Eclipse Community Forums: Objectteams » support for callin-interception of constructor-calls

Help

Home

Home » Language IDEs » Objectteams » support for callin-interception of constructor-calls

Show: Today's Messages :: Show Polls :: Message Navigator

support for callin-interception of constructor-calls [message #1045722]

Sat, 20 April 2013 15:25

Kai Zimmermann
Messages: 16
Registered: October 2012

Junior Member

https://bugs.eclipse.org/bugs/show_bug.cgi?id=316616

Is this function already implemented?

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1045727 is a reply to message #1045722]

Sat, 20 April 2013 15:33

Stephan Herrmann
Messages: 872
Registered: July 2009

Senior Member

Unfortunately not yet, but thanks for bringing this to my attention Smile

Assuming we'd add support for callin bindings for constructors with the only option to define "after" bindings, would that already create a benefit for you?

Stephan

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1045730 is a reply to message #1045727]

Sat, 20 April 2013 15:38

Kai Zimmermann
Messages: 16
Registered: October 2012

Junior Member

Only if you got an idea for a workaround Wink

I want to insert another label into the string field (super call):

package org.eclipse.pde.internal.ui.editor.feature;

imports ...

public class PluginSection extends TableSection
  implements IPluginModelListener
{
  private OpenReferenceAction fOpenAction;
  private TableViewer fPluginViewer;
  private Action fNewAction;
  private Action fDeleteAction;
  private SortAction fSortAction;

  public PluginSection(PDEFormPage page, Composite parent)
  {
    super(page, parent, 128, new String[] { PDEUIMessages.FeatureEditor_PluginSection_new, 0, PDEUIMessages.FeatureEditor_SpecSection_synchronize });
    getSection().setText(PDEUIMessages.FeatureEditor_PluginSection_pluginTitle);
    getSection().setDescription(PDEUIMessages.FeatureEditor_PluginSection_pluginDesc);
    getTablePart().setEditable(false);
  }

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1046272 is a reply to message #1045730]

Sun, 21 April 2013 13:23

Stephan Herrmann
Messages: 872
Registered: July 2009

Senior Member

That's a tough one.

For an *after* callin we can weave some code at the end of the constructor body. I just started looking into this and a solution might still fit into the schedule for Kepler.

Weaving anything *before* the constructor body would necessarily be restricted to binding a static role method, because we don't yet have a base instance to bind to. For this some new weaving scheme needs to be created - possible but unlikely for Kepler.

For a *replace* binding I haven't found a way how the JVM would let me do that. The super constructor call must be the first thing we do in the constructor body. There's no way we could cut that out, delegate to the role and let the role invoke the super constructor instead. Any code patterns we'd like to create in this vein would be rejected by the VM.

Sorry, constructors are very special for the VM, there's little we can do to harmonize them with callin bindinds -- unless someone comes up with a real cool weaving scheme, which we haven't yet thought of.

regards,
Stephan

BTW: your particular problem can probably be solved by intercepting createViewerPart(String[] buttonLabels),
that's what I did in the hot fix for https://bugs.eclipse.org/374789 Smile

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1046279 is a reply to message #1046272]

Sun, 21 April 2013 13:36

Kai Zimmermann
Messages: 16
Registered: October 2012

Junior Member

I can not say I was particularly knowledgeable with JVM or bytecode, but for testing I use powermock. Somehow the Lib bypasses this problem ..

http://powermock.googlecode.com/svn/docs/powermock-1.5/apidocs/org/powermock/api/mockito/PowerMockito.html#whenNew(java.lang.Class)

.. maybe their code gives you an idea.

Thanks for the hint to your hot fix.

[Updated on: Sun, 21 April 2013 13:37]

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1046305 is a reply to message #1046279]

Sun, 21 April 2013 14:30

Stephan Herrmann
Messages: 872
Registered: July 2009

Senior Member

From a brief look at the PowerMock code I can't find anything that solves the problems I mentioned.
One theory is: they don't actually weave into the constructor but into each constructor *call*, is that possible?
(Object Teams performs all weaving only in the bound base classes, never in clients of these base classes).

I'm not even sure PowerMock solves the problems at hand: when you intercept a constructor, can you really create an instance of the original class just using a different constructor body?? Returning a mock instead is an entirely different matter.

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1047136 is a reply to message #1046305]

Mon, 22 April 2013 17:03

Kai Zimmermann
Messages: 16
Registered: October 2012

Junior Member

On topic: Intercepting createViewerPart works Very Happy

Off topic: I can create all kinds of methods, but with a single callin method the source code can not be formatted anymore Confused

Has this ever happened to you?

My ide configuration:
Version: Juno Service Release 2
Build id: 20130225-0426
With

Object Teams Development Tooling 2.1.2.201301271217
Object Teams Equinox Integration 2.1.2.201301271217
Eclipse JDT Plug-in Developer Resources 3.8.2.v20130116-090414-8-8nFu3FNOfwKLRuqgXKIy9z0I83
Eclipse PDE Plug-in Developer Resources 3.8.2.v20130116-091538-7c7wFj0FFt6Zr9bd4AM1JEQMS
Eclipse for RCP and RAP Developers 1.5.2.20130211-1820

import org.eclipse.pde.internal.ui.parts.StructuredViewerPart;
import base org.eclipse.pde.internal.ui.editor.feature.PluginSection;

@SuppressWarnings("restriction")
public team class MyTeam{

	protected class MyPluginSection playedBy PluginSection {

		 callin StructuredViewerPart createViewerPart(String[] strings) {
		 return base.createViewerPart(strings);
		 }
		 }}

Edit (23.4 07:47 CEST): Save Action "Corrent intention" works, but manually not Shocked

[Updated on: Tue, 23 April 2013 01:48]

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1047628 is a reply to message #1047136]

Tue, 23 April 2013 09:03

Stephan Herrmann
Messages: 872
Registered: July 2009

Senior Member

Glad to here that the createViewerPart-trick helped.

Regarding the formatter: I admit that this part is not quite as well-tested as the rest of the OTDT.
Thanks for reporting, it turns out to be caused by the base call inside the callin method.

See https://bugs.eclipse.org/406316.

thanks,
Stephan

Report message to a moderator

Re: support for callin-interception of constructor-calls [message #1053835 is a reply to message #1047628]

Tue, 07 May 2013 15:33

Stephan Herrmann
Messages: 872
Registered: July 2009

Senior Member

Stephan Herrmann wrote on Tue, 23 April 2013 15:03

Regarding the formatter: I admit that this part is not quite as well-tested as the rest of the OTDT.
Thanks for reporting, it turns out to be caused by the base call inside the callin method.

See https://bugs.eclipse.org/406316.

The fix for this bug will be available in 2.2 M7 (due end of this week).

Report message to a moderator

Previous Topic:	Book chapter published: Confined Roles and Decapsulation in Object Teams -- Contradiction or Synergy
Next Topic:	"Object Teams Patch for JDT/Core" issue on update

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

]

Current Time: Mon Oct 07 06:30:35 EDT 2013