Canvas dispose

[Praveen Nayak2]

Hi,

I use JSVGCanvas in an applet to show SVG diagrams. When the page is navigating away from the current page, I disposed the canvas in the applet stop method. I earlier had:
canvas.dispose();

I have many pages each with an applet in it, with the SVG diagram, and users can switch between these pages, sometimes quite rapidly. I noticed there was memory leaking away. I then included a call to stopProcessing, which improved things. So I now have
canvas.dispose();
canvas.stopProcessing();

1. Is this sufficient?

Also, at times during navigation, I see errors such as these logged in the console (I believe if I navigate away before the SVG is rendered):
java.lang.NullPointerException
        at org.apache.batik.dom.events.EventListenerList.removeListener(Unknown Source)
        at org.apache.batik.dom.events.EventSupport.removeEventListenerNS(Unknown Source)
        at org.apache.batik.dom.AbstractNode.removeEventListenerNS(Unknown Source)
        at org.apache.batik.bridge.FocusManager.removeEventListeners(Unknown Source)
        at org.apache.batik.bridge.FocusManager.dispose(Unknown Source)
...
(or a null pointer exception at setGraphicsNode) Can these add to a leak, and can I handle these somehow?

Thanks,
Praveen

[thomas.deweese]

Hi Praveen,

Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 02:46:10 AM:

> I have many pages each with an applet in it, with the SVG diagram,
> and users can switch between these pages, sometimes quite rapidly. I
> noticed there was memory leaking away. I then included a call to
> stopProcessing, which improved things. So I now have
> canvas.dispose();
> canvas.stopProcessing();

    I don't think this is a good sequence.  I would stop at
canvas.dispose().  The call to 'stopProcessing()' is actually
likely acting to prevent the action of canvas.dispose().

    I've seen Swing hold onto the canvas much longer than
desirable because it was the 'focus' component when it was
not replaced by a new focus component.  So you might try
removing the canvas from the applet and adding a JButton in
it's place.

> 1. Is this sufficient?
>
> Also, at times during navigation, I see errors such as these logged
> in the console (I believe if I navigate away before the SVG is rendered):
> java.lang.NullPointerException

    These may be caused by your calling stopProcessing.

[Praveen Nayak2]

Hi Thomas,

Sorry for the wrong ordering below, my bad. The call is actually
canvas.stopProcessing();
canvas.dispose();

I hope that order is alright. I also have code removing the canvas from the applet:
this.getContentPane().remove(canvas);
canvas = null;

Any idea on console errors in the above scenario? Appreciate your help.

Thanks,
Praveen

thomas.deweese@...

28/10/2009 04:12 PM

Please respond to
batik-users@...

To	batik-users@...
cc	batik-users@...
Subject	Re: Canvas dispose

Hi Praveen,

Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 02:46:10 AM:

> I have many pages each with an applet in it, with the SVG diagram,
> and users can switch between these pages, sometimes quite rapidly. I
> noticed there was memory leaking away. I then included a call to
> stopProcessing, which improved things. So I now have
> canvas.dispose();
> canvas.stopProcessing();

   I don't think this is a good sequence.  I would stop at
canvas.dispose().  The call to 'stopProcessing()' is actually
likely acting to prevent the action of canvas.dispose().

   I've seen Swing hold onto the canvas much longer than
desirable because it was the 'focus' component when it was
not replaced by a new focus component.  So you might try
removing the canvas from the applet and adding a JButton in
it's place.

> 1. Is this sufficient?
>
> Also, at times during navigation, I see errors such as these logged
> in the console (I believe if I navigate away before the SVG is rendered):
> java.lang.NullPointerException

   These may be caused by your calling stopProcessing.

[thomas.deweese]

Hi Praveen,

Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 07:39:39 AM:

> Sorry for the wrong ordering below, my bad. The call is actually
> canvas.stopProcessing();
> canvas.dispose();

   Internally canvas.dispose ends up calling 'stopProcessing()'
so your early call is only likely to cause problems.  I strongly
recommend that you remove it.

> I also have code removing the canvas from the applet:
> this.getContentPane().remove(canvas);
> canvas = null;

    In my experience simply removing the canvas from the
Swing hierarchy is not enough.  You need to give it something
else to 'point at'.  You can take a look at
test-sources/org/apachage/batik/swing/JSVGMemoryLeakTest.doSomething
to see the sequence I have used in the past to 'help' the canvas
go to GC.

> Any idea on console errors in the above scenario? Appreciate your help.

    I suspect the problem is the extra call to stopProcessing.  That
method is async and I think you end up calling dispose on the BridgeContext
twice in two separate threads - which isn't good.  I don't think it will
lead to memory leaks as one of the threads will likely run to completion,
and in any case I'm confident that everything will go to GC if the canvas
goes to GC (which I suspect it isn't currently).

> thomas.deweese@...
> 28/10/2009 04:12 PM
>
> Please respond to
> batik-users@...
>
> To
>
> batik-users@...
>
> cc
>
> batik-users@...
>
> Subject
>
> Re: Canvas dispose
>
>
>
>
> Hi Praveen,
>
> Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 02:46:10 AM:
>
> > I have many pages each with an applet in it, with the SVG diagram,
> > and users can switch between these pages, sometimes quite rapidly. I
> > noticed there was memory leaking away. I then included a call to
> > stopProcessing, which improved things. So I now have
> > canvas.dispose();
> > canvas.stopProcessing();
>
>    I don't think this is a good sequence.  I would stop at
> canvas.dispose().  The call to 'stopProcessing()' is actually
> likely acting to prevent the action of canvas.dispose().
>
>    I've seen Swing hold onto the canvas much longer than
> desirable because it was the 'focus' component when it was
> not replaced by a new focus component.  So you might try
> removing the canvas from the applet and adding a JButton in
> it's place.
>
> > 1. Is this sufficient?
> >
> > Also, at times during navigation, I see errors such as these logged
> > in the console (I believe if I navigate away before the SVG is rendered):
> > java.lang.NullPointerException
>
>    These may be caused by your calling stopProcessing.

[Praveen Nayak2]

Hi Thomas,

Thanks! The test case helped!

Praveen

thomas.deweese@...

28/10/2009 05:43 PM

Please respond to
batik-users@...

To	batik-users@...
cc	batik-users@...
Subject	Re: Canvas dispose

Hi Praveen,

Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 07:39:39 AM:

> Sorry for the wrong ordering below, my bad. The call is actually
> canvas.stopProcessing();
> canvas.dispose();

  Internally canvas.dispose ends up calling 'stopProcessing()'
so your early call is only likely to cause problems.  I strongly
recommend that you remove it.

> I also have code removing the canvas from the applet:
> this.getContentPane().remove(canvas);
> canvas = null;

   In my experience simply removing the canvas from the
Swing hierarchy is not enough.  You need to give it something
else to 'point at'.  You can take a look at
test-sources/org/apachage/batik/swing/JSVGMemoryLeakTest.doSomething
to see the sequence I have used in the past to 'help' the canvas
go to GC.

> Any idea on console errors in the above scenario? Appreciate your help.

   I suspect the problem is the extra call to stopProcessing.  That
method is async and I think you end up calling dispose on the BridgeContext
twice in two separate threads - which isn't good.  I don't think it will
lead to memory leaks as one of the threads will likely run to completion,
and in any case I'm confident that everything will go to GC if the canvas
goes to GC (which I suspect it isn't currently).

> thomas.deweese@...
> 28/10/2009 04:12 PM
>
> Please respond to
> batik-users@...
>
> To
>
> batik-users@...
>
> cc
>
> batik-users@...
>
> Subject
>
> Re: Canvas dispose
>
>
>
>
> Hi Praveen,
>
> Praveen Nayak2 <prnayak2@...> wrote on 10/28/2009 02:46:10 AM:
>
> > I have many pages each with an applet in it, with the SVG diagram,
> > and users can switch between these pages, sometimes quite rapidly. I
> > noticed there was memory leaking away. I then included a call to
> > stopProcessing, which improved things. So I now have
> > canvas.dispose();
> > canvas.stopProcessing();
>
>    I don't think this is a good sequence.  I would stop at
> canvas.dispose().  The call to 'stopProcessing()' is actually
> likely acting to prevent the action of canvas.dispose().
>
>    I've seen Swing hold onto the canvas much longer than
> desirable because it was the 'focus' component when it was
> not replaced by a new focus component.  So you might try
> removing the canvas from the applet and adding a JButton in
> it's place.
>
> > 1. Is this sufficient?
> >
> > Also, at times during navigation, I see errors such as these logged
> > in the console (I believe if I navigate away before the SVG is rendered):
> > java.lang.NullPointerException
>
>    These may be caused by your calling stopProcessing.