Prism criticism

[Maneesh Yadav]

On Thu, 3 Jun 1999, Jim Little wrote:

> Maneesh Yadav wrote:
> > 
> > You're goals seem clearer to me now, and I agree with them for the most
> > part.  However I really don't see what you are trying to sya here.  Is the
> > meaning part just a part of the code?
> 
> I'm not sure if I understand your question, but I'll try to answer it
> anyway.
> 
> Source code is a model of a program.  It has an intrinsic meaning: the
> program it represents -- and an extrinsic meaning: the program you think
> it represents.
> 
> The purpose of a metamodel is to define the meaning of a set of models
> so that when you look at the model, the extrinsic meaning (what you
> think it is) is the same as the intrinsic meaning (what it really is).
> 
> To put it another way, a metamodel is a language specification.
> 
> 
> > How is this different than writing out a c struct and a comment
> > beside it?
> 
> It's not really any different.  I could have used C's basic data types
> to model programs, but I chose not to.  C isn't really that well suited
> for the purpose -- the basic types already have an intrinsic meaning (an
> 'int' is a number, a 'char' is a character), and they're fairly limited
> (no collections).  C++ has collections, but now we're talking code, and
> it just makes things messy.  As it stands, the Prism meta-metamodel is
> fairly pure, almost orthogonal, and (most importantly) small and simple.

So tell me, wouldn't C++ augumented with a bit datatype be practically
equivalent?  Just about any datatype you can think of, you'd prbably want
to model using ints' char's and the like instead of bits.  And if you
really want to use bits, you can use the binary operators or define your
own bit class if you really want....

> 
>  
> > Jim give me a clear example on how using Prism's methods provides more
> > elegance/usefullness whatever, than doing it in C++....that's all that you
> > need to prove your case.
> 
> I can't give you any useful examples at this time.  Prism is done, but
> the infrastructure required be more useful than C++ (mixing several
> models into one) has yet to be started.  The best I can do is give you a
> concept.  It's a fabricated problem, but I'm trying to mix together
> multiple issues just like a real program does:
> 
> 
> Imagine writing a GUI program to find all lines beginning with "//" in a
> file.  Instead of struggling with regular expressions in C++, or visual
> development in Perl, you use multiple Prism languages.  One just for
> visual layout.  One just for regular expressions.  And one just for file
> access.  You glue them altogether with an event-driven messaging
> language. 
> 
So you are saying that prism will allow you to deifne your own language
grammar?  I really don't see how this follows from your spec, not only
that it can already be done in LISP variants among other languages...


> We're not assuming you've ever used any of these things before,
> either... just that you're familiar with the basics of Prism, C++, and
> Perl.  Rather than thumbing through massive tomes trying to figure out
> how to write a regular expression parser (for C++) or how to layout
> GUI's (for Perl), you find just the metamodels you're interested in,
> scan their five page specifications, spend a bit more time understanding
> their related (and minimal) languages, and get the entire project done
> in a couple hours. 

And this is different from using normal libraries how?

 It looks something like this:
> 
> # This is a comment
> ** USE Prism/Event **   # 'use' tells compiler what language
> SearchMain {
>    AT START {
>       TELL SearchGui start;
>    }
>    EVERY (1) SECOND {
>       # Just showing some event metamodel possibilities, hee hee
>    }
> }
>    
> ** USE Prism/Event **
> ** EMBED Prism/Gui **
> SearchGui {
>    WINDOW frameWindow CONTAINS {
>       LISTBOX resultsList BELOW {
>          TEXTFIELD filenameField LEFTOF
>          BUTTON processButton
>       }
>    }
>    WHEN TOLD start {
>       TELL frameWindow display;
>    }
>    WHEN processButton PRESSED {
>       TELL SearchEngine match(GET FilenameField contents);
>    }
>    WHEN TOLD displayResults(results) {
>       ** EMBED Prism/Loop **
>       ** EMBED Prism/Collection **
>       TELL resultsList clear;
>       ITERATE i = results;
>       WHILE (results.HASMORE) {
>          TELL resultsList display(results.NEXT);
>       }
>    }
> }
> 
> ** USE Prism/Event **
> ** EMBED Prism/File **
> ** EMBED Prism/Regex **
> SearchEngine {
>    WHEN TOLD match(filename) {
>       ** EMBED Prism/Loop **
>       ** EMBED Prism/Conditional **
>       ** EMBED Prism/Collection **
>       ** EMBED Prism/String **
>       TEXTFILE file = filename;
>       NEW LIST result;
>       UNTIL (file.EOF) {
>          STRING line = file.GETNEXTLINE;
>          if (line MATCHES [^\w*//]) {
>             APPEND result(s);
>          }
>       TELL SearchGui displayResults(result);
>       }
>    }
> }
> 
>From what I can make of it, this is really not terribly  different than
using a high level library under and existing language?


> Anyway, that's an example.  Not the best one in the world, and I think
> real programming in Prism will be different, but it's the best one I can
> give at this stage.
> 
> If you're wondering why you don't see any _Bits, _Streams, or _Maps in
> the above code, it's because atoms are the infrastructure, not the
> goal.  I'm still building the infrastructure, so naturally that's what I
> focus on.  When I'm done, normal programmers won't see the
> infrastructure; they'll just deal with languages.  Only expert
> programmers will deal with the Prism infrastructure, and only when they
> make new metamodels and languages for ordinary programmers to use. 
> (Sort of like library developers and application developers today.  Most
> people don't create template classes in C++, they just use the ones that
> expert library programmers have created.  Same concept for Prism
> programmers and metamodels.)
> 
> Jim
> 
> Prism is at http://www.teleport.com/~sphere
>