GeoTools, the Java GIS toolkit

Vincenzo Marchese wrote:

> Hi all,
> I am using a StyleMapPane with layers built from shapefiles.
> I wanted to display  tif images on the background an I succeded to do 
> this using a GridCoverage with the BufferedImages read from disk
> The image is added to the Panel with
>
> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); // oGC 
> is the GridCoverage Object
>
> Now I use this code to remove and add layers:
>
>            Rectangle2D oArea = oMapPane.getVisibleArea(); // get 
> visible area
>            
> context.removeLayer(oLayer);                                  // 
> remove Maplayer from DefaultMapContext (or addLayer...)
>            oMapPane.setMapContext(context);                        // 
> reset map context to SyledMapPane
>            oMapPane.setVisibleArea(oArea);                           
> // reset visible area to prevent zoom reset
>
> As soon as I remove the first layer my background images disappear.
> So my questions are :
> Is it possible to keep the image from disappearing in the renderer ?
> Is there a way to act on RenderedLayers instead of MapLayers when 
> showing and adding Layers ?

Vincenzo, I coded the StyledMapPane over one year ago in an effort to 
make the j2d renderer usable in
Geotools (it was not developed in the gt2 source base, but outside it 
and added only later), but frankly
I have not used it much besides some simple teaching example so I'm not 
surprised it does not behave
as expected.
Your best plan of action is, in my opinion, to get the StyledMapPane 
source code understand it, and fix it.
By the way, to remove a layer it should be not necessary to re-set the 
map context into the map pane,
but only to remove the layer from the MpaContext itself (an event should 
be triggered informing the StyledMapPane
of the change and it should remove the correspondent rendered layer by 
himself).

Best regards
Andrea Aime

Andrea,

I'm sending this directly to you because my mail to the list
keeps bouncing for unexplained reasons.

On Wed, 2005-02-16 at 06:27, Vincenzo Marchese wrote:
> Hi all,
> I am using a StyleMapPane with layers built from shapefiles.
> I wanted to display  tif images on the background an I succeded to do 
> this using a GridCoverage with the BufferedImages read from disk
> The image is added to the Panel with
> 
> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); // oGC 
> is the GridCoverage Object
> 
> Now I use this code to remove and add layers:
> 
>             Rectangle2D oArea = oMapPane.getVisibleArea(); // get 
> visible area
>             
> context.removeLayer(oLayer);                                  // remove 
> Maplayer from DefaultMapContext (or addLayer...)
>             oMapPane.setMapContext(context);                        // 
> reset map context to SyledMapPane
>             oMapPane.setVisibleArea(oArea);                           // 
> reset visible area to prevent zoom reset
> 
> As soon as I remove the first layer my background images disappear.
> So my questions are :
> Is it possible to keep the image from disappearing in the renderer ?
> Is there a way to act on RenderedLayers instead of MapLayers when 
> showing and adding Layers ?

Here are some code snippets I've used:

// add a grid coverage layer.  info is a container object I made
RenderedGridCoverage rendered = new
    RenderedGridCoverage(info.getGridCoverage());
rendered.setZOrder(-1);
coverageLayers.add(rendered);
mapPane.getRenderer().addLayer(rendered);

// remove all layers
for (Iterator iter = coverageLayers.iterator(); iter.hasNext();) {
    RenderedLayer element = (RenderedLayer) iter.next();
    mapPane.getRenderer().removeLayer(element);
    element.dispose();
    iter.remove();
}

I was working with aerial photos (several files of 57MB), and found
that if I didn't remove them in the above manner, the program would
run out of memory.  With the aid of a profiler and the source code,
I learned that GridCoverage->PlanarImage pairs are cached in a static
WeakHashMap.  Even though WeakHashMap entries should be garbage
collected when they are no longer referenced by anything else, that
didn't seem to be happening for some reason.  Also, it ran out of
memory faster than you might think because when a RenderedGridCoverage
gets projected to your map pane's coordinate system, both the
old and new images are cached.  (Particularly annoying in my case,
because the map pane was in PROMISCUOUS mode.)

I also was not able to get the coverages (which were in WGS84)
to project to the shapefiles' coordinate system (NAD_1983_HARN).
I believe this problem will be gone after the current coordinate
system refactor.  Am I right?

A developer who's connected with RenderedGridCoverage might want to
reexamine the pros and cons of caching in that class.  For users,
be wary when using RenderedGridCoverage with large data.

--
Trevor Stone ~ <tstone@...>
Software Developer ~ Tyler Technologies, Eagle Division
Denver Office ~ 1-866-554-4434 / 303-271-9100 ext 208
Curse of the moment: Thou lumpish bunched-backed manikin!
Sometimes a cigar is just a cigar. -- Freud

Trevor Stone wrote:

>Here are some code snippets I've used:
>
>// add a grid coverage layer.  info is a container object I made
>RenderedGridCoverage rendered = new
>    RenderedGridCoverage(info.getGridCoverage());
>rendered.setZOrder(-1);
>coverageLayers.add(rendered);
>mapPane.getRenderer().addLayer(rendered);
>
>// remove all layers
>for (Iterator iter = coverageLayers.iterator(); iter.hasNext();) {
>    RenderedLayer element = (RenderedLayer) iter.next();
>    mapPane.getRenderer().removeLayer(element);
>    element.dispose();
>    iter.remove();
>}
>
>I was working with aerial photos (several files of 57MB), and found
>that if I didn't remove them in the above manner, the program would
>run out of memory.  With the aid of a profiler and the source code,
>I learned that GridCoverage->PlanarImage pairs are cached in a static
>WeakHashMap.  Even though WeakHashMap entries should be garbage
>collected when they are no longer referenced by anything else, that
>didn't seem to be happening for some reason.  Also, it ran out of
>memory faster than you might think because when a RenderedGridCoverage
>gets projected to your map pane's coordinate system, both the
>old and new images are cached.  (Particularly annoying in my case,
>because the map pane was in PROMISCUOUS mode.)
>
>I also was not able to get the coverages (which were in WGS84)
>to project to the shapefiles' coordinate system (NAD_1983_HARN).
>I believe this problem will be gone after the current coordinate
>system refactor.  Am I right?
>
>A developer who's connected with RenderedGridCoverage might want to
>reexamine the pros and cons of caching in that class.  For users,
>be wary when using RenderedGridCoverage with large data.
>
>  
>

Trevor,
unfortunately due to severe lack of spare time I'm no longer an active 
developer, so
you may want to create a JIRA issue that will be hopefully managed by 
Martin Desriusseaux...
I think he's the only developer currently interested in mantaining the 
j2d renderer, since the
Refractions guys have preferred to use and mantain my simpler lite 
renderer due to better
scalability...

Best regards
Andrea Aime

Andrea Aime a =E9crit :

> unfortunately due to severe lack of spare time I'm no longer an active=20
> developer, so
> you may want to create a JIRA issue that will be hopefully managed by=20
> Martin Desriusseaux...
> I think he's the only developer currently interested in mantaining the=20
> j2d renderer, since the
> Refractions guys have preferred to use and mantain my simpler lite=20
> renderer due to better
> scalability...


Of course I'm defending my baby, so I'm not impartial here. But we are=20
obviously in disagreement. Unfortunatly, I have been stuck with my Ph.D.=20
thesis much longer than I initially though (at least one year longer),=20
and I'm very sorry to have been unable to done any serious J2D work for=20
that long.

That said, I strongly believe that J2D renderer has way much more=20
potential than LiteRenderer, at the cost of its complexity. But J2S=20
renderer is complex because the problem *IS* complex. LiteRenderer is=20
doing some false assumptions regarding AffineTransform and projections=20
among other (I told that to Refractions Research guys in last OGC meeting=
).

I have been back as an active developper only since the begining of=20
February, as you can see from my commit. I must finish CRS first, then=20
work on GridCoverage, and J2D renderer will be next.

This is true that LiteRenderer is much mpre Geotools aware than J2D=20
renderer at this time. But the J2D architecture is closer than=20
LiteRenderer to GO-1, which is itself close in spirit to commercial=20
packages like ESRI in my understanding. J2D renderer will be ported to=20
GO-1; I'm not sure that LiteRenderer can. Please remember that I'm=20
strongly involved in GeoAPI development, up to the point that I have=20
sent feedback to some working groups, and some of my suggestions have=20
been accepted by OGC for the next ISO 19111 revision. I'm not putting=20
all those energy in standards (ISO 19111, ISO 19107, ISO 19115) for=20
ignoring them!!!!

So yes LiteRenderer is much more standard-aware than J2D renderer for=20
now, but expect the situation to be inversed before the end of this year.

As for performance and scalability, I swear you that J2D renderer will=20
blast LiteRenderer in performance (except at startup time) with=20
comparable scalability and better standard-compliance. But there is=20
months of work ahead before I reach this point.

In any case, the API will be significantly different after I make the=20
switch to GO-1. So no matter which renderer you use (LiteRenderer or J2D=20
renderer), I'm afraid that some significant refactoring will be hard to=20
avoid. On the bright side, the refactoring should be toward standards.

     Martin.

Vincenzo Marchese a =E9crit :
> I am using a StyleMapPane with layers built from shapefiles.
> I wanted to display  tif images on the background an I succeded to do=20
> this using a GridCoverage with the BufferedImages read from disk
> The image is added to the Panel with
>=20
> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); // oGC =
is the GridCoverage Object
>=20
> Now I use this code to remove and add layers:
>=20
>     Rectangle2D oArea =3D oMapPane.getVisibleArea(); // get visible are=
a
>     context.removeLayer(oLayer);        // remove Maplayer from Default=
MapContext (or addLayer...)
>     oMapPane.setMapContext(context);    // reset map context to SyledMa=
pPane
>     oMapPane.setVisibleArea(oArea);     // reset visible area to preven=
t zoom reset
>=20
> As soon as I remove the first layer my background images disappear.
> So my questions are :
> Is it possible to keep the image from disappearing in the renderer ?
> Is there a way to act on RenderedLayers instead of MapLayers when=20
> showing and adding Layers ?

could you try the following before and after the layer removal?

    System.out.println(oMapPane.getRenderer());

It should display the list of RenderedLayers in the map pane, so we can=20
check if the RenderedGridCoverage layer stay there.

	Martin.

Trevor Stone a =E9crit :
> I was working with aerial photos (several files of 57MB), and found
> that if I didn't remove them in the above manner, the program would
> run out of memory.  With the aid of a profiler and the source code,
> I learned that GridCoverage->PlanarImage pairs are cached in a static
> WeakHashMap.  Even though WeakHashMap entries should be garbage
> collected when they are no longer referenced by anything else, that
> didn't seem to be happening for some reason.

I noticed that with J2SE 1.4, the incremential garbage collector=20
(-Xincgc) has more trouble with weak references than the default garbage=20
collector. I got OutOfMemoryError too, and fixed it just by using the=20
default garbage collector instead of the incremential one. I'm not sure=20
if it will be of any help in yours case however.




   Also, it ran out of
> memory faster than you might think because when a RenderedGridCoverage
> gets projected to your map pane's coordinate system, both the
> old and new images are cached.  (Particularly annoying in my case,
> because the map pane was in PROMISCUOUS mode.)

This sound like a bug. It would be worth to open a JIRA task, so I can=20
remember that when I will restart the work on J2D renderer.



> I also was not able to get the coverages (which were in WGS84)
> to project to the shapefiles' coordinate system (NAD_1983_HARN).
> I believe this problem will be gone after the current coordinate
> system refactor.  Am I right?

There is some chances. The new CRS framework has some more capability=20
than the old one. It should also be a little bit more rigorous.


> A developer who's connected with RenderedGridCoverage might want to
> reexamine the pros and cons of caching in that class.  For users,
> be wary when using RenderedGridCoverage with large data.

I will try to keep that in mind when I will do the port of GridCoverage=20
to GeoAPI and J2D-Renderer to GO-1. A Jira task may help just to make=20
sure that I remember. Thanks for the report,

	Martin.

Andrea Aime wrote:

> Vincenzo Marchese wrote:
>
>> Hi all,
>> I am using a StyleMapPane with layers built from shapefiles.
>> I wanted to display  tif images on the background an I succeded to do 
>> this using a GridCoverage with the BufferedImages read from disk
>> The image is added to the Panel with
>>
>> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); // 
>> oGC is the GridCoverage Object
>>
>> Now I use this code to remove and add layers:
>>
>>            Rectangle2D oArea = oMapPane.getVisibleArea(); // get 
>> visible area
>>            
>> context.removeLayer(oLayer);                                  // 
>> remove Maplayer from DefaultMapContext (or addLayer...)
>>            oMapPane.setMapContext(context);                        // 
>> reset map context to SyledMapPane
>>            oMapPane.setVisibleArea(oArea);                           
>> // reset visible area to prevent zoom reset
>>
>> As soon as I remove the first layer my background images disappear.
>> So my questions are :
>> Is it possible to keep the image from disappearing in the renderer ?
>> Is there a way to act on RenderedLayers instead of MapLayers when 
>> showing and adding Layers ?
>
>
> Vincenzo, I coded the StyledMapPane over one year ago in an effort to 
> make the j2d renderer usable in
> Geotools (it was not developed in the gt2 source base, but outside it 
> and added only later), but frankly
> I have not used it much besides some simple teaching example so I'm 
> not surprised it does not behave
> as expected.
> Your best plan of action is, in my opinion, to get the StyledMapPane 
> source code understand it, and fix it.
> By the way, to remove a layer it should be not necessary to re-set the 
> map context into the map pane,
> but only to remove the layer from the MpaContext itself (an event 
> should be triggered informing the StyledMapPane
> of the change and it should remove the correspondent rendered layer by 
> himself).
>
> Best regards
> Andrea Aime
>
Andrea,
thanks for your response.
I managed to get the correct behaviour with:
             StyledMapRenderer oSMR = 
(StyledMapRenderer)oMapPane.getRenderer();
             oSMR.removeLayer(oLayer);
the removeLayer method behaves the proper way but this works only with 
the casting to StyledMapRenderer
since there is no method removeLayer(MapLayer) in the Renderer class but 
only removeLayer(RenderedLayer)
Btw a simple
MapContext.removeLayer(...) doesn't work and moreover I didn't find any 
removeLayer(MapLayer) (that I need for the shapefile layers)
in it. Is there a way to fetch the RenderedLayers from the MapLayers 
added to a context ?

Best Regards
Vincenzo

-- 
   Sometimes it seems things go by too quickly. We are so busy watching out for
what's just ahead of us that we don't take the time to enjoy where we are.
		  -- Calvin
-----------------------------------------------------
      Vincenzo  Marchese
        ARSRETIA S.r.l.
      Via D. Sansotta, 97
        00144 Roma (IT)
e-mail: v.marchese@...
     Tel.: +39 06 52270097
     Fax: +39 06 52272313

Martin Desruisseaux wrote:

> Vincenzo Marchese a =E9crit :
>
>> I am using a StyleMapPane with layers built from shapefiles.
>> I wanted to display  tif images on the background an I succeded to do=20
>> this using a GridCoverage with the BufferedImages read from disk
>> The image is added to the Panel with
>>
>> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); //=20
>> oGC is the GridCoverage Object
>>
>> Now I use this code to remove and add layers:
>>
>>     Rectangle2D oArea =3D oMapPane.getVisibleArea(); // get visible ar=
ea
>>     context.removeLayer(oLayer);        // remove Maplayer from=20
>> DefaultMapContext (or addLayer...)
>>     oMapPane.setMapContext(context);    // reset map context to=20
>> SyledMapPane
>>     oMapPane.setVisibleArea(oArea);     // reset visible area to=20
>> prevent zoom reset
>>
>> As soon as I remove the first layer my background images disappear.
>> So my questions are :
>> Is it possible to keep the image from disappearing in the renderer ?
>> Is there a way to act on RenderedLayers instead of MapLayers when=20
>> showing and adding Layers ?
>
>
> could you try the following before and after the layer removal?
>
>    System.out.println(oMapPane.getRenderer());
>
> It should display the list of RenderedLayers in the map pane, so we=20
> can check if the RenderedGridCoverage layer stay there.
>
>     Martin.
>
>
> -------------------------------------------------------
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from real users=
.
> Discover which products truly live up to the hype. Start reading now.
> http://ads.osdn.com/?ad_ide95&alloc_id=14396&op=3Dclick
> _______________________________________________
> Geotools-gt2-users mailing list
> Geotools-gt2-users@...
> https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users

Martin,
thanks for your response.
I do a dump of the layers collection each time I hide/show a layer.
The fact is that if I remove a Layer from the context I lose the=20
RenderedLayer of the raster images (the do not appear any more in the=20
collection).
I didn't have a look at the code but i think that this is because I=20
didn't add the RenderedGridCoverage to the context but only to the=20
renderer of the MapPane.
What I do not understand is the fact that the layer is lost, in other=20
words the situation is:
- I add the RenderedGridCoverage to the renderer and it appears in the=20
renderer dump
- I remove another layer from the context (a SLDRenderedGeometries layer=20
created from a shapefile) and the raster RenderedGridCoverage layer is lo=
st

As I wrote to Andrea I managed to get the proper behaviour with
            StyledMapRenderer oSMR =3D=20
(StyledMapRenderer)oMapPane.getRenderer();
            oSMR.removeLayer(oLayer);
(sorry to repeat, you can find my post in the thread)
but only because the StyledMapRenderer has a method=20
removeLayer(RenderedLayer)
Best Regards
Vincenzo


--=20
I am a conscientious man, when I throw rocks at seabirds I leave no tern
unstoned.
		-- Ogden Nash, "Everybody's Mind to Me a Kingdom Is"
-----------------------------------------------------
      Vincenzo  Marchese
        ARSRETIA S.r.l.
      Via D. Sansotta, 97
        00144 Roma (IT)
e-mail: v.marchese@...
     Tel.: +39 06 52270097
     Fax: +39 06 52272313

Vincenzo Marchese wrote:

> Martin Desruisseaux wrote:
>
>> Vincenzo Marchese a =E9crit :
>>
>>> I am using a StyleMapPane with layers built from shapefiles.
>>> I wanted to display  tif images on the background an I succeded to=20
>>> do this using a GridCoverage with the BufferedImages read from disk
>>> The image is added to the Panel with
>>>
>>> oMapPane.getRenderer().addLayer(new RenderedGridCoverage(oGC)); //=20
>>> oGC is the GridCoverage Object
>>>
>>> Now I use this code to remove and add layers:
>>>
>>>     Rectangle2D oArea =3D oMapPane.getVisibleArea(); // get visible a=
rea
>>>     context.removeLayer(oLayer);        // remove Maplayer from=20
>>> DefaultMapContext (or addLayer...)
>>>     oMapPane.setMapContext(context);    // reset map context to=20
>>> SyledMapPane
>>>     oMapPane.setVisibleArea(oArea);     // reset visible area to=20
>>> prevent zoom reset
>>>
>>> As soon as I remove the first layer my background images disappear.
>>> So my questions are :
>>> Is it possible to keep the image from disappearing in the renderer ?
>>> Is there a way to act on RenderedLayers instead of MapLayers when=20
>>> showing and adding Layers ?
>>
>>
>>
>> could you try the following before and after the layer removal?
>>
>>    System.out.println(oMapPane.getRenderer());
>>
>> It should display the list of RenderedLayers in the map pane, so we=20
>> can check if the RenderedGridCoverage layer stay there.
>>
>>     Martin.
>>
>>
>> -------------------------------------------------------
>> SF email is sponsored by - The IT Product Guide
>> Read honest & candid reviews on hundreds of IT Products from real user=
s.
>> Discover which products truly live up to the hype. Start reading now.
>> http://ads.osdn.com/?ad_ide95&alloc_id=14396&op=3Dclick
>> _______________________________________________
>> Geotools-gt2-users mailing list
>> Geotools-gt2-users@...
>> https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
>
>
> Martin,
> thanks for your response.
> I do a dump of the layers collection each time I hide/show a layer.
> The fact is that if I remove a Layer from the context I lose the=20
> RenderedLayer of the raster images (the do not appear any more in the=20
> collection).
> I didn't have a look at the code but i think that this is because I=20
> didn't add the RenderedGridCoverage to the context but only to the=20
> renderer of the MapPane.
> What I do not understand is the fact that the layer is lost, in other=20
> words the situation is:
> - I add the RenderedGridCoverage to the renderer and it appears in the=20
> renderer dump
> - I remove another layer from the context (a SLDRenderedGeometries=20
> layer created from a shapefile) and the raster RenderedGridCoverage=20
> layer is lost
>
> As I wrote to Andrea I managed to get the proper behaviour with
>            StyledMapRenderer oSMR =3D=20
> (StyledMapRenderer)oMapPane.getRenderer();
>            oSMR.removeLayer(oLayer);
> (sorry to repeat, you can find my post in the thread)
> but only because the StyledMapRenderer has a method=20
> removeLayer(RenderedLayer)


Of course I meant removeLayer(MapLayer)
sorry for the mistype

> Best Regards
> Vincenzo
>
>


--=20
   Sometimes it seems things go by too quickly. We are so busy watching o=
ut for
what's just ahead of us that we don't take the time to enjoy where we are=
.
		  -- Calvin
-----------------------------------------------------
      Vincenzo  Marchese
        ARSRETIA S.r.l.
      Via D. Sansotta, 97
        00144 Roma (IT)
e-mail: v.marchese@...
     Tel.: +39 06 52270097
     Fax: +39 06 52272313