Area series is not rendered correctly in complex chart

Hi BIRT-experts.

I have put together with a fairly complex charts with loads of different data series that all give different values over time.

I am plotting CPU usage as one data series and I use an area series for it. Also the issue exists with and without aggregation. As the CPU usage data series has only very few data points the area does not show. The data series is plotted as if all points on the X-axis where there is no value for CPU usage assume value 0. This is not only not good looking, but is also very confusing as it suggests that the machine has idle between the actual values that appear as peaks now. I attach a picture as an example.

Hope you can help.

Kind regards
Frank

Find more posts tagged with

Comments

mwilliams

Hi Frank,

Is there any way you can create a similare report with some sample data in a csv file or something, so that I can run the report? Are there no values for the times in between those spikes?

fbrueseke

I have managed to assemble a simple example. You must adjust the path to the csv file I included in the zip-file when trying to run this example.

Hope you find something out.

Kind regards
Frank

mwilliams

Frank,

I would say that the area chart is working correctly. Are you only reporting the cpu usage when there's a change in usage, I take it? You'd need to have a value for each value on the x-axis, whether you gather that for your database or create a computed column to do so in BIRT, otherwise, BIRT has nothing else to assume but 0. Could you post a small sample of your actual data? Or data like your data with changed values if you can't post actual data in here? So I can see what you're working with. Thanks.

fbrueseke

Hi.

In my real reports I have a variety of data series: CPU usage for user, CPU usage for system, response time, GC, ...

So what I'm trying to do is to draw each of these data series over a time scale as if they were independent. They all stem from the same monitoring system, but each piece of data is timestamped independently. If I wanted consistent data series that have values for the same time stamps I would have to do a great deal of aggregation.

If I have two data series I want BIRT to interpret it independently. I give you a constructed example. I will show two series one showing CPU usage over time and the other memory over time.

Current interpretation:
Y: 20% ___ 25% ___
X: 01s 02s 03s 04s

Y: ___ 57M ___ 59M
X: 01s 02s 03s 04s

Interpretation that I want:
Y: 20% 25%
X: 01s 03s

Y: 57M 59M
X: 02s 04s

So what should the consequences be? BIRT should not assume that there must be a y-value for any X-values given. It should assume that there is a valid data point if in the data set there is a non-null Y-value for an x-value. So, if there is a null Y-value, don't draw a point.

Hope that helps you to understand what I'm up to.

Kind regards
Frank

fbrueseke

As I worked with BIRT today I found out that this problem seems to be exclusive for area series. I suspected this from the beginning, but was not entirely sure.

Also have a look at my updated example. It shows the same diagram as area series and line series.

Kind regards
Frank

mwilliams

Frank, 
 
I see exactly what you're saying now. Please log a bug for this at <a class='bbc_url' href='http://www.eclipse.org/birt/phoenix/reportabug.php'>BIRT: Reporting Bugs and Requesting Enhancements</a>. Be sure to include this example you attached to show the issue in the bug report. Also, if you could post any bug information in this thread for future reference, that'd be great. Thanks.

fbrueseke

Hi, 
 
I filed a bug in: 
<a class='bbc_url' href='https://bugs.eclipse.org/bugs/show_bug.cgi?id=287653'>https://bugs.eclipse.org/bugs/show_bug.cgi?id=287653</a>

mwilliams

Thanks, Frank!