Oracle：自连接 XMLTable 输出

Question

我正在尝试将某些 XML 解析为 WITH 子句中的 table，然后在主查询中，将 table 与其自身连接。但是，即使是最小的测试用例，我也没有得到任何记录。

这是示例 xml:

<xml> 
  <entry timestamp="20170330100429" effective="20170329"> 
    <field name="Name"> 
      <ov> <![CDATA[Fran]]> </ov> 
      <nv> <![CDATA[Frank]]> </nv> 
    </field> 
    <field name="Zip"> 
      <ov> <![CDATA[13583]]> </ov> 
      <nv> <![CDATA[13853]]> </nv> 
    </field> 
  </entry> 
  <entry timestamp="20170401094783" effective="20170331"> 
    <field name="MI"> 
      <ov> <![CDATA[J]]> </ov> 
      <nv> <![CDATA[A]]> </nv> 
    </field> 
    <field name="Suffix"> 
      <ov> <![CDATA[Jr]]> </ov> 
      <nv> <![CDATA[III]]> </nv> 
    </field> 
  </entry> 
</xml>

这里是精简的 SQL，将 XML 扁平化为 WITH 子句中的 table：

with myxml as
(
  select 1 xml_id,
         '<xml> <entry timestamp="20170330100429" effective="20170329"> <field name="Name"> <ov><![CDATA[Fran]]></ov> <nv><![CDATA[Frank]]></nv> </field> <field name="Zip"> <ov><![CDATA[13583]]></ov> <nv><![CDATA[13853]]></nv> </field> </entry> <entry timestamp="20170401094783" effective="20170331"> <field name="MI"> <ov><![CDATA[J]]></ov> <nv><![CDATA[A]]></nv> </field> <field name="Suffix"> <ov><![CDATA[Jr]]></ov> <nv><![CDATA[III]]></nv> </field> </entry> </xml>' x from dual
),
HRH as
(
SELECT au.xml_id,
       ts.tran_ts as chg_timestr,
       to_date(substr(ts.tran_ts, 0, 8), 'yyyymmdd') chg_date,
       cast(substr(ts.tran_ts, 9, 6) as int) chg_time,
       f.fieldname,
       f.ov old_value,
       f.nv new_value
  FROM myxml au,
     xmltable('/xml'
              PASSING XMLTYPE(au.x)
              COLUMNS entrynode XMLTYPE PATH 'entry'
     ) as x1,
     xmltable('/entry'
              PASSING x1.entrynode
              COLUMNS tran_ts   VARCHAR(16) PATH '@timestamp',
                      fieldnode XMLTYPE     PATH 'field'
     ) as ts,
     xmltable('/field'
              PASSING ts.fieldnode
              COLUMNS fieldname VARCHAR(250) PATH '@name',
                      nv VARCHAR(250) PATH 'nv',
                      ov VARCHAR(250) PATH 'ov'
     ) as f
)
select * from HRH

此查询给出以下输出：

XML_ID CHG_TIMESTR    CHG_DATE   CHG_TIME FIELDNAME OLD_VALUE NEW_VALUE
====== ============== ========== ======== ========= ========= =========
     1 20170330100429 2017-03-30   100429 Name      Fran      Frank    
     1 20170330100429 2017-03-30   100429 Zip       13583     13853    
     1 20170401094783 2017-04-01    94783 MI        J         A        
     1 20170401094783 2017-04-01    94783 Suffix    Jr        III     
====== ============== ========== ======== ========= ========= =========

如果相反，我使用以下任一方法：

select * from HRH h1, HRH h2
select * from HRH h1 join HRH h2 on 1=1

我收到一条错误消息：

Line Pos Text
==== === ================================================
17   23  ORA-19032: Expected XML tag , got no content
ORA-06512: at "SYS.XMLTYPE", line 310
ORA-06512: at line 1
==== === ================================================

我怎样才能让它工作？我的总体目标是为我的数据集获取给定 xml_id-field 组合的最新更新。

Answer 1

不确定您为什么会看到这种行为；实现第一个 CTE 会停止错误，但随后找不到任何数据。

不直接相关，但您可以将查询（至少对于您显示的数据）简化为单个 XMLTable 调用：

SELECT au.xml_id,
       x.tran_ts as chg_timestr,
       to_date(substr(x.tran_ts, 0, 8), 'yyyymmdd') chg_date,
       cast(substr(x.tran_ts, 9, 6) as int) chg_time,
       x.fieldname,
       x.ov old_value,
       x.nv new_value
FROM myxml au
CROSS JOIN xmltable('/xml/entry/field'
  PASSING XMLTYPE(au.x)
  COLUMNS tran_ts   VARCHAR2(16) PATH './../@timestamp',
          fieldname VARCHAR2(250) PATH '@name',
          nv VARCHAR2(250) PATH 'nv',
          ov VARCHAR2(250) PATH 'ov'
) x;

    XML_ID CHG_TIMESTR      CHG_DATE     CHG_TIME FIELDNAME  OLD_VALUE  NEW_VALUE 
---------- ---------------- ---------- ---------- ---------- ---------- ----------
         1 20170330100429   2017-03-30     100429 Name       Fran       Frank     
         1 20170330100429   2017-03-30     100429 Zip        13583      13853     
         1 20170401094783   2017-04-01      94783 MI         J          A         
         1 20170401094783   2017-04-01      94783 Suffix     Jr         III

尝试自联接获取数据，但有很多空值。也奇怪。

但即使您的实际查询太复杂而无法做到这一点（毕竟您说您已经将其精简），您说的目标是获取最新更新，您可以这样做：

with myxml as
(
  select /*+ materialize */ 1 xml_id,
         '<xml> <entry timestamp="20170330100429" effective="20170329"> <field name="Name"> <ov><![CDATA[Fran]]></ov> <nv><![CDATA[Frank]]></nv> </field> <field name="Zip"> <ov><![CDATA[13583]]></ov> <nv><![CDATA[13853]]></nv> </field> </entry> <entry timestamp="20170401094783" effective="20170331"> <field name="MI"> <ov><![CDATA[J]]></ov> <nv><![CDATA[A]]></nv> </field> <field name="Suffix"> <ov><![CDATA[Jr]]></ov> <nv><![CDATA[III]]></nv> </field> </entry> </xml>' x from dual
),
HRH as
(
  SELECT au.xml_id,
         x.tran_ts as chg_timestr,
         to_date(substr(x.tran_ts, 0, 8), 'yyyymmdd') chg_date,
         cast(substr(x.tran_ts, 9, 6) as int) chg_time,
         x.fieldname,
         x.ov old_value,
         x.nv new_value,
         rank() over (partition by au.xml_id, x.fieldname
           order by x.tran_ts desc) rnk
  FROM myxml au
  CROSS JOIN xmltable('/xml/entry/field'
    PASSING XMLTYPE(au.x)
    COLUMNS tran_ts   VARCHAR2(16) PATH './../@timestamp',
            fieldname VARCHAR2(250) PATH '@name',
            nv VARCHAR2(250) PATH 'nv',
            ov VARCHAR2(250) PATH 'ov'
  ) x
)
select * from HRH
where rnk = 1;

因为每个 ID/field 您只有一组数据，可以为您提供与样本相同的结果，但如果有多个，则应该为每个数据提供最新的数据。您还应该查看 rank() 和 dense_rank() 分析函数之间的区别，并决定您想要查看的内容是否存在联系 - 如果您的数据中可能存在联系。

Oracle：自连接 XMLTable 输出

Oracle: Self join XMLTable output

oracle

self-join

common-table-expression

xmltable