[news.eclipse.tools.jdt] Re: Question on Marker use (APT vs. JavaModel)

[Mike Yawn <myawn@xxxxxxxx>]

Mike Yawn wrote:
> Walter Harley wrote:
> > "Mike Yawn" <myawn@xxxxxxxx> wrote in message 
> > news:fbpcnd$4qj$1@xxxxxxxxxxxxxxxxxxxx
> > > Question 1: Is there a way to retrieve the markers created by 
> > > EclipseMessager.printFixableError starting from an ICompilationUnit?
> >
> > Hmm.  It must be possible to pull them off the resource in some way; 
> > after all, the JDT Problems view manages to do so.  In my experience 
> > the challenge is to tell them apart from the other JDT compilation 
> > problem markers.
> >
> > Take a look at AbstractImageBuilder.storeProblemsFor(), and 
> > IncrementalImageBuilder.updateProblemsFor(), in 
> > org.eclipse.jdt.internal.core.builder; and at 
> > APTProblem.getMarkerType(), in org.eclipse.jdt.apt.core.internal.env.
> You're right; I was thinking they must be visible since the problems 
> view shows them.  I think I was trying to see them as I typed; they 
> don't actually become visible to the Problems view (or in my action code 
> until I save the file after making the edit that results in the error.
>
> And then, as you indicate, getting at the content to distinguish the 
> error is an issue.
>
> I looked the at code that creates the markers (EclipseMessager 
> implementation), and it appears that what I'm creating _ought_ to be an 
> APTProbem -- which implements IProblem, so I should be able to cast it 
> to an IProblem, and then use getArguments to retrieve the plugin id & 
> error id.   However, what I'm seeing from .findMarkers is just an 
> org.eclipse.core.internal.resources.Marker, and if I try to cast it to 
> an IProblem I get a ClassCastException.  The Marker errorids seem to be 
> random (they are sequential and unrelated to the errorId I passed in at 
> creation time).
>
> Any ideas of how I can retrieve these markers in a way that allows them 
> to be cast to IProblems (or some other clever way to get at the content 
> -- I'm not sure what lurks behind marker.getAttribute for those who have 
> the attribute name decoder ring)
>
> Mike

Well, I decided to dump out the contents of the 'getAttributes()' map to 
see if that was of any use, and it turns out to have the arguments I was 
trying to get from IProblem.

So just to document this for anyone else that might run into the same 
question, here's what works for me:

public IMarker[] findJavaProblemMarkers(ICompilationUnit cu) {
   try {
      IResource javaSourceFile = cu.getUnderlyingResource();
      String aptProblemMarker = "org.eclipse.jdt.apt.core.compile.problem";
      IMarker[] markers =
         //IJavaModelMarker.JAVA_MODEL_PROBLEM_MARKER does't see APT 
error markers
         javaSourceFile.findMarkers(aptProblemMarker,
            true, IResource.DEPTH_INFINITE);
         for (IMarker marker : markers) {
            Map attrs = marker.getAttributes();
            String arguments = (String) attrs.get("arguments");
            String arg2 = arguments.split("#")[1];
            if (arg2.equals(CodegenAnnotationProcessor.DAO_REGEN_NEEDED))
               System.err.println("DAO regen required");
            else if 
(arg2.equals(CodegenAnnotationProcessor.CGDOI_REGEN_NEEDED))
               System.err.println("CGDOI regen required");
         }
         return markers;
   } catch (JavaModelException e) {
      e.printStackTrace();
   } catch (CoreException e) {
      e.printStackTrace();
   }
   return null;
}