Yes, that is the correct track. There have been improvements in the last 10 days.
(1) Get hold of the latest night build. (2) Find the email thread titled: "Performance Question with CTSheetDataImpl.java" in the archives. The short version is that Bryce Alcock analyzed where bad performance occurred and then later Yegor made some targeted improvements which yield significant improvements for "in order" creation of XSSF files. These are very likely the improvements that are sought. Regards, Dave On Jun 9, 2010, at 12:28 PM, <[email protected]> <[email protected]> wrote: > This was suggested by Yegor Kozlov on Dec 7, 2008: > > Yegor - I created an example demonstrating how to generate large workbooks > and avoid OutOfMemory: > http://svn.apache.org/repos/asf/poi/trunk/src/examples/src/org/apache/poi/xssf/usermodel/examples/BigGridDemo.java > > If you search this list you can find his original email explaining it in more > detail. > > Justin > > -----Original Message----- > From: K raghavendra Rao [mailto:[email protected]] > Sent: Tuesday, June 08, 2010 3:06 PM > To: [email protected] > Subject: Re: Performance Issue with POI 3.6 as compared to 2.5.1 > > Nick, David, > Thank you both for your response. > > I got held up with other work and hence couldn't respond earlier. > > Please clarify the following. >>> My hunch is that you'll find HSSFWorkbook from > 3.6 to be slightly faster than from 2.5, or otherwise little different.<< > > My understanding is that, to be able to generate .xlsx files I need to use > XSSFWorkbook and NOT HSSFWorkbook. Hence that rules out the possibility of > using HSSFWorkbook. Please correct me if I am wrong. > > Based on David's reply, here is what I tried. > <This is the first time I am working with NIGHTLY builds. Please correct me > if my approach is wrong> > I downloaded the poi-3.7-SNAPSHOT-20100528.jar NIGHTLY build and replaced the > earlier one: poi-3.6-20091214.jar > > So now I have 2 environments with the following settings to test the > PERFORMANCE between POI 3.6 and 2.5.1. > > Env1: > POI version 3.6 with XSSFWorkbook (updated with the above mentioned NIGHTLY > build jar. Other 3.6 jars are the same) > > Env2: > POI version 2.5.1 with HSSFWorkbook > > The report that I am generating has a SQL SELECT query which returns 65,000 > records in 2 seconds. Env2 provides the file in less than 10 seconds. Env1 > takes around 15 mins!! > > The BIG QUESTION for my project team is: Can POI EFFICIENTLY support > generation of MS Excel 2007 (.xlsx) files which have more than 66,000 > records? Now, I need to be able to make this decision to accordingly convey > to the management. We had migrated from .CSV files to POI due to the user > preference of native MS Excel files over CSV. > > If anybody has managed to achieve this, PLEASE HELP. > > Let me know if you need any further details. > > Regards, > Raghu > > > > ----- Original Message ----- > From: [email protected] > To: [email protected] > Sent: Tuesday, May 25, 2010 5:33:51 AM GMT -05:00 US/Canada Eastern > Subject: Re: Performance Issue with POI 3.6 as compared to 2.5.1 > > On Mon, 24 May 2010, K raghavendra Rao wrote: >> I was using POI 2.5.1 to generate .xls files until the record count >> crossed the 65k+ limit set by Excel 2003. At this point, I switched to >> POI 3.6 and to XSSFWorkbook (from the previous HSSFWorkbook) > > This'll be the main cause. My hunch is that you'll find HSSFWorkbook from > 3.6 to be slightly faster than from 2.5, or otherwise little different. > > XSSFWorkbook is xml based (the whole of the ooxml file format is), and > processing it needs a bit more memory and cpu than the older binary > format. > > Otherwise, see David's reply about some recent xssf performance > improvements > > Nick > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
