[antlr-interest] Reduction in code base after yanking AST stuff

[Terence Parr]

On Jan 3, 2012, at 5:25 AM, Eric wrote:

> Ter,
> 
>> From my personal perspective I commend the decision. While I know that AST
> rewrites have been a personal work of yours for years, I am sure it took a
> lot to make the move.

It's important to admit when you've been going in the wrong direction, no matter how painful. I'm just embarrassed I didn't see this point earlier. most (but not all) of the applications I've built over the last decade or two would have been easier with listeners

> When looking at the pipeline as Lexer -> Parser -> Tree manipulation (v3,
> AST rewrite vs v4 SAX, DOM, visitor) -> Output (String Template),
> I have had the most trouble with AST rewrite, and routinely resorted to
> doing tree manipulation via calls to the API.  I even notice mention in v4
> of outputting the tree in LISP format using toStringTree(), though I
> haven't tested it yet. While I have tinkered with converting the
> s-expressions to LISP or PROLOG in v3 and then manipulating, that is a
> route I plan to explore more with v4. Also of note here would be
> http://strategoxt.org/

Yeah, those guys are very smart and it's a cool tool.

> As a hint to Sam Harwell, I have considered creating an F# library for the
> tree manipulation phase for the C# target for v4.
> 
> I hope that once users see how to make effective use of the new ways of
> doing tree manipulations, more ideas will come out on how to effectively
> apply the new mechanisms.

Yep, I'm interested to know what people think about rewriting parse trees. certainly that's what most of the European tools do.

Ter