Re: [aspectj-users] Capturing data at a pointcut

[Eric Bodden <eric.bodden@xxxxxxxxxxxxxx>]

Hi S (?),

you can get the field's name from the joinpoint's signature. I am not
sure if it's qualified but if it's not then you can qualify it
yourself with the class of the target object. This page here could
help:

http://blogs.codehaus.org/people/tirsen/archives/000690_undo_in_aspectj.html

Eric

2009/4/22 100ji <itz100ji@xxxxxxxxx>:
> Hi Andrew,
>
>
> On Wed, Apr 22, 2009 at 11:55 AM, Andrew Eisenberg <andrew@xxxxxxxxxxxx> wrote:
>> Hi,
>>
>> A little confused about what you mean by #1.  What is the fully
>> qualified name of a local variable?  Or are you only looking to do
>> this for fields?  If so, how are you planning on distinguishing them?
>>
>
> Thanks for the response. Sorry about the ambiguity. I need a fully
> qualified qualified name if it's a field and the name if it's a local
> variable. I am doing some dynamic analysis to figure out which fields
> of type java.util.collection in an object are shared by multiple
> threads. For a successful analysis, I have to also take aliases into
> account. For this I am capturing the object pointed by each of these
> fields and consider all fields that point to same object as aliases.
> Even though this technique is not general it works for me.
>
>> It is not hard to get the type of the field/variable:
>> thisJoinPointStaticPart.getSignature().getDeclaringType()
>>
>> But getting its name and the class that declares it is harder.  This
>> is just not information that AspectJ stores.  One of the reasons that
>> this is not stored is because aliasing can make this information
>> meaningless.  For example:
>>
>> class C1 {
>>   Set s = new HashSet();
>>
>>   private void doNothing() {
>>      Set t = s;
>>      t.add(null);
>>      s.add(null);
>>   }
>> }
>>
>> Here objects t and s are the same, but have different names.  What
>> would you want to do here?  Do you really want to know where the
>> variable was declared or do you want to know what the variable points
>> to?
>
> I want to know the name of the variable/field and what it points
> to(it's value). If there are aliasing effects(s and t), that's ok. I
> can work around that. Also, since one can work around this limitation,
> I think AspectJ should expose this information. This makes AspectJ a
> good tool using which one can implement dynamic program analysis. Btw,
> is anyone aware of any publicly available dynamic program analysis
> tools built on top of AspectJ.
>
> You said that getting the name of the local variable/field and the
> class is much harder. Which thankfully means it's not impossible. How
> can I get that information?
>
> Thanks,
> -S-
>
>>
>> On Wed, Apr 22, 2009 at 12:31 AM, 100ji <itz100ji@xxxxxxxxx> wrote:
>>> Hi all,
>>>
>>> I am implementing some dynamic programming analysis using AspectJ. For
>>> example, consider the program:
>>>
>>> Set<Object> s = new HashSet<Object>();
>>> s.add(o1);
>>> s.remove(o1);
>>>
>>> In this program at each add and remove method call, I need the
>>> following information:
>>>
>>> 1) a string showing the type of s and a fully qualified name of the
>>> field with the name of the field. "java.util.Set com.x.y.z.a.b.c.s".
>>> 2) a reference to object o1.
>>> 3) a reference to set s.
>>> 4) The lineno and the file name at which the call is taking place.
>>>
>>> I can get (2), (3) and (4) using the argument, target and
>>> thisJoinPoint.getStaticpart.getLineNo() respectively. How can I get
>>> (1) the name of the field s and it's type from a point cut?
>>>
>>> Thanks,
>>> -S-
>>> _______________________________________________
>>> aspectj-users mailing list
>>> aspectj-users@xxxxxxxxxxx
>>> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>>>
>> _______________________________________________
>> aspectj-users mailing list
>> aspectj-users@xxxxxxxxxxx
>> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>>
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>



-- 
Eric Bodden
Sable Research Group, McGill University
Montréal, Québec, Canada