Multiply RangeAxis with one dataset

[FPawlak]

Hi,
is it possible to assign multiply range axis to one dataset? I know that it is possible with to assign different range axis to different datasets. But i don't want to double data (domain axis).
Let's say, I have to plot weather conditions - wind and temperature - each datapoints probed simultaneously, so there are no need to use more than one dataset.

Do you have some suggestions?

[david.gilbert]

No, the each dataset is matched to a single axis, so if you want to use multiple axes you'll have to have multiple datasets.

[FPawlak]

Thanks for replay.
I've already looked through the source code and I have some idea how to solve my problem implementing this feature. So I will later post here my solution. I think it may be useful to someone else.

I'm going to extends XYPlot class and override draw() method changing two for loops which are responsible for rendering data - I plan to query dataset for each series to obtain correct axis. So I need also add getRangeAxis(int series) or getRangeAxisId(int series) to my dataset.

[david.gilbert]

That will most likely work, but I would try to avoid mixing presentation-related attributes (the axis index to use etc) with your data model (the dataset).

[paradoxoff]

But i don't want to double data (domain axis).
Let's say, I have to plot weather conditions - wind and temperature - each datapoints probed simultaneously, so there are no need to use more than one dataset.

But you will need two series anyways, one for the wind, the second for the temperature.
You can implement your requirement by using two datasets, each with a single series, and two renderers. I do not believe that this is more complicated to set up than using only one dataset with two series, a single renderer, and implementing your XYPlot extension.
If you have to plot a large amount of data, and if you want so save the memory for the second set of x-coordinates (which is identical to the first set), your approach has the potential to save some memory (dxepending on the type of dataset you are using), but even in that case, I would rather implement an XYDataset where multiple datasets can share the same set of x values.

[FPawlak]

That will most likely work, but I would try to avoid mixing presentation-related attributes (the axis index to use etc) with your data model (the dataset).

This was my first solution in order to see if it will work. And yes, it works. So to make to solution better and more clear i will add to dataset only getSeriesCategory(int series) and create class AxesManager which will map series and axes. So if XYPlot has reference to AxesManager it will during plotting for each series 1) obtain series category from dataset 2) obtain correspondent Axis from AxesManager 3) plot series using that axis.

But you will need two series anyways, one for the wind, the second for the temperature.
You can implement your requirement by using two datasets, each with a single series, and two renderers. I do not believe that this is more complicated to set up than using only one dataset with two series, a single renderer, and implementing your XYPlot extension.
If you have to plot a large amount of data, and if you want so save the memory for the second set of x-coordinates (which is identical to the first set), your approach has the potential to save some memory (dxepending on the type of dataset you are using), but even in that case, I would rather implement an XYDataset where multiple datasets can share the same set of x values.

Yes, I need more than one series, so my dataset has one "x series" common for every "y series". I think it will be easier to manage only one dataset - I need to load new while panning, add/remove series while reviewing chart. And of course I will have more then two series - five or six axes/series category and ca. 40 series

[david.gilbert]

OK, sounds like you've thought it through and you are on a good track.

[olbrich]

Hi David,

this thread describes a very common feature.
Why not adress this in the next version?

Markus

[david.gilbert]

Because it would take a lot of time (which I don't have), it would break the API, and the existing mechanism (multiple datasets and renderers, with datasets being mapped to arbitrary axes) provides a perfectly workable solution.