[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [wtp-dev] JSDT inference: type of "this"


On Apr 8, 2008, at 4:56 PM, Phil Berkland wrote:

The default InferenceEngine will create an anonymous type when is sees a Object Literal.
Assuming that you have override handleFunctionCall() in you inferengine to process the "qx.Class.define", in that code
you would "drill down" to that object initializer and call "traverse()" in its field initializers, thus bypassing the default handling of visit(IObjectInitializer).
The handleFunctionCall() should return false, so the InferEngine does not visit the children of the function call .

I've tried this now. Unsuccessful, so far.

I made certain that visit(IObjectLiteral) in my inferencer is not called. While debugging, I even throw Exceptions in the overridden method, to be sure that the call doesn't pass there unnoticed. But still the original problem exists: The inner object literal has an own anonymous type.

I went into debugging to see where it is created:
The anonymous InferredType object that is passed to my QooxdooInferEngine is not created by the QooxdooInferEngine. It is created by another instance of InferEngine (itself, no subclass). So I never get the chance to do something regarding the type creation for ObjectLiterals.


However, it seemed awkward to me why I should reimplement the ASTVisitor aspect of the InferEngine anyway. Having an anonymous type as a placeholder during the type gathering is ok for me, and not uncommon from what I know about type inferencing in other languages. I just want to be able to tell it that it is the same type as "that one".
I didn't find anything in InferredType that suggests this functionality. So I set the superclass of the anonymous type to the known class. This works now for the functionality I intended right now, but is not formally correct and will come back and bite me as soon as I introduce visibility validation.


Do you think it would be reasonable to introduce a field in InferredType that allows me to say what it actually is? It would be close to the referenceClass, only not for array members but for "this" type.

Best Regards,
Matthias Kempka




--
Dipl.-Inform. Matthias Kempka
Innoopract Informationssysteme GmbH
mkempka@xxxxxxxxxxxxxx
Tel: 0721 - 66 47 33 - 0
Fax: 0721 - 66 47 33 29
========================= Legal Disclaimer =====================
According to Section 80 of the German Corporation Act
Innoopract Informationssysteme GmbH must indicate the following information:
Address: Stephanienstrasse 20, 76133 Karlsruhe Germany
General Manager: Jochen Krause, Eric von der Heyden
Registered Office: Karlsruhe, Commercial Register Karlsruhe HRB 7883
============================================================