Giuseppe Bilotta wrote:
Wednesday, November 2, 2005 Alex Janssen wrote:
The test .SXC file, from the artical at
http://blogs.zdnet.com/Ou/?p=119, took about 23 minutes to load on my
2.1 GHZ Celeron Dell Inspiron 1100 running Ubuntu 5.10 and OOo 1.9.129
(Ubuntu has not packaged OOo2.0 yet). It amounts to 3,386,656 cells
spread out evenly on 16 sheets in one file. This is a massive amount of
XML interpretation. Dealing with interpretation of language file in
ascii format is always going to be slower than a memory image binary
format file. That's also the reason processor native binary formatted
computer programs run so much faster than interpreted language programs
stored in ascii format files. Imagine how slow OOo would be if it was
written in an interpreted language!
Actually I think quite a bit of the time is spent on
decompression of the file: the content.xml is some 283
megabytes of stuff. Try renaming the file in .zip and taking
content.xml out. See how much time it takes. Ok, parsing
such an humongous time also takes quite a lot. Interesting,
by contrast, the MS XML is 'only' 192 megs, and with more
liberal whitespace too. (I don't actually have MSO 2003 or
12 or whatever so I can't run the runtime comparison,
though.)
OpenOffice.org's fileformat is rather verbose even for an
XML format (I think I raised this issue once, after
discovering this while I was trying to hand-edit it.) The
usual reply has been that since it's compressed it didn't
count much. I wasn't much convinced then, and I'm not much
more convinced now either.
So the only test I run was opening the file in 1.1.3 and 2.0
(the latter took considerably less time) and then saving it
under ODS from version 2.0; did you actually try saving the
file? It takes quite a lot of time too (and eats up a LOT of
disk space while saving ... temp files, I guess, although
I'd highly prefer these were handled in memory if possible)
and the result is that: the ODF file is *even bigger* than
the SXC file!
content.xml went from
293,059,603 to
344,629,118
That's 117.6%, if you don't want to calculate it for
yourself :)
This is *most definitely* an area that needs a lot of work.
It is not the compression at all. I just did a test on two really
large ods/sxw files and in both cases the compression was less than 10
seconds. Both files take minutes to load or save in calc. I believe
the one file may be the example file from the ZDNet article but I
cannot confirm at this time.
<200264-l.sxc>
I tried to open the content.xml file in an external editor and this
took allot of time. The content.xml file in the one test file is 234M
in size. I never timed it but I am going to add these comments to the
issue about slow loading.
The parsing of the XML is the issue.
--
Robin Laing
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
