Re: Online explanation of Dijkstra's Guarded Command Language

(Chris Glur) wrote:
> #
> # Google didn't find me a 'Dijkstra's Guarded Command Language'
> # description. Can some one point me to an online reference.


SM Ryan wrote:
> University of Texas has his papers on-line:
> http://www.cs.utexas.edu/users/EWD/
> Perhaps somewhere in there.


Yes, Thanks !!
3 articles on 'Dijkstra's Guarded Command Language' seen.
2 in *.pdf format [images of his type-written manuscripts] I fetched.
It's good quality stuff. Tedious to re-type and try to understand at the
same time.
Some trivialising of Dijkstra's work which I read on the net after his
recent death, definitely seems unjustified.
----
> # I'm interested in methods of 'transforming code' , other than lisp.
>
> # I've started analysing M. Brandis' these of 1995 re.
> # `guarded single-assignment form' compiler. [ for Oberon]


Vidar Hokstad wrote:-
> I think you need to explain what you are looking for a bit better -
> it's not very clear how the three things you mention are related to
> what you are looking for.


It's very vague/fuzzy:
* oberon just happens to be my language (& OS) of choice, and some
        description which I read about this GSA form left an idea that
        'what I describe below' was acheived/approached.
* Guarded Command Language was from Dijkstra, an originator of
      'structured programming'. I too seek to remove the 'art'.
* A look [long time ago] at notations of 'formal representation',
      showed a big-job to transform it to existing source code.
* I was browsing a book: Knowledge-Based Program Construction
              David Barstow 1988
    "Included within this review is an example of the
operation of the system and details of some of the internal rules
which are applied to transform a high-level design of an algorithm
into a Lisp implementation. The expert system refines both data and
looping structures by pinpointing constraints under which they act. "


* I beleive/hope there exists a method of coding at a higher more
        formal method [preferably by just picking components off a
        menu-tree] and allowing automatic code generation/optimisation.


Perhaps I'm unduly optimistic, from a project which I did in the 70's
re. 8-bit micros: p-code interpreters [for various machines] which
had 2 stages of 'transformation':
1. simulate a stack machine,
2. the compiler was an augmented-finite-state-machine [original
    idea, although I later read that Perle has done it], which was dead
    easy to program and maintain. The nested syntax diagrams of a
    subset of Pascal was just laid out, and when the 'train passed the
    various stations', the appropriate actions occured.


So by the 2 transformations, one could work at a very
comfortable/naive level [toy train diagrams] to build the compiler[s]
for the different CPUs.


I was thiking/hoping that GSA might show a more cannonical/formal
representation of algorithms PLUS the transformation route(s) to
executable code.


== Chris Glur.

Return to the comp.compilers page.
Search the comp.compilers archives again.