有一种方法以该结果转换为这一个？：

IDE |CODE_LEVEL1 |CODE_LEVEL2 |CODE_LEVEL3 |CODE_LEVEL4 | 
-----|-------------|-------------|-------------|-------------| 
1 |A01   |A01   |A01   |A01   | 
1 |A01   |A02   |NULL   |NULL   | 
1 |A01   |A03   |A01   |NULL   | 
8 |A02   |NULL   |NULL   |NULL   | 

在预期的结果如上所述，IDE列被示出对应于在FK_COLUMN三个ocurrences关键1三次，和第一个结果集中键8的一次（这一个没有子元素，因此必须显示在结果集中）。

任何帮助将不胜感激。

Answer 1

起初我没有得到这个问题。你不需要PIVOT，你需要在不同的列中显示每个分支的路径。无论如何，根据级别的数量，你可能需要一些动态的sql

with test_dynamic as (
select 1 ide, 1 id_level, 'A01' code, 0 fk_ide from dual 
union all select 2, 2, 'A01', 1 from dual 
union all select 3, 3, 'A01', 2 from dual 
union all select 4, 4, 'A01', 3 from dual 
union all select 5, 2, 'A02', 1 from dual 
union all select 6, 2, 'A03', 1 from dual 
union all select 7, 3, 'A01', 6 from dual 
union all select 8, 1, 'A02', 0 from dual 
), 
tree as (
select code, level lvl, sys_connect_by_path(code, '/') path, connect_by_isleaf lf, connect_by_root(ide) root_ide 
from test_dynamic t 
start with t.fk_ide = 0 
connect by nocycle prior t.ide = t.fk_ide 
) 
select root_ide, path, 
     regexp_substr(path, '[^/]+', 1, 1), 
     regexp_substr(path, '[^/]+', 1, 2), 
     regexp_substr(path, '[^/]+', 1, 3), 
     regexp_substr(path, '[^/]+', 1, 4) 
from tree where lf = 1; 

Answer 2

如果级别的数量是动态的，那么无论如何您都需要使用动态SQL来创建具有动态列的查询。

至少有2种方法如何创建这样一个查询：

1）甲骨文与PIVOT（因为11克我想）

with test_dynamic as (
select 1 ide, 1 id_level, 'A01' code, 0 fk_ide from dual 
union all select 2, 2, 'A01', 1 from dual 
union all select 3, 3, 'A01', 2 from dual 
union all select 4, 4, 'A01', 3 from dual 
union all select 5, 2, 'A02', 1 from dual 
union all select 6, 2, 'A03', 1 from dual 
union all select 7, 3, 'A01', 6 from dual 
union all select 8, 1, 'A02', 0 from dual 
), 
tree as (
select level lvl, t.ide, t.code 
from test_dynamic t 
start with t.fk_ide = 0 
connect by nocycle prior t.ide = t.fk_ide 
), 
pivot_data as (
select * from tree pivot xml (max(code) lvl_code for lvl in (select distinct lvl from tree)) 
) 
select ide, 
     extractvalue(lvl_xml,'/PivotSet/item[1]/column[2]') code_level1, 
     extractvalue(lvl_xml,'/PivotSet/item[2]/column[2]') code_level2, 
     extractvalue(lvl_xml,'/PivotSet/item[3]/column[2]') code_level3, 
     extractvalue(lvl_xml,'/PivotSet/item[4]/column[2]') code_level4 
from pivot_data t order by 1; 

（人们说，extractValue一起有可能成为过时的和推荐使用XMLTABLE，但我从来没有与这个工作）

2）的Oracle DECODE

with test_dynamic as (
select 1 ide, 1 id_level, 'A01' code, 0 fk_ide from dual 
union all select 2, 2, 'A01', 1 from dual 
union all select 3, 3, 'A01', 2 from dual 
union all select 4, 4, 'A01', 3 from dual 
union all select 5, 2, 'A02', 1 from dual 
union all select 6, 2, 'A03', 1 from dual 
union all select 7, 3, 'A01', 6 from dual 
union all select 8, 1, 'A02', 0 from dual 
), 
tree as (
select level lvl, t.ide, t.code 
from test_dynamic t 
start with t.fk_ide = 0 
connect by nocycle prior t.ide = t.fk_ide 
) 
select ide, 
     max(decode(lvl, 1, code, null)) code_level1, 
     max(decode(lvl, 2, code, null)) code_level2, 
     max(decode(lvl, 3, code, null)) code_level3, 
     max(decode(lvl, 4, code, null)) code_level4 
from tree group by ide order by 1;