PL\SQL XMLTable性能

Question

我有一个解析XMLTYPE变量的函数,对于每个消息,为XMLTYPE变量中的每个标记附加一个具有特定结构的CLOB.像这样:

FUNCTION myFunc (px_Header      IN     VARCHAR2,
                                px_Block       IN     XMLTYPE,
                                pn_numLines      OUT PLS_INTEGER)
      RETURN CLOB
   IS
      lcl_return   CLOB := EMPTY_CLOB;
   BEGIN
      pn_numLines := 0;

      FOR item
         IN (          SELECT RPAD (NVL (RECEIPTNUMBER, ' '), 20) AS RECEIPTNUMBER,
                              RPAD (NVL (COMPANYCODE, ' '), 3) AS COMPANYCODE,
                              RPAD (NVL (BRAND, ' '), 3) AS BRAND,
                              RPAD (NVL (POLICYNUMBER, ' '), 20) AS POLICYNUMBER,
                              RPAD (NVL (CLAIMNUMBER, ' '), 20) AS CLAIMNUMBER,
                              RECEIPTAMOUNT
                                 AS receiptAmount
                         FROM XMLTABLE (
                                 'INT_DATA/Item'
                                 PASSING px_Block
                                 COLUMNS RECEIPTNUMBER   VARCHAR2 (20)
                                                            PATH 'RECEIPTNUMBER',
                                         COMPANYCODE     VARCHAR2 (3)
                                                            PATH 'COMPANYCODE',
                                         BRAND           VARCHAR2 (3) PATH 'BRAND',
                                         POLICYNUMBER    VARCHAR2 (20)
                                                            PATH 'POLICYNUMBER',
                                         CLAIMNUMBER     VARCHAR2 (20)
                                                            PATH 'CLAIMNUMBER',
                                         RECEIPTAMOUNT   VARCHAR2 (15)
                                                            PATH 'RECEIPTAMOUNT'))
      LOOP
         lcl_return:=
               lcl_return
            || px_Header
            || 'B2'
            || item.RECEIPTNUMBER
            || item.COMPANYCODE
            || item.BRAND
            || item.POLICYNUMBER
            || item.CLAIMNUMBER
            || item.RECEIPTAMOUNT
            || CHR (13)
            || CHR (10);
         pn_numLines := pn_numLines + 1;
      END LOOP;

      RETURN lcl_return;
   END myFunc ;

如果我有一个小的,这很好用px_Block.但我有一个案例,我可以有一个大的XMLTYPE,这个功能需要很长时间.我是新用的XMLType和XMLTable.我能做些什么来提高性能.也许使用BULK COLLECT语句？

菲利普,提前谢谢

EDIT1:这是仅针对两个实例的XML示例.

<INT_DATA>
 <Item>
  <RECEIPTNUMBER>1</RECEIPTNUMBER>
  <COMPANYCODE>148</COMPANYCODE>
  <BRAND>006</BRAND>
  <POLICYNUMBER>72972</POLICYNUMBER>
  <CLAIMNUMBER>2015101504</CLAIMNUMBER>
  <RECEIPTAMOUNT>-10.00</RECEIPTAMOUNT>
 </Item>
 <Item>
  <RECEIPTNUMBER>1</RECEIPTNUMBER>
  <COMPANYCODE>148</COMPANYCODE>
  <BRAND>006</BRAND>
  <POLICYNUMBER>73785</POLICYNUMBER>
  <CLAIMNUMBER>2015101505</CLAIMNUMBER>
  <RECEIPTAMOUNT>-22.50</RECEIPTAMOUNT>
 </Item>
</INT_DATA>

EDIT2:我的功能发生了一些变化,改进了20%.我改变了使用临时CLOB附加CLOB的方式.但更好的解决方案将会有所帮助.

FUNCTION myFunc (px_Header      IN     VARCHAR2,
                                px_Block       IN     XMLTYPE,
                                pn_numLines      OUT PLS_INTEGER)
      RETURN CLOB
   IS
      lcl_return   CLOB := EMPTY_CLOB;
      v_tmp_clob    CLOB := EMPTY_CLOB;
   BEGIN
      pn_numLines := 0;

      FOR item
         IN (          SELECT RPAD (NVL (RECEIPTNUMBER, ' '), 20) AS RECEIPTNUMBER,
                              RPAD (NVL (COMPANYCODE, ' '), 3) AS COMPANYCODE,
                              RPAD (NVL (BRAND, ' '), 3) AS BRAND,
                              RPAD (NVL (POLICYNUMBER, ' '), 20) AS POLICYNUMBER,
                              RPAD (NVL (CLAIMNUMBER, ' '), 20) AS CLAIMNUMBER,
                              RECEIPTAMOUNT
                                 AS receiptAmount
                         FROM XMLTABLE (
                                 'INT_DATA/Item'
                                 PASSING px_Block
                                 COLUMNS RECEIPTNUMBER   VARCHAR2 (20)
                                                            PATH 'RECEIPTNUMBER',
                                         COMPANYCODE     VARCHAR2 (3)
                                                            PATH 'COMPANYCODE',
                                         BRAND           VARCHAR2 (3) PATH 'BRAND',
                                         POLICYNUMBER    VARCHAR2 (20)
                                                            PATH 'POLICYNUMBER',
                                         CLAIMNUMBER     VARCHAR2 (20)
                                                            PATH 'CLAIMNUMBER',
                                         RECEIPTAMOUNT   VARCHAR2 (15)
                                                            PATH 'RECEIPTAMOUNT'))
      LOOP
         v_tmp_clob :=
           TO_CLOB (px_Header)
        || TO_CLOB ('B2')
        || TO_CLOB (item.RECEIPTNUMBER)
        || TO_CLOB (item.COMPANYCODE)
        || TO_CLOB (item.BRAND)
        || TO_CLOB (item.POLICYNUMBER)
        || TO_CLOB (item.CLAIMNUMBER)
        || TO_CLOB (item.RECEIPTAMOUNT)
        || CHR (13)
        || CHR (10);

     lcl_return := lcl_return || v_tmp_clob;

         pn_numLines := pn_numLines + 1;
      END LOOP;

      RETURN lcl_return;
   END myFunc ;

Answer 1

问题XMLTABLE在于Oracle使用DOM Parser来读取XML.这意味着整个XML必须以很大的开销加载到内存中.看看DBMS_XMLSTORE.该软件包使用SAX Parser,它是事件库(或基于流).

使用SAX Parser,您可以使用普通PC读取大小为几千兆字节的XML文件.

但是,直到 - 比方说50-100 MByte - DOM解析器应该运行良好.