RE: [aspectj-users] within

["Alan Brown" <abrown@xxxxxxxxxxxxxxxxx>]

Thanks for the help.  Fear not about the naming of the pointcuts.  That was sheer frustration coming through at the end.  I actually changed the name from it’s original ‘wibble’ for group consumption.  But the package advice is well taken.

 

alan

 

-----Original Message-----
From: Ron Bodkin [mailto:rbodkin@xxxxxxxxxxxxxx]
Sent: Wednesday, February 04, 2004 12:16 PM
To: aspectj-users@xxxxxxxxxxx
Subject: Re: [aspectj-users] within

 

Alan,

 

You are matching *execution* join points (i.e., execution of setValue). The easiest fix for your code is to use call join points (since the call is within your aspect, whereas the execution is NOT):

 

pointcut subTest(OpsBaseAttribute attr) : call(public void setValue(String))

                                                        && this(attr)

                                                        && cflow(execution (public void Session.set*(String)))

                                                        && !within(aspects.DateSub);

 

If you *really* need to use an execution join point, then you need to exclude any execution in the cflow of your aspect, e.g.,

 

 

pointcut subTest(OpsBaseAttribute attr) : execution(public void setValue(String))

                                                        && this(attr)

                                                        && cflow(execution (public void Session.set*(String)))

                                                        && !cflowbelow(adviceexecution() && within(aspects.DateSub));

 

The AspectJ FAQ describes more about call vs. execution join points (http://dev.eclipse.org/viewcvs/indextech.cgi/~checkout~/aspectj-home/doc/faq.html?rev=HEAD&content-type=text/html#q:comparecallandexecution).

 

Also, two style points:

1) I'd suggest using a more descriptive pointcut name; it's considered good style to name what is happening at a pointcut (i.e., setting a string value), not what your implementation uses it for. This also enables reuse (and simplifies maintenance).

2) I'd suggest not using an aspects package. If you want to segregate aspects for build purposes, I'd prefer using a separate source folder (e.g., aspect/src) but use "natural" packages. This becomes especially relevant when using package-level (default) access. 

 

Ron

 

Ron Bodkin

Chief Technology Officer

New Aspects of Software

m: (415) 509-2895

 

 

------------Original Message------------

From: "Alan Brown" <abrown@xxxxxxxxxxxxxxxxx>

To: <aspectj-users@xxxxxxxxxxx>

Date: Wed, Feb-4-2004 12:07 PM

Subject: [aspectj-users] within

I must be misunderstanding what how to use this operator.

 

I have the following test code…

 

pointcut wibble() : execution (public void Session.set*(String));

 

pointcut subTest(OpsBaseAttribute attr) : execution(public void setValue(String))

                                                        && this(attr)

                                                        && cflow(execution (public void Session.set*(String)))

                                                        && !within(aspects.DateSub);

 

 

void around (OpsBaseAttribute attr) : subTest(attr)  {

    System.out.println("supposedly calling set on an attribute in the Session object for name ..." + attr.getName() + "...");

    if (attr.getName().equals("LastLogin") ||

        attr.getName().equals("Created") ||

        attr.getName().equals("LastModified")) {

        System.out.println("resetting for " + attr.getName());

        attr.setValue("2/02/2004 2:22:22 PM");

    }

  }

 

as you can see I’m making a call in my around advice that will re-trigger the pointcut.  Hence my stating “!within(DateSub)”, where DateSub is the name of the Aspect that all this code exists within.

 

My question is simple:  Why do I get infinte recursion when ‘within’ should mask out calls to setValue() from within the pointcut…

 

Thanks for your help.

 

alan