- The MTF parser with no hooks should be done by the time we all come back
from CHI. This MTF parser will read in a whole text file and store information
in the model server or report error and stop when an error is found.
- To move in the direction to support interactive input to the model,
in addition to the MTF parser there will be some routines which can be called
to parse one attribute line at a time, to instantiate an object with no
attribute, etc. To make the long story short, this will require Kurt to pull
some tricks on the MTF parser to separate parsing of Expressions and
attribute lines from the object parsing. You shouldn't know the difference
when inputting a file at a time. This separation will happen after the whole
file parsing is complete and at the same time we pin down the syntax for
Expressions.
To help you understand why the separation, here's the plan I have from the task
side. I'm trying to help myself input more task mtf files for testing
my own task mrl and the concepts in the task model. Since I'm a visual person,
putting tons of tasks in a sequential file is driving me nut and I can't tell
whether things will be put in the model properly or not. I'm going to write
a simple command-line interface which will let me do some simple commands like:
create task A
Task A
pre-conditions = [.....]
effects = [.....]
create task B
Task B
add parameter B1
Task A
add parameter A1
....
I'll use Amulet to help me display the task structure as it is incrementally
placed in the server. If you guys want to be challenged, I can try to use
your browser.mtf as a way to display a task individually. Want to try????
At any point, I can write to an mtf file from the input I accumulated so far.
Making the parser cooperate with my little command-line module will place the
parser at a stage where it can work with a visual environment in an
incremental fashion. No work is lost here. I'll explain this in a more
diagrammatic fashion when we meet at CHI.
Also, on a separate note, we've hired a GRA, Yussuf, to work with Siemens in
cooperation with the Mastermind project. He's starting looking into building
a PC version of a task modeling environment. Matthias and I are still debating
whether doing a PC Amulet version or a visual C++ version would be more
beneficial to Siemens. For us, I think the choice is clear.
--Noi