Apache OpenOffice (AOO) Bugzilla – Issue 67350
errors in grid line and y-label positioning
Last modified: 2013-02-24 21:19:12 UTC
I am seeing what looks like a related issue. overlaying two sets of data with identical x and y ranges produces axes that do not overlay perfectly as would be expected. both sets of data have y range 0.000 - 7.500 creating a chart for each of the data sets and overlaying them reveals a bug in the placing the grid lines in y: only the second of every three lines matches, the other two have a small offset. in fact it is impossible to overlay the base lines precisely. Even entering the same numerical value for y posistion produces a one pixel difference in the baseline positions. Using cm or inch units gives same result. There is also a variable error in placing y labels on the grid. In the above, 0.000 and 5.000 have a 2 px offset , 1.000 and 6.000 seem to be the only perfect overlay. In short it is impossible to overlay data for direct comparison. This is a severe limit to the usefulness of Calc. Seen using 2.02 and 2.0.3 milestone on Gentoo linux. Thx.
the test data is an engine cam profile (crudely a gaussian hump in form). I am comparing two nearly identical profiles, one peaks at 7.24 the other at 7.15 . This means the automatic axis creation produces a display of 0 - 7.5 for both data sets. It is these grids that are not coming out the same although one would expect them to be indentical.
related defect: both charts have identical y titles: "lift / mm" when overlayed these are offset by the width of one letter such that the "l" of one lines up with the "i" of the other. this error is therefore several pixels and a lot larger then the grid errors.
Created attachment 37756 [details] screenshot of defective alignment issues.
Final note , the offset of the y axes was intentional to align the data, this is not part of the issues I am reporting here. Comments on the misalignment of the Y labels refers only to the vertical positioning which , as noted in the original post, is variable. hope that all gives you enough info to pinpoint the effect. regards.
@OOOer: Pls attach the calc doc. Thx
Created attachment 37760 [details] ods for above issue, see sheet "overlay" for charts
PS. now you have the source data you will find a third chart (displayed as descrete data points) behind the two visible ones. This also should overlay and does not. If you bring this one to the front it will overlay the other two and reveals a further issue matching the scaling in x. I mention this in passing incase you want to look at it at the same time since it may all be part of the same problem and may provide you with some more input on the defect, otherwise I will open a separte bug report. pls post if you req more details on the 3rd chart issue.
It is indeed strange, that creating two charts in the same way does not yield the same layout. However, there are always rounding errors when displaying floating point numbers on a display using pixels. So, both charts in the overlay are probably correct solutions to the rounding of values to pixels. I don't know why they do it different. Another important thing is that Calc uses Pixels for rendering, while the chart uses 100th of mm internally. Therefore, there are additional rounding problems. And finally, if you edit the chart, you see the output rendered by the drawing layer, whereas when the chart is deactivated, you see a replacement image (metafile) which might also differ in details. I see those differences as not important, as the data is not forged, i.e. not correct. Why don't you create a chart that renders the differences between your two plots? This way, you wouldn't have to search for differing pixels (with a magnifying glass) but could look at differences at any scale you like, usually a finer one.
thanks for your reply. >>I see those differences as not important, as the data is not forged, i.e. not correct. I see no way of knowing if the data displays the same errors but I see no good reason for assuming it is not distorted in the same way as the axes. >>Why don't you create a chart that renders the differences between your two plots? Because that was not the point. I have determined what I needed from this data. The reason I created a bug report was that there was a rather messy error that will show up on ANY overlay of two data sets. Not only does this make Calc looks second rate but prevents any serious scientific use in this way. Even a business presentation would be unacceptable with misaligned axis and blurred axes labels. I was not looking for a work-around. I assumed that would be an error that O.O.org would wish to correct. If not , sorry for wasting your time. You may (or may not) also wish to look at the third chart I mentioned above that is also in that data. I was unable to scale the axes to be anything close to the other two. There were several mm of off-set. Many more alignment issues come up if you add the legend to the charts. This obviously needs a fair bit of work still. Since it is apparently getting active attention at the moment I assumed your would appreciated some detailed feedback on these issues. thanks again. PS you may wish to change the status to confirmed since you have reproduced the results on the supplies data. bye.
Changed Prio. and Type.
Looking at your document you have different sizes and positions for the charts: Yellow chart frame pos: 1,02cm 10,21cm size: 25,40cm 13,49cm White chart frame pos: 1,06cm 10,22cm size: 25,40cm 13,49cm Yellow diagram pos: 1,56cm 1,51cm size: 20,12cm 11,70cm White diagram pos: 1,56cm 1,51cm size: 20,11cm 11,71cm Thus it is no suprise that you cannot overlay them nicely. I'll attach an corrected document.
Created attachment 40645 [details] corrected positions and sizes
Created attachment 40646 [details] screenshot of the correction
You can use the toolbar button 'Reorganize Chart' to bring the chart elements to their default positions and sizes. That is easier than opening and editing each Position and Size dialog for all titles, legend and diagram. You other problem is simply that the second chart has no transparent background. Edit your RH intel chart in this way: Choose main menu -> Format -> Chart Area. Click to the Area tab page and select 'None' in the listbox for the filling. That should help.
closed as worksforme