RJLRef: $CASE/04s522/statusAJLopez040809Fdbk.txt

> From ajlopez@cs.uml.edu  Mon Aug  9 22:13:51 2004
> Subject: Project works fine (SetUp Phase and Run Phase)
> 
> Good evening, Dr. Lechner!
> 
> The project is working pretty good, even though the source code I have is 
> not 100% accurate. I would like to show you what I've done. We just need 
> to set a meeting.

We can meet Thur around 530M.
I need to go to NH from Westford 215PM, return about 515 estimated.

You do not specify the new or revised .dat and .c/.h  files 
that are produced by setup and compiled or consumed by run phase?
Please reply ASAP and give me those file names.

Your revised programs must be state-model-driven. The example code can onlybe 
hand-tailored to fit the application by modifying StateModelCreateCompletely()
and its friend functions 
which are based on the quoted state transition event string initializations.  


> 
> I said "not 100% ac-curated" because the Setup phase doesn't generate the 
> same archive that we worked with in our last meeting (HominidDB.dat). I 
> think the source code is not the one which generates that archive, instead 
> it generates a smaller one.
> 
> The main change to the Setup phase is switching from function address to 
> function index, and function array. The main change in the Run phase 
> is executing function calls with the function pointer array.
> 
> I also made some changes to my initial source code -when I realized how 
> olcarch files, and pr_util files work-. And, of course, I have some 
> question to make in order to improve what I did:
> 

> One is that I have to create different function arrays in order to keep 
> different states. So, I must check which array to use. In this case, I 
> create HDFunctTbl, and HLFunctTbl, and I check if STcurr->ActName == "l". 
> It's not a great way to work.
> 
I suppose you mean different arrays for different State Models?
Note there are two levels of complexity:
(1) One func ptr array per SM-table-row  and one SM-table-row per AC-row
and its subclass (HD or HL in this example).
(2)  One func ptr array per SM-table-row and MULTIPLE State Models (SM-rows)
per Active Class AC (like bde, which has one SM per method and multiple methods
per class: 
	cd $RL/src; grep '(switch\(state\))' *ops.cc
(bendptops.cc captionops.cc graphops.cc linkops.cc nodeops.cc textops.cc)

I mentioned multiple SM-rows per AC-row before. It has to do with sub-models
called from a top-level class model for behavior.  E.g., the left-side OOMenu 
in bde calls sub-machines for create/delete/resize/restyle/move methods
for most drawable object classes in the bde schema.
I haven't decided whether this state model tree (like a pop-up menu tree) 
requires more than one SM per AC. 

One URL that is fundamental but not specific to this question is to a paper by 
parrt@#usfca.edu: Enforcing Strict Model-View Separation in Template Engines.
I have a .pdf copy.

> I have some other, but I think it's better ask you during our meeting.
> 
> You can check how the program works. There is a folder setup having the 
> LCPsetup.exe generating HominidDB.exe. It needs a parameter (--setup) in 
> order to work. And there is a Run executable file LCPmain (at the 
> executable folder) reading the HominidDB.dat file.
> 

Surely you mean generating HominidDB.c/.h not .exe? gcc generates the latter.

> To be sure that the program works the same that the one we check, I update 
> the old HomonidDB.dat file. I changed function address by array index, 
> and all output is exactly the same.
> 

You don't say how the new event dispatcher was modified for run phase?
you don't say how two Makefiles handle separate compilation of setup 
and run phases?


> I hope you'd spent a nice vacation.
> 
> Alfredo Lopez
>