Subj : baja.c / unbaja.c
To   : Angus McLeod
From : Digital Man
Date : Thu Sep 08 2005 06:09 pm

  Re: baja.c / unbaja.c
  By: Angus McLeod to Digital Man on Thu Sep 08 2005 09:44 pm

 >   Re: baja.c / unbaja.c
 >   By: Digital Man to Angus McLeod on Thu Sep 08 2005 12:44:00
 > 
 >  >  > I meant the use of Dijkstra's *shunting* algorithm or a variant thereo
 >  >  > to convert infix with nested sub-expressions to postfix, for simplifie
 >  >  > evaluation by the low level processor (in this instance, the BAJA
 >  >  > low-level execution code).  I've always found it a little strange seei
 >  >  > parenthesis actually embedded in the low level code as they are (in
 >  >  > tokenized form) in Sync's .BIN files.  These are usually removed by th
 >  >  > (semi-)compiler by application of some variation of Dijkstra's
 >  >  > shunting algorithm.  It's a very basic sort of code-optimization.
 >  >
 >  > Honestly, I wasn't aware of any such algorithm and it didn't occur to me 
 >  > the compiled AR strings needed much opimization. :-)
 > 
 > I don't mean to imply that ARS evaluation requires any sort of runtime
 > optimization.  But I've studies compilers and compiler design and I had
 > never come across any expression-evaluation code that didn't reduce infix
 > to postfix.  Not until I first examined the BAJA compiler and it's output,
 > years ago.
 > 
 > You may recall that I wrote a BAJA de-compiler long ago. 

That does seem vaguely familiar. Do you still have that code around anywhere?

 > That was before
 > SBBS source code became publicly available.  At the time you sent me a
 > copy of the header file containing all the current op-codes (probably
 > cmdshell.h, unless the name has changed).  It worked, but unfortunately,
 > by the time it was finished, it was obsolete.  The rate of development of
 > BAJA was high at the time, and new 'verbs' were being introduced faster
 > than I could keep up with, and I decided not to make myself a nuisance by
 > demanding a fresh copy of the header file every two/three days.  :-)

Oh how nice it is to have CVS. :-)

 > Anyway, I remember *then*, spelunking my way through .BIN files, how
 > surprised I was to find ARS expressions with parenthesised sub-expressions
 > in the low level code.  Since Dijkstra's shunting algorithm eliminates
 > sub-expressions, finding tokenized parenthesis in the low-level code was,
 > somehow, disturbing!  :-)

Can you give me an example of a postfix version of the following ARS:

((LEVEL 80 OR FLAG S) AND AGE 18) OR LEVEL 90

or

LEVEL 90 OR (TIME 12:00 AND TIME NOT 18:00)


 >  > In hind-site, I could've made a *lot* of different/better design decision
 > 
 > Of course!  And let me quote myself:
 > 
 >  >  > It wasn't a criticism, just an observation.

I didn't take it that way.

 > I have never looked back on any non-trivial code I ever wrote without
 > experiencing the disturbing thought that I could have done better.  Even
 > with code that worked flawlessly over a long lifespan without any problems
 > ever being isolated, I always end up feeling that I could improve on it.

Me too! I wrote some programs that looked and worked great on the surface but
under the hood were horrendous examples of programming. <shrug>

                                                  digital man

Snapple "Real Fact" #190:
Thailand means "Land of the Free." 

.