You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@openoffice.apache.org by Rob Weir <ro...@apache.org> on 2011/10/10 22:24:44 UTC
Slooooow spreadsheet documents
Does anyone have a good example of a really slow spreadsheet document?
Preferably ones that was slow due to computation/calculation, and one
that is not merely slow due to size alone.
I was talking to someone earlier who had ideas for speeding up
spreadsheet computation through parallelization, and was looking for
some notoriously slow test documents.
-Rob
---------------------------------------------------------------------
To unsubscribe, e-mail: ooo-users-unsubscribe@incubator.apache.org
For additional commands, e-mail: ooo-users-help@incubator.apache.org
Re: Slooooow spreadsheet documents
Posted by Rob Weir <ro...@apache.org>.
On Mon, Oct 10, 2011 at 5:35 PM, Joost Andrae <Jo...@gmx.de> wrote:
> Hi Rob,
>
> just a remark:
>
> In my private opinion it would make more sense to invest resources into
> impementing a formula reference which points to expanded formula cells
> (/edit/fill) like it is already done for formatting templates within the
> sheet.
This is probably true. But this isn't my time investment ;-) It is
someone that has a tool for analyzing code and refactoring it for
parallel computation. They sounded willing to contribute their
changes back to AOOo. That's why I'm looking for examples of slow
spreadsheets that are computation-bound rather than memory bound, even
thought that might not be the typical case.
> A slow sheet doesn't necessarily mean that calculation is slow. Better
> resolving references within the sheet (in my opinion) has more potential to
> make such documents faster. I believe by using formula references the
> documents' memory footprint could be lowered as well because a formula
> reference doesn't cost as much as a cell content containing a formula and
> containing edit engine data.
>
> At one of the OOo conferences I vaguely remember there was a university
> project that already parallelized computation of Calc sheets. Unfortunately
> I cannot remember the project/university name anymore. Probably having a
> look at the conference session drafts might give us some more insight.
>
>>
>> I was talking to someone earlier who had ideas for speeding up
>> spreadsheet computation through parallelization, and was looking for
>> some notoriously slow test documents.
>
> Kind regards, Joost
>
>
Re: Slooooow spreadsheet documents
Posted by Joost Andrae <Jo...@gmx.de>.
Hi,
> At one of the OOo conferences I vaguely remember there was a university
> project that already parallelized computation of Calc sheets.
> Unfortunately I cannot remember the project/university name anymore.
> Probably having a look at the conference session drafts might give us
> some more insight.
...it wasn't directly about parallelization but about using Calc as a
frontend for grid computing. I've found the conference abstract about
the Tavernalc project on the OOo Marketing conference page of the
conference in Barcelona.
http://marketing.openoffice.org/ooocon2007/programme/thursday_171.pdf
Kind regards, Joost
Re: Slooooow spreadsheet documents
Posted by Joost Andrae <Jo...@gmx.de>.
Hi Rob,
just a remark:
In my private opinion it would make more sense to invest resources into
impementing a formula reference which points to expanded formula cells
(/edit/fill) like it is already done for formatting templates within the
sheet.
A slow sheet doesn't necessarily mean that calculation is slow. Better
resolving references within the sheet (in my opinion) has more potential
to make such documents faster. I believe by using formula references the
documents' memory footprint could be lowered as well because a formula
reference doesn't cost as much as a cell content containing a formula
and containing edit engine data.
At one of the OOo conferences I vaguely remember there was a university
project that already parallelized computation of Calc sheets.
Unfortunately I cannot remember the project/university name anymore.
Probably having a look at the conference session drafts might give us
some more insight.
>
> I was talking to someone earlier who had ideas for speeding up
> spreadsheet computation through parallelization, and was looking for
> some notoriously slow test documents.
Kind regards, Joost
Re: Slooooow spreadsheet documents
Posted by 依瑪貓 <im...@mail.imacat.idv.tw>.
Dear Rob,
Last time I encounter it, it was an Excel document with 15000+ empty
(and hence invisible) text frames. I suppose this was cause by many
users copy-paste without noticing it. And I got this complaint since
user does not notice this in MS Excel, which I assume that 15000+ text
frames does not trouble MS Excel.
I guess this is about mouse moving triggered event, so I suppose
this is not memory-bound, but calculation-bound. I shall see if I can
find this fine example Excel document.
On 2011/10/11 04:24, Rob Weir said:
> Does anyone have a good example of a really slow spreadsheet document?
> Preferably ones that was slow due to computation/calculation, and one
> that is not merely slow due to size alone.
>
> I was talking to someone earlier who had ideas for speeding up
> spreadsheet computation through parallelization, and was looking for
> some notoriously slow test documents.
>
> -Rob
--
依瑪貓 imacat ^_*' <im...@mail.imacat.idv.tw>
PGP Key http://www.imacat.idv.tw/me/pgpkey.asc
旅舍依瑪 http://www.imacat.idv.tw/ 《女聲》電子報 http://www.wov.idv.tw/
台灣女子自由軟體工作小組 http://wofoss.blogspot.com/
OpenOffice.org http://www.openoffice.org/
EducOO/OOo4Kids台灣 http://www.educoo.tw/
Re: Slooooow spreadsheet documents
Posted by Rob Weir <ro...@apache.org>.
On Mon, Oct 10, 2011 at 5:33 PM, Andreas Säger <vi...@t-online.de> wrote:
> Am 10.10.2011 22:24, Rob Weir wrote:
>>
>> Does anyone have a good example of a really slow spreadsheet document?
>> Preferably ones that was slow due to computation/calculation, and one
>> that is not merely slow due to size alone.
>>
>
> Database surrogate (complete madness in millions of xls files):
> X1 =IF(ISNA(VLOOKUP(value;range;2;0)));"";VLOOKUP(value;range;2;0))
> Y1 =IF(ISNA(VLOOKUP(value;range;3;0)));"";VLOOKUP(value;range;3;0))
> Z1 =IF(ISNA(VLOOKUP(value;range;4;0)));"";VLOOKUP(value;range;4;0))
> Performs 3 lookups at least. If the test for error is negative then it
> performs 6 lookups to find the exact same row.
>
So we don't cache the partial evaluations internally?
> Objects overkill:
>>
>> http://user.services.openoffice.org/en/forum/viewtopic.php?f=9&t=44259
>
>
> This one does not even have any data nor formulas:
> File>New>Spreadsheet
> Ctrl+Space (select entire column)
> Shift+<right arrow> x4 (expand 4 columns to the right)
> Format>AutoFormat...
> Choose "Yellow", [OK]
> Get a cup of tea.
> killall soffice.bin
>
Extremely large sheets are a problem. Remember this file?
http://www.zdnet.com/blog/ou/openofficeorg-20-is-here-but-is-it-a-pig/119
The issue is users are so comfortable with spreadsheets that they use
them for problems that are better handled in a database. We're cursed
by the success of the spreadsheet interface.
I did some optimizations of that particular case a while back,
focusing on optimizing the markup we write out. I'll see if I can
find it.
But thanks for the examples. Very useful.
>
Re: Slooooow spreadsheet documents
Posted by Andreas Säger <vi...@t-online.de>.
Am 10.10.2011 22:24, Rob Weir wrote:
> Does anyone have a good example of a really slow spreadsheet document?
> Preferably ones that was slow due to computation/calculation, and one
> that is not merely slow due to size alone.
>
Database surrogate (complete madness in millions of xls files):
X1 =IF(ISNA(VLOOKUP(value;range;2;0)));"";VLOOKUP(value;range;2;0))
Y1 =IF(ISNA(VLOOKUP(value;range;3;0)));"";VLOOKUP(value;range;3;0))
Z1 =IF(ISNA(VLOOKUP(value;range;4;0)));"";VLOOKUP(value;range;4;0))
Performs 3 lookups at least. If the test for error is negative then it
performs 6 lookups to find the exact same row.
Objects overkill:
> http://user.services.openoffice.org/en/forum/viewtopic.php?f=9&t=44259
This one does not even have any data nor formulas:
File>New>Spreadsheet
Ctrl+Space (select entire column)
Shift+<right arrow> x4 (expand 4 columns to the right)
Format>AutoFormat...
Choose "Yellow", [OK]
Get a cup of tea.
killall soffice.bin
Re: Slooooow spreadsheet documents
Posted by Pedro Giffuni <pf...@apache.org>.
Hi Rob;
If they want to design their own acid tests ...
Point them to the Matrix Market:
http://math.nist.gov/MatrixMarket/
And if they want cool algorithms point them to
MTL:
http://www.osl.iu.edu/research/mtl/
and the Coin-OR OSI interface:
https://projects.coin-or.org/Osi
Pedro.
--- On Mon, 10/10/11, Rob Weir <ro...@apache.org> wrote:
> From: Rob Weir <ro...@apache.org>
> Subject: Slooooow spreadsheet documents
> To: ooo-users@incubator.apache.org, ooo-dev@incubator.apache.org
> Date: Monday, October 10, 2011, 3:24 PM
> Does anyone have a good example of a
> really slow spreadsheet document?
> Preferably ones that was slow due to
> computation/calculation, and one
> that is not merely slow due to size alone.
>
> I was talking to someone earlier who had ideas for speeding
> up
> spreadsheet computation through parallelization, and was
> looking for
> some notoriously slow test documents.
>
> -Rob
>
---------------------------------------------------------------------
To unsubscribe, e-mail: ooo-users-unsubscribe@incubator.apache.org
For additional commands, e-mail: ooo-users-help@incubator.apache.org
Re: Slooooow spreadsheet documents
Posted by Pedro Giffuni <pf...@apache.org>.
Hi Rob;
If they want to design their own acid tests ...
Point them to the Matrix Market:
http://math.nist.gov/MatrixMarket/
And if they want cool algorithms point them to
MTL:
http://www.osl.iu.edu/research/mtl/
and the Coin-OR OSI interface:
https://projects.coin-or.org/Osi
Pedro.
--- On Mon, 10/10/11, Rob Weir <ro...@apache.org> wrote:
> From: Rob Weir <ro...@apache.org>
> Subject: Slooooow spreadsheet documents
> To: ooo-users@incubator.apache.org, ooo-dev@incubator.apache.org
> Date: Monday, October 10, 2011, 3:24 PM
> Does anyone have a good example of a
> really slow spreadsheet document?
> Preferably ones that was slow due to
> computation/calculation, and one
> that is not merely slow due to size alone.
>
> I was talking to someone earlier who had ideas for speeding
> up
> spreadsheet computation through parallelization, and was
> looking for
> some notoriously slow test documents.
>
> -Rob
>