Oh yeah, we also tested with the DBMS_XMLQuery package - it seemed to take about as long as the Oracle Serializer java code. On Tue, Aug 2, 2011 at 6:19 PM, Jeremy Schneider < jeremy.schneider@xxxxxxxxxxxxxx> wrote: > Well we're not making any 1GB files yet, but I can't really imagine what > anyone does with a 400MB file either. Anyway - we were doing some testing > with a much smaller "test" dataset (8MB file), and we tried Marco's idea. > Here are the results: > > Original Code: 96 sec > Oracle Serializer: 40 sec > Marco's Code: 3 sec > > :) > > Sweet! However, it seems that there's an implicit conversion from XMLType > to CLOB and it's not immediately obvious how to control the XML that's > generated. I guess it's time for me to learn some more about XMLDB. :-/ > Here's what we're trying to figure out how to do with XMLType: > > - xmlQuery.keepCursorState(true); > - xmlQuery.setMaxRows(pRowLimit_variable); // Note: we continue with > cursor and split output to multiple files > - xmlQuery.setRowTag(pRowHeader_variable); > - xmlQuery.setRowsetTag(pRowHeader_variable+"_MYSUFFIX"); > - xmlQuery.setDateFormat("yyyy-MM-dd'T'HH:mm:ss"); > - xmlQuery.setStylesheetHeader(pXSLtURI_variable); > > Any tips would be appreciated... > > -J > > > PS - this blog post by Marco is excellent: > http://www.liberidu.com/blog/?p=369 > > > On Tue, Aug 2, 2011 at 4:41 PM, Marco Gralike <Marco.Gralike@xxxxxxx>wrote: > >> ONE 1++GB in XML document, are you serious. >> Wouldn't be surprised though. I see it more and more often. ;-) >> >> No the stuff showed below is in C and/or even part of the C kernel. It is >> possible to do in SQL and yes due to the fact that it is 1 XML document you >> will get in trouble regarding those mentioned sizes per XML document. I am >> guessing, although never really checked that if you switched it again for a >> C based SAV serializer it would be even faster or does it reside in the JVM >> database kernel? >> >> I wonder which part picks up those big XML documents and tries to do >> something useful with it. Work to be done with those documents afterwards >> most be very resource intensive ;-) >> > -- http://www.ardentperf.com +1 312-725-9249 Jeremy Schneider Chicago