Hi Joe:

we probably need some interaction on this but here is 
information for Doug, in case he asks.

Our tool development strategy is described at:
http://www.ccs.neu.edu/research/demeter/tools/DAJ/tool-plans.html

We have big plans for DAJ.

Information about DAJ is at:
http://www.ccs.neu.edu/research/demeter/DAJ/

Applicability to DRE systems:

1. DRE systems need to follow the Law of Demeter in most parts.
See e.g.
http://www.ccs.neu.edu/research/demeter/related-work/pragmatic-programmer/book-recommendation.txt

DAJ makes it easier to follow the Law of Demeter.

2. Here we need some discussion:

We plan the following addition to DAJ:

declare traversal t1 = "from Sender to Contract";

from this we generate an AspectJ aspect:

aspect TypePattern {
  public pointcut target_t1():
    target(Sender||...||Contract);
  public pointcut this_t1():
    this(Sender||...||Contract);
}

Then we can use the target_t1() pointcut, for example,
to log calls whose target is any of the classes between
Sender and Contract.

The reason why we need to make this detour through defining "standard"
pointcuts is that AspectJ currently does not support named type patterns.

This approach was actually proposed by one of our open source developers
in China.

Would this be useful in your context so that you can talk about
a set of classes robustly?
Rather than saying Sender||A||B||C||D||E||F||Contract
you say
from Sender to Contract

-- Karl


From jloyall@bbn.com  Mon Jul  1 11:07:42 2002
Date: Mon, 01 Jul 2002 11:03:56 -0400
To: lieber@ccs.neu.edu (Karl Lieberherr)
From: Joe Loyall <jloyall@bbn.com>
Subject: DAJ
Cc: wand@ccs.neu.edu

Hi Karl -

At the last PI meeting, we had to identify Go-No-Go decision points for our 
work for July 2002. For your work, we identified the following one for the 
connection work and one for the DAJ work that you are doing. We decided to 
punt on further AC-connection work, at least for now, so I wouldn't worry 
about that one (we made a no-go decision). For the DAJ work, you provided 
the following "Go-No-Go" decision point:

Alpha version of DAJ is demonstrated successfully in small DRE examples to 
reduce tangling

What's the status of this? Have you demonstrated DAJ in any DRE examples? 
You mentioned some metrics or comparisons that you had done during the 
visit last week. Were any of these in DRE examples? What do you want me to 
say if Doug asks the status of this?

Joe


--
Joseph P. Loyall                            phone:  617-873-4679
BBN Technologies                            fax:    617-873-4328
10 Moulton Street                           e-mail: jloyall@bbn.com
Cambridge, MA 02138

http://www.dist-systems.bbn.com/people/jloyall/