Community
Participate
Working Groups
SourceType#get*QualifiedName() methods return name that are non-standard (there is no standard) for local types, and the javadoc in IType doesn't specify what we should expect from these method in these cases. For anonymous classes, SourceType#get*QualifiedName() methods return names like '<OuterTypeName>$<number>'. This is the format that almost all Java compilers use to name anonymous classes, but this format is not specified, Java compilers can add the identifier they want after the '$'. Even if the compiler uses this format, the way it orders the anonymous classes can be different than the way JDT does it. For local classes, SourceType#get*QualifiedName() methods return names like '<OuterTypeName>$<LocalTypeName>'. This format isn't used by any compiler that I know of. In fact this the format usualy used by compilers form inner (no-local) classes. Compilers usualy use '<OuterTypeName>$<number>$<LocalTypeName>' to name local classes, but as for anonymous classes, there is no specification. I don't know what should be returned, but I'm not sure that what is actualy returned helps anybody. May be the javadoc in IType should say that for local types in source types, the names returned are not 'official' names.
Why is the output so important to you ? These are mostly for presentation purpose. Are you trying to infer back things based on them ? I agree that local type name look weird.
We are using this name when adding and removing breakpoints. In the process of adding/removing breakpoint, we are using IClassFile#getElementAt() or ICompilationUnit#getElementAt() to get a first idea of in which type, the user is trying to add/remove a breakpoint. And we use the name of this type to check if the user is trying to add breakpoint (not existing breakpoint at this location) or to remove a breakpoint (existing breakpoint at this location). I pretty sure that some time ago, you were not returning elements like local types when calling the getElementAt() method, so we had no problems. Now you are more precise in your results, it's great, we just have to manage it. I modified our code to only use type with 'well-know' names (see bug 52385), because it doesn't matter what you decide to do with this bug, you will never return the right names for local types, because there is no 'right' names. I filed this bug report mainly because the implementation does something that is not specified in the javadoc of the interface.
Thanks. I'll update the spec then.
Improved the spec of IType#getTypeQualifiedName(char). Other get*QualifiedName () methods refer to this one.
Verified for 3.0 M9 with build I200405180816.
