Apache OpenOffice (AOO) Bugzilla – Issue 17565
Bad performance for opening files with very large tables.
Last modified: 2013-08-07 14:38:26 UTC
File with very large table included in attachment (219 A4 Landscape pages in doc format crated by MS Word) is opened more than 3 minutes even when converted to native sxw format. Compare this to MS Word which opens this file in a few seconds an renumbers pages in about 10 seconds on the same machine. Another problem: original MS Word 219 pages size is converted to 239 pages by OO word processor.
Created attachment 8113 [details] Example of large table for issue 17565
Maybe this preformance problem is not important for home usage of OO, but it prevents us from introducing OpenOffice to our business.
Confirmed -- WinXPPro/OOo 1.1rc2 ZIP file is 418 KB Document is 2475 KB CPU usage 100% when opening .doc file, saving in .sxw format, and reloading resulting .sxw file. Time estimates are accurate -- three to four minutes each to open .doc file, to save in .swx format and to reload .sxw. OOo document contains 239 pages Intel Celeron - 2.30Ghx - 512 MB RAM
1. I measured performance of loading this doc document on much slower machine: Pentium 2 400MHz 256 MB RAM and got times in the same range as reported for Celeron 2.3 GHz! Maybe the problem is in communication between threads, sleeps or other events synchronised by external clock? During my tests paging space was not used at all by Windows 2000. 2. I tested the same file with older OpenOffice 1.0.2 included as a part of RedHat 9 distribution. Results are similar (slower: 6.5 minutes for loading and displaying - swap was only lightly used by Pentium 2 400MHz 256 MB RAM machine, also usage of 100% of processor power was reported), so this performance problem is possibly OS independent.
HI->CMC: It is a little bit faster when the doc is resaved with OO.
Slow alright, I have a cache to speed this up. I can try increasing it, otherwise need to rethink table parsing.
cmc->mmaher: Performance enhancements are always welcome. Tables and slow insertion of hard coded properties is the problem. The .sxw format is about as slow to load...
*** Issue 26785 has been marked as a duplicate of this issue. ***
I almost filed a new issue, but decided to try hopping on the CC list for this one and add another data point. Unfortunately I cannot attach the document as it's confidential info for the co. This slow loading problem is getting to be a big issue for us, because it's an important specification document that is looked at by many people and frequently edited. Everyone's just dealing with it for now but there are increasing amounts of grumbling as the document grows and load times increase. OOo versions: 1.1.4 and 1.9.91 behave the same. Saved as either .sxw or .odt, same behavior too. It contains ~20 relatively complex drawings from OOo Draw. After I initiate an open of the file, OOo completely freezes for 75 sec. Then the progress bar starts to slowly make its way to the right. After 3:37 (217 sec) the document is opened. Until it is opened, the CPU is pegged. Far, far too long to wait. OOo versions: 1.1.4 and 1.9.91 behave identically in this regard. OSes: Windows XP pro, Linux CPU/memory: P-III mobile 1.4-1.6 GHz, 512 MB memory (not slow processors) Document file size: 569 Kbytes. unzip -v shows uncompressed size of 9.3 MB. Document properties: pages:83, tables:24, graphics:2, OLE objects:20, paragraphs:1838, words:20209, characters:124765 The OLE objects are all from OOo Draw. Note this document is considerably smaller than most others for which this problem has been reported, yet takes as long or longer to load.
mmaher->flr: Yours I think
*** Issue 55616 has been marked as a duplicate of this issue. ***
Update for OO 2.0: HW: Pentium 4 2.8 GHz 1 GB RAM Opening doc attachement takes about 9:30 minutes on Windows XP SP2 and about 8:30 using Open SuSE 10.0 distro (OO compiled by SuSE). Previous OO 1.1RC2 made this in less than 4 minutes on Pentium 2 400MHz 256MB RAM!!! I wonder what will be the time in next main OO release. Opening odt file created from doc attachment takes about 1:30 minutes on both platforms (no improvement to previous result for Pentium II 400MHz and OO 1.1RC2 - Pentium 4 2.8GHz has 7 times faster clock - just mentioning only this difference). For comparison: MS Word 2002 SP2 opens this doc in 2-3 seconds on the same platform. Oryginal number of 219 pages in MS Word is converted by OO 2.0 to 224 pages. This doc attachement is a part of "Integrated tariff of the European Communities" - 25 countries now.
This old and bad performance issue already exist since 2003. and the comparisson is bad for OOo 9:30 minutes on OOo and 2-3 seconds only on microsoft Word and this documents used by 25 countries now. I hope OOo fix this. Thanks
Well Andreas is doing the table refactoring --- I guess.
->FLR: You have analyzed this problem and you detect the root cause in our old table model?
*** Issue 95894 has been marked as a duplicate of this issue. ***
Update for OO.3.0 Windows XP SP3 PL: open time for doc from attachmentid=8113 is about 3 minutes. for Word 2002 SP3 English it is about 1-2 seconds. The hardware is: Intel Core2 Duo T7200 2GHz, 2GB RAM.OO opens odt version about 40 seconds. One processor core is used in 100% during open operation. The performance improvement compared to the previous 9:30 time may come mostly from faster HW I am afraid.
Tables don't have to be very large for noticeable slowness--see Issue 95894 attachment..