Re: [Geotools-gt2-users] MouseListener, Multiple Coordinate Systems...

[Martin D. <mar...@te...>]

Casson Stallings a =E9crit :

> Map data does not have a defined coordinate system until it is added to=
=20
> a mapContext with a defined coordinate system. I could find no way to=20
> define the coordinate system at the DataStore or FeatureCollection=20
> level. (After some more reading I did find the SLDRenderedGeometries,=20
> but its not clear to me if this is its role.)

Right, for now SLDRenderedGeometries could be used as a workaround, but=20
it should not be its role. I think that DataStore should be the right=20
place for defining a coordinate system. They were a discussion on this=20
topic a while ago. For now DataStore just ignore CS, but it shouldn't. I=20
don't remember who was in charge of introducing CS concept in DataStore.=20
Andrea may know more.



> When you pull a renderer off of the StyledMapPane with a defined=20
> coordinate system, the renderer does not keep the knowledge of the=20
> coordinate system. It must be assigned separately if the renderer needs=
=20
> to know it. In the output see "Renderer Coorsys 1" and "Renderer Coorsy=
s=20
> 2". The first is pulled off of the renderer before the separate=20
> assigning of the coordinate system.

It sound like a bug. We should fill a JIRA task for it.



> NEW QUESTIONS
> =20
> Should the renderer maintain the coordinate system of the MapPane it is=
=20
> pulled from?

Yes.


> At least in my tests (as an applet under eclipse)  the coordinates=20
> returned by MouseEvent.getMapCoordinates were always initially the pixe=
l=20
> coordinates. See output log below. The mapPane knew the appropriate=20
> boundaries in real world coordinates. It was not until I resized the=20
> applet window or zoomed in on the data that getMapCoordinates started=20
> returning the correct coordinates. I tried doing .resets .validates,=20
> and  resetting the bounds of the data in the program, but these did not=
=20
> seem to help. Any idea what is going on?

Sound like a bug again... It is somewhat strange, since I didn't=20
experimented this behavior in my own application ('getMapCoordinate' was=20
working properly right from the first draw). In may be worth to monitor=20
the value of the 'zoom' AffineTransform field in ZoomPane. I suspect=20
that it is left to the identity transform when is should have been some=20
more complex transform (but in this case, I find it strange that the=20
rendering is done properly). If the ZoomPane.zoom value is okay (not the=20
identity transform), then a second possibility is to check if its value=20
is consistent with Renderer.mapToText (an other AffineTransform, but=20
should have the same value than zoom after a rendering as been=20
completed). If the values are consistent as they should be, then we will=20
advise (there is some other possibilities).



> Clearly there is something I am missing. I do not see how you can set u=
p=20
> a mappane to display multiple set of data stored in different coordinat=
e=20
> systems if you cannot define separate coordinate systems or transforms=20
> for them. How is this done unless you use SLDRenderedGeometries to=20
> create RenderedLayers? However, the documentation on this class does no=
t=20
> seem to indicate it is for general use. Maybe I'll have to test it.

Setting of CS should be done at DataStore or Feature level. If there is=20
no method for that in the current API, this is because the CS and=20
Geotools framework in Geotools were developped by different peoples, and=20
the Feature / DataStore framework has not yet taken CS in account. But=20
they should. An other issue is that the CRS framework is under=20
refactoring right now, in order to mirror the change in OpenGIS=20
specification. We are in process of switching to the interfaces defined=20
in the GeoAPI process.

	Regards,

		Martin.