Home » Archived » M2M (model-to-model transformation) » [QVTO] Number Conversion Issue
[QVTO] Number Conversion Issue [message #104811] |
Sun, 10 May 2009 20:38  |
Eclipse User |
|
|
|
This is my first post to the group, so I hope this is a worthwhile
place to post this question (and my potential resolution). During some
headless QVTO testing I was getting errors about casting Floats to
Doubles (between two EMF models with generated and pre-compiled Java
classes). I had a look into the QVTO runtime source and found that in
the org/eclipse/m2m/internal/qvt/oml/evaluator/NumberConversions .java
file the test for number conversion was checking if a target
attribute's instance type were of the Object type (e.g. Integer.class,
Double.class, Boolean.class etc.) and our problem was that the
attributes reported their instance class as being the primitive class
(e.g. int.class, double.class and boolean.class)
I made a change to the NumberConversion.java file so that this would be
taken into account (assuming that Java 1.5+ autoboxing would take care
of the conversions to the primitives) and this patch to the
NumberConversions.java file fixed this issue for us. I was wondering
whether this is something that should be included in the main QVT
source or if that behaviour is what is expected (and so whether we need
to work around it).
I generated a patch in Eclipse which you can grab here:
http://cimphony.org/org.eclipse.m2m.qvt.oml.patch
Cheers
Alan
|
|
|
Re: [QVTO] Number Conversion Issue [message #104840 is a reply to message #104811] |
Mon, 11 May 2009 06:08   |
Eclipse User |
|
|
|
Hi Alan,
Thanks for catching the problem. You even digged into the source code and
provided
a fix, which is great!
Yes, this is this the right and I guess the only place to post such a
question ;),
however the bugzilla is great too, especially if you provide a patch.
Could you please raise a new bugzilla here?
https://bugs.eclipse.org/bugs/enter_bug.cgi?product=M2M& component=QVT_OML-Engine
It's not just to add more work for you ;), but to help us easily track
contributions
so you can appear in the IP log of our project.
Would be great if you provide a QVT code snippet to reproduce the issue.
Regards,
/Radek
On Mon, 11 May 2009 11:33:45 +0200, wrote:
> On Mon, 11 May 2009 02:38:19 +0200, Alan McMorran <alan@mcmorran.co.uk>
> wrote:
>
>> This is my first post to the group, so I hope this is a worthwhile
>> place to post this question (and my potential resolution). During some
>> headless QVTO testing I was getting errors about casting Floats to
>> Doubles (between two EMF models with generated and pre-compiled Java
>> classes). I had a look into the QVTO runtime source and found that in
>> the org/eclipse/m2m/internal/qvt/oml/evaluator/NumberConversions .java
>> file the test for number conversion was checking if a target
>> attribute's instance type were of the Object type (e.g. Integer.class,
>> Double.class, Boolean.class etc.) and our problem was that the
>> attributes reported their instance class as being the primitive class
>> (e.g. int.class, double.class and boolean.class)
>>
>> I made a change to the NumberConversion.java file so that this would be
>> taken into account (assuming that Java 1.5+ autoboxing would take care
>> of the conversions to the primitives) and this patch to the
>> NumberConversions.java file fixed this issue for us. I was wondering
>> whether this is something that should be included in the main QVT
>> source or if that behaviour is what is expected (and so whether we need
>> to work around it).
>>
>> I generated a patch in Eclipse which you can grab here:
>> http://cimphony.org/org.eclipse.m2m.qvt.oml.patch
>>
>> Cheers
>>
>> Alan
>>
>
|
|
| |
Re: [QVTO] Number Conversion Issue [message #104952 is a reply to message #104937] |
Tue, 12 May 2009 06:25  |
Eclipse User |
|
|
|
Hi Alan,
Thanks!
No need for the actual models, just wanted to know the context of usage,
like
assignment, cast or numeric operation etc., exactly as you did in the
bugzilla.
Regards,
/Radek
On Tue, 12 May 2009 11:17:21 +0200, Alan McMorran <alan@mcmorran.co.uk>
wrote:
> On 2009-05-11 12:08:03 +0200, "Radek Dvorak" <radek.dvorak@borland.com>
> said:
>
>> Hi Alan,
>> Thanks for catching the problem. You even digged into the source code
>> and provided
>> a fix, which is great!
>
> Not a problem, figured I might as well make use of it being open-source
> :-) Still happier digging about in Java code than QVT at the moment
> anyway!
>
>> Yes, this is this the right and I guess the only place to post such a
>> question ;),
>> however the bugzilla is great too, especially if you provide a patch.
>> Could you please raise a new bugzilla here?
>> https://bugs.eclipse.org/bugs/enter_bug.cgi?product=M2M& component=QVT_OML-Engine
>
> It's
>> not just to add more work for you ;), but to help us easily track
>> contributions
>> so you can appear in the IP log of our project.
>
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=275824
>
> Done!
>
>> Would be great if you provide a QVT code snippet to reproduce the issue.
>
> I can't include the actual models or transforms due to contractual
> agreements but it was actually just pretty simple EFloat to a custom
> datatype of instance type double.class (i.e. a duplicate of EDouble) and
> I think the main issue has been that the transform was working with
> Genmodel generated Java class files for both models (as opposed to
> Dynamic models).
>
> Cheers
>
> Alan
>
|
|
|
Goto Forum:
Current Time: Tue Jul 22 22:13:37 EDT 2025
Powered by FUDForum. Page generated in 0.29054 seconds
|