Hi Bill:

>From bdelav@hotmail.com Sat Nov 14 11:39:44 1998
>From: "Bill de la Vega" <bdelav@hotmail.com>
>To: lieber@ccs.neu.edu
>Subject: traversal bean generator
>
>Professor, 
>Two questions about the traversal been generator:
>
>1.
>I imagined that the traversal bean generator would read the ICG, CCG, & 
>the map to create a trv file (or some similar data structure) on the 
>CCG. 
>
>Then when running the Traversal Bean would interpret this trv file to 
>deside what events to send out.
>
>The 2 alternatives you discussed in the Demeter seminar were 
>interpreting or compiling the trv file. Interpreting it involves a 
>generic piece of code that can read trv files and emit events. In the 
>compiled alternative the Bean has been generated with a set of functions 
>for traversing the CCG structure (it is hard coded to work with that 
>CCG).
>
>Is this how you imagined the project?
>

It is very close. Let's write a mathematical specification for
the traversal beans.

Input:
Class graph I and traversal strategy S for I.
Class graph C and a map from I to C that
assigns to each edge e of I a strategy T[e] in C (the midterm problem).
We assume that the nodes of S are a subset of the nodes of I and
the nodes of I are a subset  of the nodes of C. (The motivation for
the assumption is
to first solve the problem without name maps). 

Object graph O of C.
Note: I is an interface class graph. C is a concrete class graph. 

Output:
A sequence of events for traversing O (at the concrete class graph
level) following S and the applicable S[e].
The events are node-visit and edge-visit events produced by a Java bean.

I recommend that your Java bean interprets the traversal graph.
Your description of the two possibilities is 100% correct.

>
>2.
>Honestly, I have not been following the discussion concerning the 
>restrictiveness of the map (from ICG to CCG or Traversal Strategy to 
>CG). I just started reading your papers "Traversals of Object 
>Structures: Specification & Efficient Implementation" and "The 
>Refinement Relation of Graph-Based Generic Programs."  

I suggest that you first solve the following simplified problem:

Input:
Class graph I and traversal strategy S for I.
Class graph C.
We assume that the nodes of S are a subset of the nodes of I and
the nodes of I are a subset  of the nodes of C. (The motivation for
the assumption is
to first solve the problem without name maps). 

Then you can firstavoid the restrictiveness of the map.
>
>I don't think that I need to finish the Refinment paper unless you 
>expect the Traversal Bean generator to enforce the "more restrictive 
>notion of refinement" relationship. Do you agree?
>

No, you don't need to finish the refinement paper. The more restrictive
notion of refinement is checked by I-Chen.

>
>3.
>It seems the Demeter seminar is every Mon. at 3PM. Can I find out the 
>topic in advance? That would make it easier to prepare and adjust my 
>work schedule so I can attend.
>

Yes, always same time.
Sorry about that. Will announce it earlier.
>
>______________________________________________________
>Get Your Private, Free Email at http://www.hotmail.com
>

-- Karl