Re: Demo for Bob Neches

Kurt Stirewalt ([email protected])
Fri, 21 Jul 1995 13:57:10 -0400

I agree with Noi. 3 sounds pretty easy, so we should be able to complete
it with no trouble. 4 is particularly intriguing because it sounds a lot
like a tool Noi has wanted for a while to do incremental MTF evaluation.

>4- Attribute changer: a small program that takes the name of an object, the
>name of an attribute and a value, and tells the model server to set the
>value of the attribute. We might need several of these programs: set, add,
>clear, and one of each for each type.

I think there is potentially a lot of overlap with this tool and the
*back-end* I wrote for MTF. To handle the assignments in all of the weird
contexts that MTF allows, I created what I call evaluation context classes
that act as either left-hand sides or right-hand sides. The constructors
to these classes take qualified string names (like my_pres.parts[...])
and create one of these contexts that may then be evaluated.
The left-hand side context has a method called (what else) assign
that takes a right hand side as a parameter and effects the model
server with the assignment. I'd be interested to see how we need to
extend these classes to handle the attribute changer tool. It would have
the side benefit of testing existing software better and maybe generalizing
it. Is this too ambitious? It seems like it would also afford a nice
development split: (ISI builds the gui part with buttons for add, change,
delete, etc and input widgets; and GaTech builds the "attribute
assignment layer").

Waddaya think?
-- kurt.