Apache OpenOffice (AOO) Bugzilla – Issue 90561
Velocity to handle a big number of unsteady data points
Last modified: 2013-02-24 21:20:04 UTC
Dear Devolepers I found that the time that the cal application takes to make simple graphs with a lot of data points (as 35000 data points without any additional correlation), it take forever to be processed (forever respect excel in windows and Gnumeric Spreadsheet in Linux) Thanks for you work
Created attachment 54375 [details] ultra slow creation of graph and editing
duplicate *** This issue has been marked as a duplicate of 82869 ***
No, this is not duplicate to i82869 as that issue is especially about smoothed 3D lines. but the attached scrrenshot ois not a 3D chart. -> reopen
@triathman, could you please attach the spreadsheet document and describe which action exactly takes to long (for example selecting a series or leaving the chart) ? Thanks, Ingrid.
added keyword needmoreinfo
@iha with pleasure; the slow process are pretty much all, 35000 points in a 2D graph the slow steps concerns all the the graph treatment, from the creation of the graph itself to any changes done in the graph with this amount of datapoints (even the selection to have the graph option available take a long time). I feel already a noticeable speed reduction with 3500 point, with 35000 it became really too long (matter of minutes to treat it). The only part that doesn't have problem is the movement of the graph within the spreadsheet sheet. Thank you for your work
Created attachment 54448 [details] ultra slow creation of graph and editing spreadsheet, 35000 data points 2D
Sorry, I can't post the spreadsheet because it's 2.1Mb, anyway it's just a 2 columns 35000 rows long plotted as a (x,y) graph
@triathman, OpenOffice.org has a mechanism to render continues data faster. So it depends also on the values of your data how fast they are rendered. If you have completely unrelated random data probably for each of the 35 000 points a symbol will be drawn. In contrast when you have a continues curve for example a sinus a new shape is only created if it is different from its predecessor for the given resolution. Maybe can you send me your document via mail? Thanks, Ingrid.
Ah, no need to send. The attached doc is complete. And in fact the data seems to be jumping around unsteadily, which causes the performance problems. I change the summary accordingly. Thanks for the example.
Ok, you replied just in time; the data that I have are completely uncorrelated between them, they represents the location of particles and a correlation between the points doesn't exist. Could it be that it take so long because the program keeps trying to find a good usable function? Maybe it's worth to mention that I didn't select any correlation between point, not even a single line to connect them. Thank you again for you work.
The problem is that 35 000 shapes is something the current drawing engine cannot handle quickly. But there are major reworks ongoing in that area which will went in probably for 3.1. It is not clear to me how the rework will affect the speed in this case, I will check that then.
Ok, the major rework in the drawinglayer now has been made and integrated for OOo 3.1. But the speed has not increased. In contrary - see issue 98907. Besides that there are a couple of issues now already that aim to address specific performance problems with charts. I think Issue 97749 is about a feature that will be necessary to implement to speed up the attached example significantly. The idea is to have only one object with the formatting properties (color, linestyle etc. ) and some e.g. 35000 transformations to place it on different positions, instead of creating 35 000 full blown objects. As that feature issue is already linked from other issues I exceptionally set this older issue to duplicate to the newer one. Your issue will be linked from that issue then so your example document will still be found and can be used for further analysis. Thank you for reporting this problem! *** This issue has been marked as a duplicate of 97749 ***
closing as duplicate now