Import Large XML File to Table

Question

I have a large (819MB) XML file I'm trying to import into a table in the format:
<ROW_SET>
<ROW>
<column_name>value</column_name>
</ROW>
<ROW>
<column_name>value</column_name>
</ROW>
</ROW_SET>
I've tried importing it with xmlsequence(...).extract(...) and ran into the number of nodes exceed maximum error.
I've tried importing it with XMLTable(... passing XMLTYPE(bfilename('DIR_OBJ','large_819mb_file.xml'), nls_charset_id('UTF8'))) and I gave up after it ran for 15+ hours ( COLLECTION ITERATOR PICKLER FETCH issue ).
I've tried importing it with:
insCtx := DBMS_XMLStore.newContext('schemaname.tablename');
DBMS_XMLStore.clearUpdateColumnList(insCtx);
DBMS_XMLStore.setUpdateColumn(insCtx,'column1name');
DBMS_XMLStore.setUpdateColumn(insCtx,'columnNname');
ROWS := DBMS_XMLStore.insertXML(insCtx, XMLTYPE(bfilename('DIR_OBJ','large_819mb_file.xml'), nls_charset_id('UTF8')));
and ran into ORA-04030: out of process memory when trying to allocate 1032 bytes (qmxlu subheap,qmemNextBuf:alloc).
All I need to do is read the XML file and move the data into a matching table in a reasonable time. Once I have the data in the database, I no longer need the XML file.
What would be the best way to import large XML files?
Oracle Database 11g Release 11.2.0.1.0 - 64bit Production
PL/SQL Release 11.2.0.1.0 - Production
"CORE     11.2.0.1.0     Production"
TNS for Linux: Version 11.2.0.1.0 - Production
NLSRTL Version 11.2.0.1.0 - Production

Answer 1

This (rough) approach should work for you.
CREATE TABLE HOLDS_XML
        (xml_col XMLTYPE)
      XMLTYPE xml_col STORE AS SECUREFILE BINARY XML;
INSERT INTO HOLDS_XML
VALUES (xmltype(bfilename('DIR_OBJ','large_819mb_file.xml'), nls_charset_id('UTF8')))
-- Should be using AL32UTF8 for DB character set with XML
SELECT ...
  FROM HOLD_XML HX
       XMLTable(...
          PASSING HX.xml_col ...)How it differs from your approach.
By using the HOLDS_XML table with SECUREFILE BINARY XML storage (which became the default in 11.2.0.2) we are providing a place for Oracle to store a parsed version of the XML. This allows the XML to be stored on disk instead of in memory. Oracle can then access the needed pieces of XML from disk by streaming them instead of holding the whole XML in memory and parsing it repeatedly to find the information needed. That is what COLLECTION ITERATOR PICKLER FETCH means. A lot of memory work. You can search on that term to learn more about it if needed.
The XMTable approach then simply reads this XML from disk and should be able to parse the XML with no issue. You have the option of adding indexes to the XML, but since you are simply reading it all one time and tossing it, there is no advantage to indexes (most likely)