[antlr-interest] build issues: bytecode assembly generation

[Terence Parr]

On Oct 21, 2004, at 10:53 AM, Matt Benson wrote:

>
> --- Terence Parr <parrt at cs.usfca.edu> wrote:
>> My goal is basically to avoid .class file generation
>> / reading totally
>> by making the compiler do it.  All I would have to
>> do is open up the
>> ..class file and replace a blank method (there's a
>> comment as the method
>> containing the bytecodes)
>> with the real code after I assemble it. The only
>
> Hmm, seems if you have to know where the classfiles
> are anyway you might as well write them yourself.  Or
> create a part of the ANTLR runtime that would compile
> the bytecodes from byte[]s at runtime (ducking).

I just had another idea.  The original problem is "I need gotos in 
Java."  What if I could make a quick translator for a .java file with 
special comments that would fix up the .class file to allow the gotos?

Let's see...one might do this (remember, this file is *generated*):

     public static int dfa1(IntegerStream input) {
         short from, to;
         int i = input.LA(1);
         if ( (i>='a'&&i<='z') || (i>='0'&&i<='9') ) {
                 input.consume();
                 from = (short)0xBAAA; // a secret GOTO
         }
to=(short)0xBAAB; // a secret GOTO target
         return 0;
     }

We need some unique thing that doesn't get thrown out by the compiler 
to signify the goto location and the target of the goto.  The

from = (short)0xBAAA;

results in the following bytecode.

     sipush          -17750

The target is just:

     sipush          -17749

Now we probably need something a little more unique like two in a row, 
but you get the idea (they need to be short so the operand doesn't get 
dumped into the const pool).

Here's the cool translation part.  You look in a .class file for a 
binary pattern then replace goto targets with no-ops, but you know the 
byte code address now :)  The native compiler will throw these out.  
The gotos themselves are simply replaced with a goto_w or goto bytecode 
instruction, filling with no-ops.  Direct manipulation of simple binary 
patterns in a .class file--totally ignoring the format minus the 
structure of a like 3 byte code instructions.

Cool, eh?  This should work great!  That said, I do like my bytecode 
assembler ;)

Ter
--
CS Professor & Grad Director, University of San Francisco
Creator, ANTLR Parser Generator, http://www.antlr.org
Cofounder, http://www.jguru.com
Cofounder, http://www.knowspam.net enjoy email again!





 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/antlr-interest/

<*> To unsubscribe from this group, send an email to:
    antlr-interest-unsubscribe at yahoogroups.com

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/