[antlr-interest] philosophy about translation

[Andy Tripp]

Loring Craymer wrote:

> I have off and on put a fair amount of thought into rule-based pattern 
> match-and-replace systems and how they might relate to ANTLR.  ANTLR 
> trees with transformations are like a surgeon's scalpel:  you can 
> locate precisely which parts of a tree need to be changed and change 
> them in situ.

What is the in situ Java replacement for "memcpy(a, b, c);"?
How about "char *p"? Is the replacement a String, an array of chars, a 
LinkedList of Character,
or maybe an integer index into an array?
How about replacing 10 lines of C code that clearly can be replaced by a 
single call to Collections.sort()?

If you're going to have a translation that "works" - has identical 
functionality as the original, the
answers may be straightforward. But if you are going to produce code 
that looks "hand-written"
(whether identical functionality or not), you're
going to have to take a good look around at the code in order to decide 
what should be produced.

If someone's covered these issues in the literature and I just missed 
it, let me know.

Andy