[antlr-interest] conditional processing

[Harald Mueller]

Hi

after a quick scanning of your code, I'd say this should work.

However, I don't like the design that much: global variables are per se a bad decision - passing your knowledge around (in single parameters; or, when that gets too cumbersome, in my beloved "context object") is simply more state-of-the-art.

Still, I confess that this argument only is valid if your language will grow one day - then a better design will scale better ... (wouldn't you (or someone) at some time like (or need) to have nested ifs or somewhat more expressive expressions? - all little languages I saw one day wanted to do such things).

At least, you could safe-guard your global variables so that you can only assign a value once to them; then do an evaluation which clears the variables; etc. (in other words, define a clean state machine for them; and implement it - probably in some small classes). That way, any modification of your grammar or your evaluation logic which invalidates your implicit assumptions about when a variable is written and when it is read will at least fly into your face immediately ...

Regards
Harald M.

-------- Original-Nachricht --------
> Datum: Tue, 27 Nov 2007 06:50:22 -0600
> Von: "Mark Volkmann" <r.mark.volkmann at gmail.com>
> An: antlr-interest at antlr.org
> Betreff: Re: [antlr-interest] conditional processing

> On 11/27/07, Harald Mueller <harald_m_mueller at gmx.de> wrote:
> > Hi -
> >
> > You need a context that keeps knowledge about the world - variable
> values, but also results of evaluations e.g. for ifs, returns etc. Usually the
> context is a stack - see any text on fundamentals of progam language
> semantics. The context is threaded through all your calls (as a simple parameter),
> and your calls - e.g. "simpleStatement" - can then react to it:
> >
> > if (boolStack.Top) {
> >     ...do evaluation...
> > }
> >
> > Just imagine what data a virtual machine would need to do what you want
> - exactly these data are what you need in your context.
> 
> >>From what you described, it sounds like my current solution might be
> using the right approach. I don't have to worry about nested
> conditionals, so I probably don't need stacks now. Here's what I have.
> Note how the rule for simpleStatement just remembers what it saw but
> doesn't execute the statement. Also note how the rule that matches
> "IF" executes the statement if the condition evaluates to true. Does
> this seem like a reasonable way to implement this?
> 
> tree grammar MathScriptTree;
> 
> options {
>   ASTLabelType = CommonTree;
>   tokenVocab = MathScript;
>   output = template;
> }
> 
> @members {
>   enum Operation { Add, Subtract, Print }
>   Operation operation;
>   int operand;
>   int value;
> 
>   private void execute() {
>     switch (operation) {
>       case Add: value += operand; break;
>       case Subtract: value -= operand; break;
>       case Print: System.out.println(value); break;
>     }
>   }
> 
>   private static int toInt(CommonTree node) {
>     return Integer.parseInt(node.getText());
>   }
> }
> 
> script: ^(SCRIPT statement*);
> 
> statement
>   : simpleStatement { execute(); }
>   | ^('IF' c=condition simpleStatement) { if ($c.result) execute(); }
>   ;
> 
> simpleStatement
>   : ^('ADD' n=NUMBER) { operation = Operation.Add; operand = toInt(n); }
>   | ^('SUBTRACT' n=NUMBER) { operation = Operation.Subtract; operand =
> toInt(n); }
>   | 'PRINT' { operation = Operation.Print; }
>   ;
> 
> condition returns [boolean result]
>   : 'POSITIVE' { $result = value > 0; }
>   | 'NEGATIVE' { $result = value < 0; }
>   | ^('<' n=NUMBER) { $result = value < toInt(n); }
>   | ^('>' n=NUMBER) { $result = value > toInt(n); }
>   | ^('=' n=NUMBER) { $result = value == toInt(n); }
>   ;
> 
> > -------- Original-Nachricht --------
> > > Datum: Mon, 26 Nov 2007 21:02:46 -0600
> > > Von: "Mark Volkmann" <r.mark.volkmann at gmail.com>
> > > An: antlr-interest at antlr.org
> > > Betreff: [antlr-interest] conditional processing
> >
> > > Part of my AST looks like this.
> > >
> > > ^('if' condition simpleStatement)
> > >
> > > In my tree grammar, the rule for "condition" evaluates the matching
> > > content to true or false. The rule for "simpleStatement" executes the
> > > statement, for example, it could be a simple print statement. I only
> > > want simpleStatement to be executed if condition evaluates to true. I
> > > haven't been able to figure out how to make this work because when
> > > this part of my AST is matched, the rules for both condition and
> > > simpleStatement are fired.
> > >
> > > I'd appreciate any hints on how to achieve this.
> > >
> > > --
> > > R. Mark Volkmann
> > > Object Computing, Inc.
> >
> > --
> > GMX FreeMail: 1 GB Postfach, 5 E-Mail-Adressen, 10 Free SMS.
> > Alle Infos und kostenlose Anmeldung: http://www.gmx.net/de/go/freemail
> >
> 
> 
> -- 
> R. Mark Volkmann
> Object Computing, Inc.

-- 
Ist Ihr Browser Vista-kompatibel? Jetzt die neuesten 
Browser-Versionen downloaden: http://www.gmx.net/de/go/browser